Here is my interview with Dominic Cummings on why Western governments are so dangerously broken, and how to fix them before an even more catastrophic crisis.
Dominic was Chief Advisor to the Prime Minister during COVID, and before that, director of Vote Leave (which masterminded the 2016 Brexit referendum).
(00:00:00) - One day in COVID…
(00:08:26) - Why is government broken?
(00:29:10) - Civil service
(00:38:27) - Opportunity wasted?
(00:49:35) - Rishi Sunak and Number 10 vs 11
(00:55:13) - Cyber, nuclear, bio risks
(01:02:04) - Intelligence & defense agencies
(01:23:32) - Bismarck & Lee Kuan Yew
(01:37:46) - How to fix the government?
(01:56:43) - Taiwan
(02:00:10) - Russia
(02:07:12) - Bismarck’s career as an example of AI (mis)alignment
(02:17:37) - Odyssean education
(00:00:00) - One day in COVID…
Dwarkesh Patel 00:00:00
Today I have the pleasure of speaking with Dominic Cummings. He was the chief advisor to Boris Johnson when he was Prime Minister and before that, he masterminded the Brexit campaign.
Let’s start with talking about your time in Number 10 as a chief advisor. What is the thing that most people don’t understand about being in that government and that famous ministry?
Dominic Cummings 00:00:25
When you go through that door, you’re basically going into a rabbit warren of old townhouses that have been knocked together behind the scenes. It’s nothing like any kind of modern office building. It’s a very, very odd physical environment. And first of all I think you probably would be struck by the constant string of chaos. I don’t think people really appreciate what it’s like being in a building like that.
For example, in one day on COVID, the day starts off with: Are we going to have a lockdown? It then proceeds to the Prime Minister’s girlfriend going crazy about the media. It then involves Trump calling up saying, “We’ve got to go and bomb all these people in Iraq.” It then goes to the deep state coming in and saying, “We don’t think we should because it’s probably going to bomb the wrong people.” And then other parts of the system come in and say, “No, we should bomb them because we’ve got to stay friends with America.” Then there’s some other disaster on the news with something flooding. And it’s just constant.
Obviously, some days are more crazy than others, but if you haven’t been in that environment, it’s extremely hard to appreciate that you have these handful of people trying to come up with the answers to extremely hard problems. The weight of the news flooding in on you just makes it intrinsically difficult. Then on top of that, you have these just incredibly old, centralized bureaucracies actually trying to cope with all of this.
For example, when we arrived in Number 10, there wasn’t even a file sharing system such that a three man setup could at least write a press release on a Google Doc, for example. The British Prime Minister did not have access to such a system.
Dwarkesh Patel 00:02:29
Was it because of security concerns? What was the reason?
Dominic Cummings 00:02:32
A combination of bureaucracy, security concerns and arguments between. The Google Docs thing is actually a sort of interesting thing. So we arrive in summer 2019 and there is no system for file sharing. We say — This is completely insane. We are going to create a system for file sharing. Months of wrangling ensue. Two different parts of the deep state basically argue about whether Teams or Google Docs is more susceptible to China and Russia intercepting. No resolution. This actually ends up affecting COVID. We force a resolution. We get GCHQ to build a system. Still now, four years after we started this discussion, the Cabinet office and Number 10 are still fighting about Google Docs versus Teams, and have now recently resolved to hire some consultants for millions of pounds to spend a year doing a study on it.
So one very small little thing in a way, but probably certainly anybody who’s been at a kind of high functioning company would just be completely stunned by how the core of a G-7 state actually works.
Dwarkesh Patel 00:03:40
Hopefully the consultants are allowed to use Google Docs to decide whether you’re allowed to use Google Docs.
Dominic Cummings 00:03:45
Dwarkesh Patel 00:03:46
You’ve written on your blog “I had these priorities, these main things I wanted to accomplish.” But you have all these things that are coming up day to day. Do these things feel like distractions from the thing you went into the government to do? Do they all feel important? So you have this big picture idea and all these things are coming up. What does that feel like?
Dominic Cummings 00:04:08
A fundamental problem with how the British state works is this question of prioritization and the Prime Minister’s time. So you have all of these normal parts of the system that essentially can’t really do anything quickly at all, even in a crisis. So the Prime Minister’s time and the Prime Minister’s prioritization is the most important asset. But also it’s something which is constantly pulled hither and thither by all of this craziness.
One of the things that obviously we wanted to do was fundamentally reorient Number 10, away from what it’s been since Thatcher, which is a kind of press entertainment service. Where the whole building is just built to respond to what the media says and instead say, “What do we actually think is important?” And what is the management system you’re going to build that actually can maintain focus on those things whilst the inevitable chaos goes on?
Most obviously, for example, some terrorist incident happens in central London. Prime Minister can’t ignore that. It’s going to involve the Prime Minister’s time. Politics, the news, the government means that there are things like that happening all day, every day,
So if you have the normal state of affairs, then the Prime Minister is just pulled from one crisis like that after another. Now, some of them are justifiable. You’ve got to deal with a terrorist crisis. But in fact, what’s happened over the last 20 years is that that’s just been the model for how Prime Ministers spend practically all of their time. And there isn’t a kind of background system with a set of people who are just plodding on, going “Right. The actually important things are science and technology agenda, productivity reform, the Ministry of Defence, Procurement Reform, deep state this, etc.” So that’s a fundamental structural issue.
We tried to deal with that in various ways, but in January 2020, it was clear that this essential question was already a fundamental disagreement between me and Boris. So in the first week of January, Trump had just whacked that Iranian guy. I forgot his name, the head of the revolutionary whatever it was, with the drone strike in Iraq. So we’re talking in that week about the future. So what, three weeks after the election?
And he says, these various journalists are complaining that you don’t return their calls. Now’s the time for us to make friends with all of these people after the chaos of 2019.
And I said, “No, I don’t think that’s right at all. We’ve gone through hell. We’ve won an 80 seat majority. Now we have to change how this building works and we have to focus on the actual big problems facing the country. We’ve got to get Number 10 and your time, the PM’s time, away from how it’s been managed for the last 20 years, which is media entertainment service.”
And his response was, “No, no, no, no. That’s crazy. Everyone will go crazy with us if we do that. We’ve got to make friends with the media. We drove them all mad last year. Now it’s time for a great reconciliation.”
And I said, we could do that, but if we do that, we’ll just be the same as every other government for the last 20 years, which is chasing the bullshit in the media all day. And 4 years will pass and we won’t actually have done anything. And by then it’ll be 14 years of Conservative government. We’ve just won an election, partly by running against the previous decade of the Conservative Party. Can’t pull that off again. We’re actually going to have to do a whole bunch of things. And also, that’s why the people are here. We actually want to solve these problems.
Those issues get to the heart of a lot of it. You have a management question about how you actually manage priorities in such an insane environment. You have a personnel problem about a lack of talent. But you also have this fundamental question of, do the key politicians want to spend their time on the important problems, or do they want to spend their time running around all day dealing with the media? And the answer is, almost all of them want to do the second.
(00:08:26) - Why is government broken?
Dwarkesh Patel 00:08:26
There’s a lot I want to ask about there. It’s interesting because the way you’re describing it, it sounds like it’s a company with the Prime Minister as a CEO who’s dealing with all these operational things that come up, but there is no CEO who has a long term plan.
I’m curious about the media stuff. I want to understand why it’s the case that they’re so obsessed with the media. Does the media actually matter that much to how people perceive the politicians? How is the media interacting with the things that are happening in the building so that people are constantly thinking of the media? Why is the media so ever present in people’s minds?
Dominic Cummings 00:09:02
Just go back to your first point about the kind of CEO analogy. Imagine if Steve Jobs or Tim Cook or Patrick Collison or someone actually spent a large part of their day just doing photo ops as well. That’s actually the reality about how a lot of these jobs have evolved. So you have a person whose time is the single most precious asset, who, because of the dysfunctional bureaucracy, his decisions or her decisions, is the only thing that can actually break down the bureaucratic resistance and make sure something happens. Yet they’re actually standing with the ambassador from Tongazonga, whatever country, just doing photo ops for a large part of the day or going to stupid ceremonies or whatever it might be.
So again, I think that analogy also shows some of the central problems. These people have all grown up in a system where they just don’t know any better than dealing with the media all day. Now, if you actually understand communication, you know that communication is not the same as answering questions to the media, but that’s not what they think, and it’s not the environment in which they operate, and also it’s not the incentives in which they work.
Again, one of the funny conversations I had with Boris was, you know, we should say to the ministers that here’s your actual priorities as defined by us. Whether or not you get promoted and whether or not your career goes well is going to be defined by how well your department actually fulfills these goals. We don’t care about all of your interviews. We don’t care if you are on TV or never on TV. That’s not how we’re going to judge. Because they’ve all grown up in a culture where they think whether or not they’re going to be promoted really depends on: Are they seen as a good media performer? Or do they botch things on the media? Well, that’s just a fundamentally bad criteria, not least because their definitions of what’s good on the media are themselves terrible. By approaching government like that, you’re incentivizing them to think that their goal is making friends with the media. So then they get good interviews. That also incentivizes them to leak everything. So again, the culture and the incentives are self reinforcing in a very negative way.
Dwarkesh Patel 00:11:35
Let’s say the Prime Minister was somebody who agreed with you and said no more interviews. Ministers, you’re going to do your thing. Media is out. How would the media react? Would that matter?
Dominic Cummings 00:11:49
We sort of did that in quarter one of 2020. We actually stopped participating in the number one insider elite media show on BBC Radio. We just said, we think the show is rubbish and it’s a waste of everyone’s time and no ministers are going to be going on it anymore. The media went completely crazy. One of the most senior people in the BBC said to me that I was a fascist. And a lot of the MPs went crazy as well because for them, being on that show is their raison d’etre. So it was extremely unpleasant and disruptive for both the old media and the MPs who’ve grown up in that culture.
Dwarkesh Patel 00:12:30
How much does it have to do with the fact that in the British political system, the people who are running these departments have to be an MP, so they have to be an elected official? Is that why they’re so obsessed with the media? Because it’s like if a congressman had to become the Secretary of State always.
Dominic Cummings 00:12:46
That’s part of it, though I would pick up on your language, which I think is important. Because you say the people who are actually running the departments, of course they’re not running the department. They describe themselves as “I’m running the department”. The media describes them as they’re running the department. But how many people can the Secretary of State for Defense fire? Three people.
Dwarkesh Patel 00:13:20
He cannot fire the others?
Dominic Cummings 00:13:21
He legally cannot fire anybody else in the building.
Dwarkesh Patel 00:13:20
So who can he fire?
Dominic Cummings 00:13:21
The personal three people that he brings in that. He can get rid of those three people. The other hundreds of thousands of people in the system, he can’t get rid of a single one of them. So the person who actually is in charge of personnel at the Ministry of Defence is the Permanent secretary. Stressing the word permanent.
The only person in the British state who has the legal and constitutional ability to say “This senior person in the Ministry of Defence or the Department of Education is clearly failing and they must be removed, and I order that they be removed” is the Prime Minister. That itself is problematic because the large part system will say, “Well, we can’t just dismiss them, or if we do dismiss them, then there’ll be legal repercussions.” But you can actually remove them from the post. There are complications, and those complications have grown in the last few years because of the way that the legal system and various parts of the law have evolved. But the PM can actually remove someone. But only the PM can. So that gives you an idea as well of the tremendous bottlenecks that occur inside Number 10.
I’ll tell you a story about it that kind of summarizes it. At the peak of COVID craziness in March 2020, on the day itself that the PM tested positive for CoVID, a bunch of people come into Number 10 sit around the table and we have a meeting and it’s about supplies of PPE to the NHS.
They say, “None of this ppe that we’ve ordered is going to be here until the summer.”
“But the peak demand is over the next three to four weeks.”
“Sorry, Dominic, but it’s not going to be here.”
“Well, because that’s how long it takes to ship from China.”
“Why are you shipping from China?”
“Well, because that’s what we always do. We ship it from China.”
But A, we need it now and B, all of the airlines are grounded. No one’s flying anything.
“So call up the airlines, tell them that we’re taking their planes, we’re flying all the planes to China, we’re picking up all our shit, we’re bringing it back here. Do that now. Do that today. Send the planes today.”
We did that. But only the Prime Minister could actually cut through all the bureaucracy and say, Ignore these EU rules on Blah. Ignore treasury guidance on Blah. Ignore this. Ignore that. “I am personally saying do this and I will accept full legal responsibility for everything.”
You multiply that kind of problem by hundreds and thousands of problems, you get a sense of partly why COVID was so crazy. This is normal government. But in a crisis, when no part of the system can actually move fast, all of these bottlenecks end up very dramatically escalating to the PM’s office. And if you read Jared Kushner’s book, Memoir about the White House, there are very, very similar tales there. That a lot of things that obviously should have been solved elsewhere couldn’t be solved at any other part of the system. They all end up cascading upwards in these centralized bureaucracies, because ultimately only the president or only the Prime Minister can give certain kinds of orders.
Dwarkesh Patel 00:16:38
What if the Prime Minister said, I am giving a blank check. If I appoint you as minister, whatever you say, I’m going to sign it. So you can rule as if I agree on everything you said. Could they write such a blank check and basically give the minister whatever authority they might want?
Dominic Cummings 00:17:00
To some extent. The system here, in a lots of ways, the Prime Minister has more powers than the president does and is less legally constrained than the president is, because in all kinds of gaps, in lacunae, in the system, there are kind of ancient assumptions that the PM is operating with the authority of the Crown and royal prerogative. So the PM could just do a whole bunch of things, particularly in a crisis, and doesn’t have to necessarily get approval from Parliament.
The most obvious example of what you’re talking about in COVID was on the vaccine task force. I said, with others, in March 2020 that the normal system obviously can’t deal with vaccines. It can’t do anything fast. We’re watching it deal with all of these other logistical problems. We have to create a completely different entity to deal with vaccines. So we essentially appointed someone and the Prime Minister said: Ignore all the EU procurement rules. Ignore all UK procurement rules. Ignore everything. You just build the vaccine task force the way that you want to. We’re freeing you from all the normal Whiteall, HR, all the normal things which add massive friction and mean that nobody can do anything quickly and the simplest things take months and months.
To a large extent, because of the scale of the crisis that happened. But the vaccine task force was still plagued by lots of parts of Whitehall saying, “Well, we don’t want you to do this. We don’t like this.” So the reality is that it depends on the characters involved. It depends on how much the person the PM has empowered will actually push. It depends on the extent to which they come to the PM. It depends on the extent to which the PM then shows everyone in the system that I’m actually supporting them and also fundamentally is prepared to say that people will just be removed from their post if they don’t do what they’re told.
Covid showed how much faster you can do things if the PM’s authority is used and you could prepare to drive things through the system. And the PM is prepared to say, “I will deal with legal issues later.” But it also showed that even in a crisis where literally thousands of people are dying day after day after day, that large parts of the bureaucracy will still simply say “No. We are optimizing for sticking to the old rules.” And that also happened an awful lot.
Dwarkesh Patel 00:19:38
What I find interesting is you’ve written a lot after you got out of the government about all these problems, but there’s other people who have been at a high level in governments in the UK, US, wherever. And it’s weird to me that if there’s this much dysfunction, are they not noticing it? Are they just refusing to talk about it? Have they just never been in a functional environment, maybe in the private sector or something, where they realize how dysfunctional this is? Why is everybody not screaming about this as soon as they get out of the government?
Dominic Cummings 00:20:08
It’s a mix. There’s a lot of people inside the system who don’t actually think of it as being dysfunctional. That’s their life, it’s completely normal for them. So we arrive at Number 10. I’ve taken a bunch of people who are actually experts in data science, AI and blah blah. We talk to a bunch of officials about projects around data, and officials say, “Oh yeah, well, here’s some examples of some interesting projects we’ve done.” You ask them about it. Turns out that they took two and a half years.
You ask “Why did it take two and a half years?”
“Well, the actual project took like eight to twelve weeks. The rest of the two and a half years was the system emailing each other about legal permissions on what to do.”
Now for the people that I brought in from outside, this is obviously insane. No one in any normal functioning environment would spend two and a half years emailing each other to do a project taking eight weeks. But you have to realize that for everyone inside the system, this is completely standard. And a lot of the people who are most senior in the system have been in a thing like that for 30-40 years. They don’t know anything different. And it just seems completely normal to them that this is what would happen.
And in fact, In 2020, for example, when we did some things very differently, it was extremely disruptive and extremely unwelcome to the large part of the system. Hence why a lot of what we did was closed down.
Did they say, “Okay, the vaccine task force and operation warp speed and the state have been great successes. We should massively reinforce them. We should build the next generation of vaccines. We should spread the lessons of how the task force operated.”?
No, they basically closed the task force. Sewage monitoring closed. Rapid testing, basically closed and forgot to order enough tests the following year.
So if you look back at 2020, most of the people who were most wrong were given awards and honors by the system and promoted to new jobs. The people who were most obviously repeatedly right have almost all left.
What incentive is there for people to speak out about how these things work? No one expects anything to change. Even after something as big as COVID, when you see what the reaction is, everyone can now see the truth. You can have a once century pandemic. It can kill tens of thousands of people unnecessarily. It can be a complete carnage for the economy, and everyone will just basically go back to normal. MPs will ignore it and nothing much will change.
So if you’re a standard official inside the system, all the signals to you are very clear. In fact, in 2021, it was even more powerful than that. There were a whole load of legal actions brought to say that the real problem with 2020 was that we went too fast and we did things too quickly. People actually brought legal actions against the vaccine task force. They brought legal actions against rapid testing. They brought legal actions against all sorts of activities.
The system didn’t say “This is completely insane. Actually, the bureaucracy and the sloth killed thousands of voters.” It said “Yes, we’re going to investigate all of this.” Every signal propagated through the system was essentially back to normal. You will be promoted for being the most insane process, and you will be demoted and blacklisted if you say this process is insane and try to do better.
Dwarkesh Patel 00:23:54
There’s two things I’m curious about here.
One, who exactly is the person who decided that the vaccine task force has to go and we can’t continue it.
And two, and this might be related. What happens if a PM decides, I don’t give a shit about these legal challenges. Pile them on if you want. Is it just going to cost a lot in lawyer fees? Why does it matter?
Dominic Cummings 00:24:16
Manhattan Project is much in the news with the Oppenheimer movie. If you look at the very last bit of General Grove’s book on Manhattan Project, he talks about what are the most fundamental principles about why it succeeded? And one of those principles is relevant to government. Actually, they’re all relevant to government.
One of the principles is that the quality of the people is fundamental. Another one is that responsibility and authority are always delegated together. The entire British constitutional system and management structure is based on the fundamentally opposite principle. Responsibility and authority are not delegated together. So if you’re asking about something like the vaccine task force, in the normal system, nobody really is in charge of anything. Lots of people can criticize, lots of people can complain, lots of people can argue about things. Lots of people can veto. Almost nobody ever has the authority just to build something or just to do something.
Why did we create the vaccine task force the way we did? Well, because we were trying to actually embody principles like responsibility and authority pulled together. We brought one person in, we said “You are responsible.” But once we’d gone, then what happens to that entity? It’s sitting there amid Whitehall while all the normal parts of Whitehall just start going back to being normal. So what happens? They say, well, they are exempt from all of these rules on HR that the Cabinet Office imposes on every part of Government. This should change because it’s going back to normal. They have to do the following things properly. We gave them special dispensations because of the extraordinary circumstances of summer 2020, but these now come to an end.
So those sorts of things come in. The treasury says, “The spending rules and how the people in the vaccine task force make decisions, that was an emergency thing. Now the normal rules apply again.” So before you know it, all the different parts of the system have basically said, the thing that you created outside of the normal system now has to obey all of the things that it was specifically created to avoid.
Now, the system will just do that automatically unless there is a very powerful counterforce. Fundamentally, again, only the PM can say,”No, we’re not having that. In fact, I want to strengthen the vaccine task force. We want to move on to the next generation of vaccines. etc.” If they don’t do that, and if the people in charge of it can’t call on the PM’s authority, the system will just devour the new entity very, very quickly and force it to conform with all of the normal system.
I’ll give you another example of this on rapid testing. One of the things that we did to get the rapid testing to work was we got a guy who formerly was commanding officer of the SAS, British Special Forces, and this guy got a bunch of his friends from Special Forces also to work on rapid testing. When we first got this pushing from Number 10, I got the critical people from procurement, commercial HR, etc, into the Cabinet room with the Cabinet Secretary, the single most important official in the whole country, and the two of us said, “The PM wants rapid testing dealt with as if this is a wartime crisis.” We’re going to have a second wave. There’s going to be thousands more people getting CoViD, there’s NHS. People are dying, etc. We can’t have any of the normal civil service HR. We can’t have any of the normal civil service bullshit on procurement. Exactly the same as with the vaccine task force. Everyone sits around the cabinet table, they all nod their heads.
A week later, I call this guy, a former SAS boss and say, “So, how’s it going? Are you getting who you want and is everything working great?”
He says, “No, it’s all the same shit show.”
So I have to get all the people back in the same room with the country’s most senior official and say, who the fuck have we got to fire around here to make clear that these people doing testing don’t have to do all of your bullshit HR?
That’s how extreme things have to be. It was only by doing that a second time and making clear that I would get the PM to actually just start firing senior people in the Cabinet office. It’s only then that the system will kind of part and go, “Okay, this element is allowed to.” But you imagine as soon as that countervailing force is removed, all the normal sea floods back.
(00:29:10) - Civil service
Dwarkesh Patel 00:29:10
At some point we should talk about what it would actually take to change the equilibrium. It might be first useful to talk about how did this come to be? So presumably, at some point, this is not the way the government functioned. How did it end up like this? Is it just that the cludge builds up? Tell me about the actual mechanism of how does it end up this way?
Dominic Cummings 00:29:29
In Britain it started in the 1850s when people said, well, the old aristocratic system based on patronage is irrational. We’ve got a shift towards a much more meritocratic system. We should have officials that are appointed on merit. You essentially had a transition from a pre-1850s aristocratic system based on patronage, where there was all kinds of dodgy corruption in various ways, and favors done in various ways, but it also moved much faster and in all sorts of ways much more efficiently, particularly in a crisis. From that, you shifted towards what you say is like a modern system where you have permanent civil service, supposedly meritocratic.
However, over time, you see that this supposedly meritocratic system ends up actually just being a closed caste. It’s not actually meritocratic. It’s a system that promotes practically 100% internally and is therefore, by definition, closed to approximately 100% of the world’s most talented people. Now, that’s defended on constitutional grounds now, as: this is the only civilized, sensible way in which a state can operate. But of course, it’s totally self serving. And as we saw in COVID, it actually just means — massive bureaucracy, very poor people in all kinds of critical positions, and a state that’s actually paralyzed when it comes to the crunch.
Dwarkesh Patel 00:31:06
It’s like a Japanese company where you get in after college and just stay until you die. I live in San Francisco, and if somebody works in a company for three years or five years, they’re, “Oh, my gosh, you’ve been here for a while. You’re like a long term employee.” So the idea that you would be in the same place for 40 years is really odd.
Dominic Cummings 00:31:25
Sorry to interrupt, but imagine as well what the promotion system is like and who ends up getting to the top of these systems. A lot of people say, “Oh, you’re so negative about the civil service. You’re all saying that everyone there is rubbish, and it’s not fair.” That’s not my view. In fact, if you look at the civil service, you actually see a lot of very able people, but most of them are young. What happens is the young, excellent people get weeded out by self-selection, largely because they go in idealistic, they’re there for a few years, but then they look at what the process is to be promoted, and they look at their bosses, and the best of them look at it and go, “I don’t want to be like that.” I don’t want to have to make those decisions. I don’t want to have to make those compromises. I don’t want the job like that, where it’s almost all bullshit. We can’t actually build anything.
The most entrepreneurial, the kind of people who actually want to get on and do stuff now, leave and the most HR compliant, disastrous people to be in charge of supposedly fast moving agencies are the ones who are promoted to take over. And then that culture itself becomes highly self reinforcing. Once you get a whole cadre of leadership at the top that’s like that, it’s extremely difficult to break out of.
It’s not like “Well, okay. IBM has gone wrong. But you still have a whole ecosystem of competition for IBM.” Like sure, IBM dies, but the American economy evolves and continues. In government systems that obviously can’t happen. So you have — IBM is dead, but it’s still actually in charge of everything.
Dwarkesh Patel 00:33:03
This reminds me of this wonderful biography of LBJ that Robert Caro wrote. In volume three, which is about him taking over the Senate, when he gets in the Senate the seniority system is known as a senility system because you have these people who are in charge of these important committees and they’re 80 years old or something and they’re in charge of ways and means. So when he becomes a majority leader, he realizes that he can’t have this if he wants to get things done. He realizes he needs the young senators in charge and so he engineers ways to get 40 year olds or 50 year olds in charge of these committees.
Could a Prime Minister just retire a whole bunch of people and say we’re going to put 20 year olds on the path to management or path to advancement. Who would that be up to?
Dominic Cummings 00:33:50
In the British system only the PM can do that and we actually started to do that in summer 2020. After the first wave of COVID when the PM nearly dies in March, April, for about eight weeks or so, he was highly aggressive in saying, “This whole fucking disaster system has got to change. It nearly killed me. It’s killed tens of thousands of other people. We’ve got to blow the whole thing up and rebuild it.” So in summer, we actually got rid of roughly half a dozen of the senior permanent secretaries in charge of a lot of these departments.
It was described by the insiders as a “rolling coup”. Fascism. The Vote Leave fascist entity is now essentially mounting a coup, Orban style, to dismantle all democratic. So that was the reaction of large parts of the old system to saying half a dozen duffers are going to be removed and we’re going to start promoting some younger, more dynamic people and that’s part of an overall shift that needs to happen. Clearly, a whole bunch of different parts of the system have failed. Agencies need to be shut, new things need to be created.
So it is possible to do. But the tale of that also shows the problems, the reaction from the old system and the media is very severe. Nobody supported us at all, including all the Conservatives. Ironically, the people who were most in favor of it was actually the deep state itself and the most entrepreneurial elements of the deep state because they’d all seen this system completely implode. So the closer you were to Number 10 and the closer you were to actual power, the more you actually appreciated the full horror story of how the apex of the British state had completely imploded.
A lot of them, particularly the younger ones, said, well, now it’s time you’ve had a once century pandemic, the biggest crisis the state Britain has faced since Hitler in 1945, and you’ve got Brexit done and you’ve got the Vote leave team in Number 10 who’ve said before they arrived here they’re going to change civil service. Okay, well, now, obviously this is all going to happen. And it should happen. Now we can argue about the details of it.
Clearly, the personnel system needs opening up. The civil service is going to have to have the biggest shakeup that it’s had in a century or more. Fundamental problem was that the PM then thought that this is going to annoy so many people that I can’t face it. Also other dynamics, which I’m not sure how much you want to go into, but the fact that Starmer is just so obviously rubbish also meant that Boris and his wife thought, we don’t actually have to do very much. So it goes back to this question of are you actually trying to change a lot or not?
Our response to summer 2020 was, we’ve said for years that a whole bunch of things are wrong and that the next big crisis will reveal it. Well, we’ve had the next big crisis and it has revealed it, and it’s obvious that we’re right. Now, we’ve obviously got to make all of these changes. And by the way, in the election six months ago, we also promised we were going to make all of these changes. Boris, Kerry and the majority of the Conservative Party’s attitude was far more “Keir Starmer’s rubbish. Changing anything big will just be very disruptive and create lots of enemies. Why are all of these people talking about disruption and firing people and rebuilding things? We just want to go back to normal. We hate COVID. We don’t want all these arguments about Brexit. We kind of want to go back to normal.”
And for them, what normal is, is like the 90s. The kind of default mode for how people in politics think about normality is, can we go back to that lovely time between the fall of the wall in 1989 and the fall of the towers. That period is what people kind of psychologically gravitate back to.
So in summer 2020, our attitude was that the state needs to be fundamentally reengineered but most of the systems’ was — We’re tired. We just want to go back to normal. We want to do what we always do, which is: chat with each other, give interviews to the media, and not change much. And given that Starmer is so obviously rubbish on the other side, we don’t actually have to change very much either.
(00:38:27) - Opportunity wasted?
Dwarkesh Patel 00:38:27
On your blog, when you’ve written about all these other reformers throughout history, they were able to make their changes in a time of crisis. Lee Kuan Yew after the British leaves or Bismarck after Prussia loses the Napoleonic wars, In this case, you have COVID, you have Brexit, you have an 80 seat majority that you’ve engineered based on campaigning on these things. And you have you as chief advisor to the PM, somebody who understands these things, has thought a lot about them and, in fact, accepted the job on the condition that you would be able to do these things.
You have all these things come together. Why was the opportunity wasted? You mentioned, for example, that Boris and Kerry were not interested in changing these things. Presumably they saw that the problem was there. I mean, if you’re a PM you must see the craziness. And also, wasn’t he a fan of Churchill and he wanted to build a legacy or something? I would think if I was PM considering that most PMs are going to be forgotten, you might as well just go forward and try to do something big, and try to be remembered.
Why did they waste this opportunity? You have all these things that were going for you guys. If something was going to happen, that seemed like the perfect moment to make something happen.
Dominic Cummings 00:39:53
Part of it goes back to what I said before, that you have two fundamentally different attitudes towards the whole thing. Boris was prepared to be very aggressive and revolutionary in various ways in 2019, fundamentally because he thought it was necessary for him to survive as Prime Minister, not for the country, which he didn’t and doesn’t care about, but for himself.
In January 2020, we were already arguing about what the fundamental direction is that we’re going to go in. As I said before, my view was: We’ve won. We won saying that we’re going to do all these things. We now have to actually do all of these things. And in order to do all of these object level things, it requires facing these multi-decade long term, some of them over a century, in terms of how the actual constitutional system works, how the civil service works and everything else, versus his view, which is that it all just sounds like a lot of hard work and making a lot of enemies and it doesn’t really seem necessary because the Labour Party is a joke and I want to be friends with London. I want to be friends with insiders. I want to have a nice time.
The overwhelming majority of people in politics fundamentally prioritize social relations within the insider network. That is just what totally dominates their life day in, day out, year in, year out. Now in some ways it’s very odd. Because what’s the point of doing this for 30 years and then you kind of become a minister, but you’re not actually in charge of anything and you can’t actually do anything and then you’re spat out and you’re not even going to be even a footnote in history. What on earth is the point of it? It kind of seems completely crazy to me anyway.
I think that misunderstands how they see it. For them, the theatre and appearing on TV is the critical part of it. It’s not — Have you actually done anything? And the permanent civil service are brilliant at manipulating the theater to keep the egos of the MPs satisfied. The most obvious way to think about it is the Cabinet. You have the most famous door in the world, Number 10. Every week these characters walk up the street. The cameras are all there. The cameras work. They click, they shout out questions. The ministers smile at the cameras and they think “We’re part of the insider gap.” This is the peak thing that everyone aims at whilst behind the door.
The people actually running most of the government regard the Cabinet as just kind of bullshit theater. The meetings are literally scripted. People are given their talking points to read out by the officials. They read them out, they read those out and the conclusions are all pre-written. So the whole thing is a complete potemkin farce. But from the MP’s point of view, we’re on TV, we’re treated like we’re in charge. The fact that the people actually running the Ministry of Defense is not the Secretary of State. The fact that these ministers have fundamentally no power. The fact that in Number 10, officials who are like 28 years old working 5 meters away from the PM have usually far more power and authority over things than the ministers on TV do. But their names are never in the papers. No one knows who they are. This weird mismatch is never explored. It’s never covered.
I had a funny conversation the other day. I was talking to one of the editors of one of the biggest newspapers in the country, and I said, “In the old days, there used to be a parliament page in newspapers that would report what happens in Parliament.” Because that’s the center of political activity and what happens in Parliament is important. “Why don’t you start a page where you actually report on the deep state?” and you say, “Oh, the Prime Minister’s secretary for Economic affairs has moved from this job to that job and it’s now been replaced by a 31 year old.” because those jobs actually are far more important.
If you look at Brexit, the foreign secretary was almost completely irrelevant to all of the Brexit negotiations. The Prime Minister’s private secretary, a very able character, much more able than anyone in the current cabinet, but was totally unknown in the media. Name never in the media but far more important. So I said to this editor, “Why don’t you have a page and fix that?” And he said, “Well, everyone would think I’d lost my mind if I did that.” It would antagonize everyone. The MPs would go completely crazy because preserving this fiction, the potemkin charade of the MPs trotting up Downing street and being filmed going into the Cabinet which is taking decisions. As long as the potemkin carries on. Just don’t break the kayfabe, as they say in WWE.
Dwarkesh Patel 00:45:02
It sounds like you have two constitutional monarchies in Britain.
Dominic Cummings 00:45:06
In some ways, yes. Exactly.
Dwarkesh Patel 00:45:08
You say that you have these undersecretaries, who are these 31 year olds or 28 year olds and they’re the ones in charge. Doesn’t this contradict the idea that you have this gerontocracy of people who had to wait 40 years to get promoted to be in charge of these departments, this permanent civil service?
Dominic Cummings 00:45:27
So it’s an oddity in the system and it’s purely around the PM. The way the system works is in all the different departments, the gerontocracy is in charge and you have this appalling HR system that filters out almost all of the best people and promotes the most HR friendly types.
But the PM’s office is, for historical reasons, slightly odd. In the PM’s office, you have roughly a dozen or so traditionally young, bright people who are put into that job to do specific things as part of training them for the future. It’s just an oddity of how the system works that you have this one set of people who are formally not senior, these people don’t have very senior roles in the hierarchy of the caste system, but they have a lot of kind of unofficial power and authority because they are talking to the PM every day and they’re part of the PM’s private office. So it’s a weird way in which the system works.
Dwarkesh Patel 00:46:31
By the way, do you want to mention the able 31-year old able person who helped with Brexit?
Dominic Cummings 00:46:38
Well, there’s a few. IThere’s a guy called Jono Evans who did Brexit negotiations. There is a brilliant young woman called Hannah who coordinated deep state stuff with Mi5 and things like that. Whenever there’s a terrorist attack, if a plane is going to fly into parliament and try to kill everyone and it’s got to be shot down, then she’s the cog in the wheel that actually organizes the system to get the right call. So there’s a lot of people like that who are extremely capable.
The joke amongst the Vote Leave team used to be that if you swapped the private secretaries for the Cabinet in every single job, you would basically improve the caliber of the person by like 10X. However, lots of those people have now left.
Dwarkesh Patel 00:47:23
Because of self selection?
Dominic Cummings 00:47:25
Dwarkesh Patel 00:47:25
It’s interesting, as a foreigner, when I think about the UK and I’m just learning the basics of what the problems in the country are, the main thing is your productivity growth since 2005 has been abysmal. You guys are much poorer than you guys could otherwise be. This is the main thing that jumps out in my mind. Is there a team in Number 10? Is it anybody’s job who’s very senior to deal with this specific problem?
Dominic Cummings 00:47:52
Now, not really. Remember, as I said before, how small Number 10 is which in lots of ways is a good thing. Everything to do with economic policy growth, etc. is all, like, 99% of everything to do with all of that is elsewhere, mostly in the treasury, which is actually anti-growth in most important ways.
Now, in 2020, we started a whole bunch of things to actually make it a core part and shake up the way that the system worked. We changed how the power relations between Number 10 and the treasury, also with the Cabinet Office, and we actually created task forces to start working on lots of these most critical problems. So the science, technology, whether it’s the startup ecosystem, procurement, the planning system, which is one of the really big things that destroys so much value and there’s just such incredible low hanging fruit.
But basically, all of that was dismantled by Boris in 2021 after we left, because, for the reasons I already said, he came to the conclusion and most of the party agreed. Remember in 2020, when it came out that we were working on all of this stuff on growth and taxes and the planning system, The Daily Telegraph which is screaming about growth today was totally hostile. Most of the Tory MPs, totally hostile. Most of the Conservative Party were not agitating for an aggressive pro-growth agenda in 2020. They were actually hostile to it.
(00:49:35) - Rishi Sunak and Number 10 vs 11
Dwarkesh Patel 00:49:35
Speaking of the treasury, you helped promote Sunak to Chancellor of the Exchequer. He was, from what I understand, relatively unknown before that.
What did you notice about him that made you decide to do that? And maybe tell me more about why you think the treasury is anti growth.
Dominic Cummings 00:49:55
So Sunak, he’s obviously much brighter than the normal MP. He’d actually worked in functional private sector organizations before coming into government unlike most of the MPs. He was extremely hardworking, unlike most MPs and unlike most ministers. He actually dug through the detail and wanted to understand it and could understand it and did understand it. And he seemed much more interested than the normal MP in actually doing useful work, rather than running around the media and giving interviews and all the normal bullshit that they’re all obsessed with.
So, from our point of view, we had a chancellor then at the time, who just couldn’t do the job at all. He couldn’t manage the Department. He didn’t have the self confidence to grip the officials in the treasury, so it seemed just like an obvious change. Sunak didn’t understand politics very well. There were a lot of things he didn’t understand but from our point of view, that didn’t really matter because we didn’t want him to do the politics side of it. We wanted someone in there who also wanted to work with Number 10, right? Which is a critical thing.
One of the big structural problems in how the British state has worked is that there’s a kind of structural conflict between Number 10 and Number 11, the treasury. You normally have a Prime Minister in Number 10 that’s always wanting to spend money, and then you have a Number 11, which says we’ve got to control Number 10. But Number 11 has, in lots of ways, much more power over Whitehall than Number 10 does, because it legally signs off the cheques not Number 10 and that has lots of weird consequences.
One of the most stupid things is that Number 11, the treasury, basically hides huge amounts of financial information from Number 10, which is obviously completely insane. How can you have a Prime Minister making judgments about all sorts of things if his own treasury is hiding data? Insane. Makes no sense whatsoever, but it’s completely normal in Britain.
We said, “Fuck that. We’re going to have one team between Number 10 and Number 11.” The data will all be completely transparent. So exactly what the spreadsheets are in the Chancellor’s office are instantly completely accessible and open to the Prime Minister’s Office. Not something you’d think that ought to be very controversial. But it was highly controversial and unprecedented. And it changed literally within like 3 hours of me leaving. The Treasury went and stopped it straight away.
So our thought was, if you’re actually going to get to grips with growth and productivity and all these things, you have to bring the Number 10 and 11 systems together. They have to have an integrated diagnosis of what the problems are and an integrated plan for what they’re going to do to change it and then actually one team of people. And then you stop Whitehall also picking you off and using the normal divisions.
Dwarkesh Patel 00:52:48
What do you think of Sunak now? You noticed he was intelligent, hard working and paid attention to details. Is that enough to make a good PM?
Dominic Cummings 00:52:56
He’s trying to make the old system work and he’s treating the old system with respect. It’s political disintegration. He doesn’t control the government. He doesn’t even control Number 10, the Cabinet Office. He has no political story. He has no message. He has no grip. And he’s just buffeted hither and thither by events in the same way that every PM has been since Thatcher.
He had a fundamental choice, “Am I going to try and win?” Which involves challenging the way that the conventional system works, the conventional power system of Whitehall, and am I going to tell a story to the country which is convincing, or am I going to do what all the insiders are telling me to do and keep them happy? He chose to do B, with the inevitable consequence.
Dwarkesh Patel 00:53:56
I’m more curious about the broader lessons about personality that this reveals than him necessarily. In the American context for example, FDR clearly had control of the government, and as far as I understand, wasn’t somebody who was known for being intelligent or micromanaging details.
Dominic Cummings 00:54:15
But he had Harry Hopkins to do that for him.
Dwarkesh Patel 00:54:16
Right. So, what does that reveal about who makes a good PM? What are the characteristics needed for a person who can control the government?
Dominic Cummings 00:54:25
It’s impossible to be a good PM if you accept the way that the job currently works. It’s totally impossible because of what we’ve been talking about. You’re just buffeted by media events all day, and you don’t actually control the government. You can’t actually make anything happen. If you accept all the constraints and you accept the way that Whitehall works, it’s impossible for anybody. It doesn’t matter. You could put General Groves into that job. If he did it the way that Sunak’s doing it, General Groves would fail, too. You could put in FDR. You could put in Bismarck. The reality is nobody who actually gets a lot of things done historically operates the way that the current Prime Minister is forced to operate by the prevailing system. They are fundamentally incompatible things.
(00:55:13) - Cyber, nuclear, bio risks
Dwarkesh Patel 00:55:13
This is the same government that has nukes, that deals with biosecurity, counterterrorism, and all kinds of other things that I’m sure I’m not even aware of.
If the general government is this dysfunctional, are the people who are in charge of the nukes just as dysfunctional as those not having Google Docs and taking two years of litigation to do a two week project?
Dominic Cummings 00:55:42
It’s worse in a lot of ways. I saw recently that Peter Thiel said that we don’t see much about how the NSA works and his assumption is that in lots of ways, the NSA is worse managed than the DMV.
There is, unfortunately, a lot of truth to that. But the position is mixed there in the same way it’s mixed generally. You can’t just say it’s all a shit show and the people are all rubbish. In the world of intelligence services and special forces and things like that, there are obviously a lot of incredibly able people and incredibly public spirited people, people who make huge sacrifices believing in what they’re doing.
But it’s also simultaneously the case that a lot of the very worst, most appalling aspects of the bureaucracy happen in that world. And part of it, obviously, is that they can classify things and use classification to hide extraordinary public disasters.
For example, the situation in terms of China’s infiltration of critical infrastructure and data systems in Britain is much, much worse than practically all MPs have any comprehension of. I’ve been in meetings where these things have been discussed and the now PM, then Chancellor, have sat literally with their mouths wide agog at the extraordinary tales that they’ve been told.
“What the fuck? Are you kidding me?”
The number of MPs who know that is probably like a handful at most, and it’s almost all completely hidden.
Similarly, on the nuclear side, I spent a lot of time in 2020 in bunkers without phones, talking to officials about the state of the nuclear enterprise, weapons safety infrastructure. And the truth is absolutely horrific there as well. And it’s horrific because for year after year and administration after administration, they haven’t faced hard problems. They’ve punted off.
So you have a combination of things. You have normal catastrophic procurement, which just means it’s totally normal for everything to be fucked up. Well, that also applies to the nuclear enterprise. You also then classify a lot of that so that it’s hidden. And that means it’s even easier for things to keep going for longer. It also means that the budget problems are hidden. A lot of what happens in terms of the public discussion about M.O.D budgets and the national accounts in general is massively distorted by the fact that in reality, you have literally tens of billions of pounds that are going to have to be spent on the nuclear weapons infrastructure that don’t appear in the official accounts at all. Simultaneously, you have parts of that infrastructure that just don’t work properly. Appalling safety that’s been neglected for year after year. So that’s cyber and nuclear.
Speaking of Bio, I organized a meeting on biosecurity in summer 2020 as well, given that at the time there was COVID, and we were thinking, “Is it a lab leak? Is it not? What’s the truth about all of this?” So we organized a meeting and asked various questions. I didn’t say that one of the people that I actually took to the meeting was themselves a brilliant young scientist who’d been working in the States in the Janelia lab on neuroscience. And so all these people inside the system said, “Don’t worry about this, Dominic. This will never happen. This is impossible. This is science fiction. This is ten years away. Blah, blah, blah.” And everything was about trying to reassure me that I shouldn’t really worry about this. At the end of the meeting, I asked, “So James, what do you think about this?” Of course, these people have no idea who he was. And his answer was that pretty much everything that everyone has said is impossible or will take ten years, I have personally done in the lab in the last two or three years.
Now, does that mean that the whole system for biosecurity is a disaster? No. Does it mean that everyone involved in it is a nightmare? No. There are obviously brilliant people everywhere. But across all of these things, there are budget horrors. There is a chronic inability to build long term. There are constant bureaucratic incentives to not face reality and the truth. And that’s the case across all of these secret systems.
Dwarkesh Patel 01:00:50
Why hasn’t there been a disaster? In many countries, the systems are as much of a shit show. And I mean, this is the West. Russia has nukes. Pakistan has nukes. And you can only imagine how fucked up their systems are. What has prevented it?
You could say that the system is so fucked up that actually there was a lab leak and that was COVID. And so maybe there already has been a disaster. But what is the explanation for why other parts of the system haven’t crumbled in a disastrous way yet?
Dominic Cummings 01:01:21
If you look at just the public record on nuclear stuff, then I think that the only reasonable conclusion is that we’ve got extraordinarily lucky through the Cold War. Whether it’s the famous hydrogen bomb falling out of the plane and all of the safety devices apart from one failing. I mean, America nearly nuked itself, right? It was just completely by the grace of God that that didn’t go off.
We’ve been very lucky so far, and there’s no reason to expect that luck to continue. And if you look at what’s happening in Ukraine now, then you can see that large parts of the system are very happy to dance right on the edge of the abyss.
Dwarkesh Patel 01:02:05
Is there more you can say on the Chinese infrastructure stuff? I don’t know if you can but I’m very curious.
(01:02:04) - Intelligence & defense agencies
Dominic Cummings 01:02:04
If you imagine having a Sci-Fi novel and you said, what are a whole set of data systems that you really would not want the British state or the American state to be transferring data about? And then imagine that it turned out that these things are controlled by and owned by Chinese intelligence. That gives you a kind of nightmare picture. That is the reality. I can’t go into the specifics of it, because I think the specifics of it are illegal to discuss. But if you just wrote a story and imagined what are some of the most obvious ways in which things can go wrong?
In fact, you probably wouldn’t write such a story because you’d think that’s completely implausible. There’s no way that that would happen. There’s no way that they would transfer data between A and B about this information and then find out later that that is actually controlled by China. That would be fucking mental.
Dwarkesh Patel 01:03:05
How about the intelligence agencies? MI5, MI6, and then in America, the CIA. How much situational awareness do they have about the most important things? Honestly, if I was like the PM, maybe in my daily briefing, the top thing I would want is the training loss on the newest AI models or things like that. But on the things that matter, how much situational awareness do the intelligence?
Dominic Cummings 01:03:27
That’s an interesting example. I think you have to draw a huge contrast between two things, capabilities and analysis. There are some extraordinary capabilities that deep state has in the Western world. If you want to dig into people’s phones, if you want to acquire secret information in various ways, then there are some extraordinary capabilities which people have and can be aimed. However, in Britain, they are generally not aimed nearly as aggressively as they ought to be. The process for prioritization of things is extraordinarily awful. Again, that’s another thing that if you actually wrote it out, no MPs really know anything about it, but if you looked at the system, people will just be completely appalled.
Also, parenthetically, a lot of that kind of stuff has basically been shifted away from politicians over the last ten years. Nobody really has any visibility on large parts of that system at all now. When I dug into these things inside the Cabinet Office, I was essentially the only political person in a long time who’d actually even been discussing it with parts of the system so the officials themselves said to me.
Because it’s been pulled into the Cabinet Office away from the Ministry of Defence, away from the Foreign Office and away from the Home Office, the three parts of the system that legally in the past had a lot of oversight over what was happening. Now a lot of it happens inside the Cabinet Office where there’s essentially zero political oversight of any kind. To a large extent, that’s very bad. It means that the bureaucracy metastasizes and a lot of decisions are made without any real challenge.
So that’s bad. But going back to the main thing, amazing capabilities in various ways but badly focused, badly prioritized. But also the quality of analysis is much, much worse. In trying to analyze what people might do, how they’ll behave in a negotiation, will they start a war if they do? Those sort of questions. A lot of the work there is poor.
Also I think it’s crucial to bear in mind that there just hasn’t been a focus on AI. You and I know lots of people who’ve been watching the world of AI now since at least 2014-ish when Deepmind made a big splash, or the couple of years before that. But even in 2019-20, I went into Number 10 wearing an OpenAI t-shirt on the first day to try and make a point to people that people should be paying attention to this. In 2019-20 that was seen as extremely eccentric, never mind five years earlier.
And you’re saying, well, if you were president or Prime Minister, I’d be aiming these amazing capabilities and saying, “I assume that GCHQ and the NSA know who is running the equivalent of OpenAI in Beijing. Who is running the black projects for Chinese intelligence on training runs? Where is the black project data center and who the fuck is running it? And where is it? In which mountain and how many spy satellites are looking at it?”
Dwarkesh Patel 01:07:04
If the Prime Minister said, “I want to know how many H100s NvIdia will ship out next year. Go to TSMC, go to Taiwan, go wherever. Find out how many China is ordering. What university or state company in China is ordering and how much?”
Does the capability exist to get that information? Or is it just like nobody cares about it?
Dominic Cummings 01:07:26
The main problem is no one cares about it. But if you turned MI6 and GCHQ onto a question like — What exactly is China doing with AI? Who are the key people? How can we honey trap them and blackmail them? Then obviously you could get an extraordinary amount of interesting information.
But A, the system won’t do that by itself. B, the politicians won’t tell them to do it because the politicians aren’t really interested in it. C, there’s an incredible kind of risk aversion now in large parts of these systems, particularly after what happened with Iraq and then the legal investigations post Iraq and terrorism and whatnot. So there’s a huge kind of self censoring in large parts of the system and much less aggression than most people would assume.
Dwarkesh Patel 01:08:17
How about defense? In Ukraine, we’re seeing very asymmetric returns on different kinds of new weapons where you have cheap drones that are taking out expensive Russian tanks. How competent is defense generally, and how much are they adopting these new technologies?
Dominic Cummings 01:08:32
Generally in Britain, it’s completely shocking. Again, in 2019-20, we had arguments about this with Number 10 and the MoD. Again, it goes back to this generational thing. There’s a lot of younger people inside the MoD and obviously a lot of people in special forces who are looking at the sharp end of this and saying:
Obviously, drones are going to completely change how land war operates.
Obviously, we should be pushing this and exploring it ourselves.
Obviously, we should be thinking about how to get large numbers of relatively cheap things leveraging commercial technology, and then think about how to add deep state capabilities on top of that, etc.
But you also had a whole set of the senior people thinking, as they always do, “Fuck this. This sounds like it’s going to cannibalize our budgets. Oh, we’ve already got these drone programs.”
Yeah, but a lot of your drone programs are complete dog shit. The drones fall out of the sky and they don’t work and they’re massively expensive and they should be closed down.
There’s a drone now that Britain is trying to deploy in Ukraine called Watchkeeper. In private meetings in 2020, the MoD admitted the whole program was a complete disaster. They keep dropping out of the sky. They’re completely shit, and the whole project should be closed down.
Of course, once we left and the system went back to normal, then they didn’t close it down, they haven’t replaced it, they’ve just thrown more money at it. And lo and behold, the things fall out the sky and fuck up.
In 2019-20 a lot of these arguments were quite theoretical.There’s some things happening on the fringes. There’s the war in Armenia and whatnot, where you could kind of see some beginnings of people experimenting with some of these things. Literally, in 2019-20, people were saying to me, “Well, Dominic, our future fighter in 2040 or 2045 is still going to be manned, and there are all kinds of classified studies that show that drones are not going to be able to do this, that and the other.”
“Really? Well, let’s open up these studies. Let’s see what OpenAI and DeepMind make of these so-called studies.” Of course, it all turned out to be total bullshit. Britain is still going ahead with that, though. And our current plan is to build another fighter in the same way, optionally manned, with BAE. Completely ludicrous. So there’s huge resistance inside the system to making that kind of shift for all of the normal reasons.
Dwarkesh Patel 01:10:59
The reason I’m especially curious about this is we’re seeing how this war in Ukraine is happening. I’m very curious about what this implies if there was a conflict in Taiwan.
How easily could you take out an American or British aircraft carrier? Having seen the insight of how these things work, how confident are you in these war games and these projections? If there was an actual conflict in which Britain had to get involved, what do you think would happen?
Dominic Cummings 01:11:23
Well, in war games, if they’re dealing with any kind of serious peer opponent the British aircraft carriers immediately flee to the edge of the war game so that they don’t get sunk immediately. And the aircraft carriers are obviously a joke. I said that for the first time ten years ago. Five years ago, I had a lot of meetings about it in Number 10. Nothing persuaded me of anything other than that the whole thing was a sort of massive waste of money. And that’s just becoming more and more obvious.
All over these systems in the west, people are now starting to face the music that a lot of these things that they’ve invested billions in are just totally vulnerable to asymmetric technology and asymmetric costs, where very cheap systems are going to be able to destroy platforms worth multiple billions. A lot of senior people who’ve talked a lot of shit about it obviously can’t admit any of these things publicly, but I think there is growing realization behind the scenes of what the reality is.
But the situation in Ukraine should make us even more pessimistic in various ways. Because if you’d said in advance — Okay. Putin’s going to invade Ukraine. We are going to simultaneously encourage Ukraine to fight. Arm them and push Ukraine into a war of attrition. Simultaneously we’re going to ramp up, we’re going to ditch One China policy and aggressively ratchet up diplomacy against China on Taiwan and push the world’s biggest manufacturer into a closer relationship with Russia, whilst fighting a war of attrition against Russia, and then we are not going to actually have a defense industrial plan and a procurement change ourselves in Europe.
“Even for you, Dominic, that would be just too ludicrous a scenario to happen. Nobody in their right minds would actually get themselves into that situation.”
But that’s literally what we’ve done. We’ve pushed the world’s biggest manufacturer into alliance with Russia. We have escalated a war of attrition. And we have left a completely rotten dysfunctional procurement and manufacturing system for defense continue. The same set of people in charge of the decisions in the MoD and the Pentagon. Same set of bullshit pushed out about the advantages of aircraft carriers and blah, blah, blah the whole thing makes absolutely no sense at all
Dwarkesh Patel 01:14:13
What happens when we have to face the music and when does that happen?
I don’t know. Maybe there’s a conflict and many personnel on the aircraft carrier drown. Or maybe not even with defense specifically. Just in general, all these different parts of the government are not only dysfunctional but are getting more and more dysfunctional.
Is it just going to be a slow degradation or will there be another very clear inflection point like you had with COVID?
Dominic Cummings 01:14:42
My assumption is that there’ll just be repeats of the COVID experience and lots of them worse. There’s roughly some kind of financial crisis every decade or so. Another financial crisis could easily be worse than the last one. It wouldn’t surprise me if within the next two years there’s a worse one than the 2008 crash. Perfectly plausible. Lots of hedge fund people I talked to are kind of planning on that basis and think that something like that is quite likely.
We’re going to face the music shortly in Russia because the Ukrainian offensive is not going to be the great success that we’ve all been told. Remember last Christmas, 9 months ago, the British and American media were full of: Russia’s about to run out of ammunition. The Ukrainians are going to launch a counter offensive. They’re going to sweep all before them. The idiot Russians are all going to collapse. Well, Chickens are coming home to roost on that now, that’s going to get worse.
From China’s point of view it’s the most perfect situation imaginable because they can charge Russia inflated prices to sell them a bunch of stuff to blow up all of our shit turning Ukraine into rubble. You don’t have to be Sun Tzu to figure out that from China’s point of view this is like an absolute dream scenario and if America was insane enough to get into a war over Taiwan then obviously a lot of this would be exposed.
There’s a book by a guy who now works at Anduril called the Kill Chain. The truth is that a lot of classified Pentagon stuff has said this. A lot of people involved in the system know that if America has a war over Taiwan then it’s going to be catastrophic for America. Leaving aside the obvious risks of escalation to nuclear conflict like. Just leaving that aside, just on a conventional basis, it will be a catastrophe for America.
Dwarkesh Patel 01:16:35
In a hundred years time, I guess it’ll be 2123, what about what the government does now will matter? When you read back in history about 1923, what has mattered most? What should be the big priorities?
Dominic Cummings 01:16:50
If you run that experiment on ourselves and you look back, what is it that we care most about now? You care about people who come up with new ideas, which is not something the governments do. People are interested in what Nietzsche said, but the details of what the British government did in the 1870s and 1880s is almost totally forgotten and not really relevant.
The thing, I think, which people obviously pay most attention to is what contributes to war and what contributes to revolution, collapse, regime changes of various kinds? And if you think about what decisions we make, if there’s anyone still around in 100 years to look at it, the big things that people will look at are things related to that. What were the big things they did that they didn’t understand at the time, but which clearly affected the next war, the next Revolution’s collapse? Does the Euro collapse? Is there a revival of fascism and communism in various forms in Europe? Is there a war between America and China over Taiwan or something? Those would be the big things. And if those don’t happen, then we’ll be largely forgotten. How much do people pay attention to the government in Britain in 1890-1895 now?
Dwarkesh Patel 01:18:10
Conflict matters a lot. Technology matters. And they matter in ways that are very contingent and hard to predict and very nonlinear.
Dominic Cummings 01:18:19
And ideas matter most of all. But almost none of that really comes from the government, right? It comes, almost by definition, from people who are fringe at the time.
Dwarkesh Patel 01:18:29
Why does democracy work? Or does it work? And why do some democracies work better than other democracies?
Dominic Cummings 01:18:36
Britain’s been a “democracy” for 80 to 100 years depending on how you define it. There’s been a particular kind of regime in most of Europe post 1945 of a kind of pseudo American empire which has entrenched a certain kind of democracy in various countries. But that’s a very, very small change in the sweep of history.
It’s like asking questions about 250 BC in 100 BC about the Roman Republic. Well, the Roman Republic has lasted for a century so far. You certainly can’t say now that democracy has proved to work. I would say the one thing you see in history is regimes are constantly changing and everyone thinks in their own time that we’re not going to change. What we’ve got is going to persist. But it’s always wrong. The most that things persist is a few generations, and then there’s always chaos and then there’s always a change.
Dwarkesh Patel 01:19:47
What is the next thing going to look like? And what should it look like? And the reason I asked that question is, one of the justifications of democracy is it’s nonviolent error correction and you can fix these mistakes. But on this account of what’s going on in the government now, not only are errors not being fixed, they’re accumulating and constantly increasing. So then is there a system in which these errors are constantly pruned away? What would that system look like?
Dominic Cummings 01:20:16
If you looked at what’s happened in the West over the last few decades and you brought back some of the people from classical Athens, it would seem quite familiar to them in certain ways. Some of the Roman aristocrats would say that democracy is having its predictable effects. You have demagogues in charge. You have a constant demand for more handouts, which is gradually bankrupting the country. You have moral and spiritual decay. You have a kind of collapse of internal cohesion. Pretty much what we’d expect historically.
I think the big inflection point was the mid 19th century. Before 1848, you had a bunch of conservatives like Metternich, who had watched the French Revolution, who had watched the guillotine and the terror and bloodshed, and they said, we’ve got to try and stop this. And they were actually conservative and they really meant it and they were really trying to turn the clock back.
After 1848, those people are kind of like either fled, retired, feel themselves doomed, and the old aristocrats who call themselves conservatives basically thought — well, how do we use democracy or universal male suffrage anyway, to try and smash the liberals up, who were their real enemy at the time. Which is what Bismarck and Napoleon II both tried to do.
But if you went back and looked at those people, brought those people back to life and got them looking at our current situation, I think they’d look at the first half of the 20th century and say, yes, things proceeded pretty much as we said they would. You allowed Christianity to collapse, you allowed the Socialists and the Democrats to win. And unsurprisingly, you had the torture chambers of the Gestapo, NSS, and the NKVD.
Dwarkesh Patel 01:22:29
How does that chain necessarily imply the catastrophes of …?
Dominic Cummings 01:22:33
I think old school aristocrats would assume that if you go down the path of democracy, then you will pretty rapidly end up with the Gestapo and the NKVD. And that is in fact what happened.
Dwarkesh Patel 01:22:44
But how come? Why is that the implication of the end state of democracy?
Dominic Cummings 01:22:49
That it’ll implode? I mean, look at what happened in Athens. They tried and it only really worked as long as you had the old aristocracy managing it. I’m not saying this is necessarily what I believe, all I’m saying is that a reasonable perspective is that the old system worked because the old aristocracy managed it. Pericles being the obvious example from the old Alcmaeonidae family.
Once they lost their grip and demagogues like Cleon took over, chaos, demagoguery and endless demands on the treasury, the people voting handouts for themselves. It collapses and the cycle repeats. Similar sort of arguments about the Roman Republic. So I think they wouldn’t be surprised at the collapse post war, post 1918 in Western Europe.
(01:23:32) - Bismarck & Lee Kuan Yew
Dwarkesh Patel 01:23:32
Maybe an interesting question is then if you look at Bismarck, who you’ve done a huge dive on, in that case, you have a system where after him, the system he set up couldn’t survive and yada, yada, yada, World War I. Whereas Lee Kuan Yew, another person you’ve studied deeply, he had meticulously set up a system so that even after he gave up the leadership, it would be a competent, successful government. And I guess we’ll see how it turns out. But Singapore seems to be running fine.
They both are these strong figures in the context of a democracy. What is the correct model here? And then how does this solve the succession problem?
Dominic Cummings 01:24:22
They are two very, extremely different examples. With Bismarck, you have a system where there’s universal male suffrage, but a kind of gerrymandered constitution written by him personally on a little Baltic island, designed essentially to make him the fulcrum of all of the power, but in a sort of lots of ways, hidden away. You certainly wouldn’t call it a democracy in anything like the current situation. All the deep state stuff, to use the modern terminology, the army, the intelligence service and everything were obviously completely excluded from that whole power structure and completely in the grip of the Prussian king.
Lee Kuan Yew has created a system where I think the most important thing he’s created is a certain kind of culture amongst the leadership in Singapore, a genuinely meritocratic culture, a culture where people are actually trying to solve the problems. Not like Washington or London now. And where there’s a kind of moral reinforcement for that culture, without which, everything else doesn’t fly. Therefore, it all becomes about: Generation by generation. Is that moral leadership? Does that moral leadership maintain or does it dissolve? Because it’s very easy for that sort of thing to go.
His son’s in charge now and I haven’t studied it very much, but he seems to be preserving the fundamental culture. The people running the different agencies, their job is to actually run the agencies properly. It doesn’t seem to be corrupted. But it’s very easy for that to change, right? Say he gets shot tomorrow or has a heart attack tomorrow, and someone else takes over very quickly. The signals could go out that actually, your job is not to do pandemic preparation properly. Here’s how people would be promoted from now on. You’d always expect that as the default. Because that’s just what entropy does.
Dwarkesh Patel 01:26:20
It brings us back to what is a system in which not only is a leader somebody who understands how to take control of the government and does it effectively, but also reliably hands off power to people?
Dominic Cummings 01:26:34
There is no such system. That’s why history is what it is. The Roman Republic lasted for a few centuries much longer than America’s lasted so far, and many times longer than democracy in Western Europe so far has been operating. But everything has its time of growth and decay, and it depends on the culture of the elites. It depends on the ideas that they believe in. It depends on. To what extent are they public spirited?
When I was over in Silicon Valley, I quoted to some of the guys there a famous letter that Cicero wrote where he said, everything basically is going to the dogs, because the leading people in the Republic are spending all their time in their wonderful houses, tending to their fish ponds. And if this carries on, then the Republic is clearly doomed. And he was right.
It’s extremely hard to turn those things around.There’s no reliable way to do it. How do you turn around an elite culture? What tends to happen is it implodes slowly, then hits a crisis and is blown up, and then something new comes along amid bloodshed and disaster. How often do you see a kind of deliberate, nonviolent, non-crisis led internal rejuvenation?
Dwarkesh Patel 01:27:57
Will Durant has this quote that Rome fell for longer than most empires have lasted, but we have these crises constantly. One could have imagined COVID could have caused this. I guess a rejuvenation will have to happen with a different generation.
Since you’ve brought up Silicon Valley. I do want to ask about something.
In 2021 you had a blog post about how in America, a Silicon Valley led or funded campaign in 2024 could have a big impact and get rid of the Trump v. Biden version 2 disaster. What happened? It seems like you now more focus your efforts back to the UK. What came of that attempt to look into what you could make happen in the US?
Dominic Cummings 01:28:41
To what extent are things just going to play out the way that they normally do historically. i.e. Slow rot. Elite blindness. Sudden crisis. Collapse. Bloodshed. Chaos. That’s normal. That’s the baseline expectation for our own current situation.
To what extent can one try to preempt that? And the only way really to try and preempt it is that you have to change the nature of the elites. And the obvious elite in America to look at is the elite who are most competent at building. So my point was, if you leave the old system in Washington, New York to itself, then it’s going to produce Trump vs Biden 2 probably. Obviously people could have heart attacks or whatever, but left to its own devices, that’s where the system is heading towards.
The obvious set of people that could put the country on a different track are the Silicon Valley builders. You should try. But that leads to a basic paradox, which is that the more and more mad the old system of politics and the old media get, the stronger are the cultural, financial, personal disincentives to competent people getting involved. In fact, the opposite is happening. The madder the system gets, the more the competent people retreat to their fish ponds. They look up, they build walled gardens, and they cultivate their fish ponds, they try and build their own companies where they can do things of value. They spend their time on research, they build hedge funds, they do things which are trying to insulate themselves and other people that they care about from the chaos.
A lot of them said to me personally, if I tried to do anything about it, then my investors would go mad, my employees would go mad, my family would think I’d lost my mind. I’d have demonstrators outside the house. Who the fuck needs that? The current system is working as intended. It’s closed. The old parties and the old media are driving themselves and everyone more and more mad. But the people who could change it are very highly disincentivized from getting involved.
Look at what’s happened to Elon. Elon, generally regarded by the old system as a hero, even up to two or three years ago, as soon as he says — Shouldn’t we take the First Amendment seriously? and by the way, this Ukrainian war seems insane. He’s called a supervillain.
I think a perfect way to understand the old political system in America and Europe is that across the political world, academia, in the media, you can see on Twitter, they’re all very happy to give their personal takes on Elon Musk as manager and his startup abilities. And they take each other’s take on that more seriously than they take the historical evidence of Elon’s abilities. Now, for them, they all think that they’re all sensible and they’re all rational. For others of us looking at it, we think that that’s just a wonderful sign of the old system’s complete madness and inability to face the most obvious things. Why do junior academics and political journalists, who can’t even cheat their own expenses competently, think that they’re competent to make judgments about Elon and his management of SpaceX? It’s completely crackers. But they don’t think it’s crackers. They think it’s perfectly reasonable.
Dwarkesh Patel 01:32:29
Let’s say if you did something in the UK, US, wherever, and suppose it succeeds, and it’s not somebody like Boris, where it’s somebody you have qualms about, but better you than somebody else. Let’s say it’s a handpicked person, you think they’re super competent, and you’re now chief advisor again.
Winning the election sounds like the easy bit, given the challenges you’ve talked about in terms of actually taking control of the government. What changes? Because last time, 80 seat majority, COVID crisis, huge mandate on both counts. Yet still, things couldn’t happen. Planning reform didn’t happen.
What would actually take on the next iteration for you to actually take control of the government? If you have to do a natural regime change, as you’ve called it?
Dominic Cummings 01:33:17
Take me out of the equation. Let’s think about the general thing.
Winning an election, in a lot of ways, is the easy part because in a lot of ways, the madness of the old system actually makes it easier and easier to win an election. Because the more divorced from reality that the old system gets, the simpler it is to win an election just by actually focusing on the voters. Which sounds completely crazy, but the old parties can’t do that. Why did we win the referendum in 2016? Because we focused on the voters, and the remain campaign didn’t. Why did we manage to prevail in 2019 when everyone thought what we were doing was completely mad? Because we focused on the voters.
But remember, everyone thought everything we did at every stage was completely mad and stupid and wouldn’t work. When you do things that are actually focused on what voters want, you seem insane to the old system.
Dwarkesh Patel 01:34:08
Wouldn’t you just expect some sort of basic evolutionary or selection argument for winning elections. Such that even if one politician just actually strategically tries to win elections, wouldn’t the system at least select for that? Because that’s presumably who’s getting elected.
Dominic Cummings 01:34:24
No, it doesn’t select for that at all. I mean, the whole history of Britain since 2016 is a perfect example of it. We won the campaign. Did the establishment go, “Oh, how did we lose that? We controlled the question. We had all the money. We controlled all these institutions. We controlled everything with power in the country. And then this startup hobbled with all sorts of problems, somehow won. Let’s investigate.”
No, of course not. They created a whole conspiracy about Putin, Trump, Facebook, everything else as an excuse not to face the reality.
Then in 2019, when we came along and we just did very, very obvious things, actually focused on voters outside London, they just made all the same mistakes again and blew themselves up again in 2019.
Look at the current situation now between Sunak and Starmer. And even Trump and Biden. All of them, repeatedly, every day, do things that make no sense. If your assumption is that they’re optimizing for winning the election, why would Rishi Sunak say: judge me on whether or not I stop the boats? Then when he’s told your whole plan and your legislation to stop the boats will not work and legally can’t work because of the European Convention’s Human rights. He just ignores it and does it anyway. It totally makes no sense if you think they’re optimizing for winning an election.
And if you spent three days doing market research in America, you’d know that Donald Trump’s message should be about the economy and what he should be saying is, this is what I did on the economy last time. This is what I’ll do on the economy next time. The reason why all these legal cases are being brought against me is because people don’t want me to do this with the economy. That’s what his message should be.
What should his message definitely not be? Arguing about the 2020 election and who won. Yet what does he do? Keeps arguing about that. All you have to do is most simple market research and you can say they are objectively not doing what is rational if you assume that they are focused and optimizing for winning an election.
Dwarkesh Patel 01:36:31
Who would you say was the last US president who was in charge of the government? Who was the last UK Prime Minister who was in charge of the government?
Dominic Cummings 01:36:47
Probably FDR, you’d say.
Dwarkesh Patel 01:36:49
Not even LBJ?
Dominic Cummings 01:36:50
I don’t know enough about LBJ to say. I mean, my impression is that for all his amazing skills, he didn’t manage to grip the Pentagon and the intelligence services and had all kinds of problems dealing with them. I mean, obviously, in some ways, LBJ clearly had an unprecedented grip, at least since 1945, over parts of the system for sure.
Dwarkesh Patel 01:37:14
But his was a legislative talent, not an executive. And then who in the UK would you say?
Dominic Cummings 01:37:19
I’d say Churchill.
Dwarkesh Patel 01:37:20
Probably not even not Thatcher?
Dominic Cummings 01:37:22
No, I mean, Thatcher just objectively very clearly didn’t if you look at just the memoirs of people that worked with her closely. The single biggest mistake Thatcher made was that she never actually got to grips with the permanent state, the permanent civil service. She talked about it a bit. She did some kind of sporadic occasional firings and a bit of purging here and there, but it was very less than half hearted, ineffectual, and contributed to her downfall.
(01:37:46) - How to fix the government?
Dwarkesh Patel 01:37:46
The problems you’ve highlighted have been since World War II. Everything from the media to the legal challenges to the civil service itself pushing back. We went through the entire list.
How is this actually going to work? The person is in charge. What happens next?
Dominic Cummings 01:38:04
It can’t be just one person. There needs to be some subset of the elite or part of the elite currently not really involved with politics, that decides to get involved with politics, that actually decides that its fundamental goal is to solve a set of problems. Its fundamental goal is not maneuvering the social hierarchy of existing political and media elites. Like that’s the fundamental question. At the moment, approximately everybody in the system is optimizing for their position in the social hierarchy of insiders. That means that no one wants to face reality. No one wants to tell uncomfortable truths. By definition, you can’t fix any of these agencies because that means alienating people immediately.
All of that can only change if a set of people say — “Here’s how we define the problem and solving this is actually what we’re here for, not staying friends with the old elites.”
And that’s why this thing very rarely happens. It’s why it normally, historically, only happens after a disaster.
Dwarkesh Patel 01:39:19
How big does that group need to be? So it’s not just the PM. Is it the entire cabinet? Is it the undersecretaries? How big does that minimum group need to be to take charge?
Dominic Cummings 01:39:29
Everything depends on particular historical circumstances. I mean, you could do it with a relatively small number of people. If you had the right PM, you could start off with ten people. But it has to be people that have a common set of what their goals are and who are very able and who are then able to build a network beyond that. No kind of coup or no kind of regime change ever happens if the small group of people at the start stay small. By definition, you have to convert people forcibly or through persuasion or whatever.
Dwarkesh Patel 01:40:07
What is a new equilibrium for these existing institutions, whether it’s the civil service or the media? Suppose you successfully reverse a lot of these trends and there’s a bunch of successes that ramp up. Do they keep opposing you or do they support the new regime? What is the new equilibrium for the blob?
Dominic Cummings 01:40:27
Well, by definition, I think all bureaucracies end up operating for themselves. So you should assume that if you do a kind of like a standard approach to this, and you say — We don’t like the existing bureaucracies, we’re going to take over, we’re going to scrap them and replace them. Then obviously within not very long, those things are going to operate pretty similarly to what the old things did.
The only way, I think, that there’s any chance of long term changes, you have to build into the system institutions and public acceptance and elite acceptance of a kind of constant reinvention and rejuvenation and closing and refounding of things. It’s the only conceivable way that over a time period like 100 years.
If you go back to your Singapore example, so far it’s been successful in preserving that culture. It’s like: here’s what our real goals are and the different elements of the government system have to change to meet those goals. And they constantly are adapting and we try and face our failures honestly. Like their report on what they got wrong on COVID which is already published and actually accepting the errors and trying to fix them. Unlike Britain, where we’re just paying lawyers hundreds of millions of quid to spend years arguing about it.
The only way that you could even imagine something happening long term is if you not just replace the existing elements with new ones, but you also create institutional mechanisms whereby the elites and the public accept these things need constant rejuvenation. And that’s a mix of basically sunsetting, closing and rebuilding.
Dwarkesh Patel 01:42:29
Let’s get more specific. Let’s talk about the Ministry of Defense. Would you just lay off a huge chunk of it? And there’s specific projects you’ve talked about that you’d sunset. But what would it look like to refound the Ministry of Defense, for example, or any specific department?
Dominic Cummings 01:42:45
So what you wouldn’t do, and what’s doomed with that, same as with the Pentagon, is going into those buildings and saying, “Right, let’s sit down and talk about a reform program for your procurement system.” It’s just never going to happen.
What you have to do is you have to set up a parallel thing that says here is the new procurement system that’s going to deal with the following things, and then you have to close the existing procurement system and get rid of 95% to 99% of the people involved with it. It’s the only thing that has any chance of success.
Dwarkesh Patel 01:43:21
What percentage of the people in the current Ministry of Defense would keep their jobs? How bigger of a change would it require to have something?
Dominic Cummings 01:43:30
On the order of like 90% or so.
Dwarkesh Patel 01:43:32
Dominic Cummings 01:43:32
But it obviously depends on department by department and different units. If you’re looking at, say, across government, the place where you could be and should be most aggressive is in everything to do with communications. Everyone has massive communications teams. Everything is worse as a consequence. Lots of those should be culled by like 95% to 99%.
I actually did this in the Department of Education. When we went into the Department of Education, the communications team was over 250. When we left, it was less than 50. Everything had improved, and if it would have been less than ten, it would have improved again.
Dwarkesh Patel 01:44:06
Well, the examples you gave of people like Groves and everything, this is a very particular time in history, wartime and also depression. So there’s not many private sector alternatives.
How would you have a situation today where you’re able to compete for that talent, not just for these really ambitious people in Number 10, but across the civil service? The best and brightest are going there. And until that happens, is there a way to get the best possible work out of the people who are in the civil service now, who are maybe not necessarily like the top of the class, but certain things need to be done.
Dominic Cummings 01:44:44
Very able people are extremely interested in politics and extremely interested in government. I don’t think the problem is in trying to make them interested in it. The fundamental problem is that they don’t want to get involved because rightly they say, “I’ll just be in endless stupid meetings. I can’t actually do anything.”
If you asked them “Would you like to run MI6?” A lot of them would think that would be a great job. Running Mi6 would be a great job and more interesting even than my own extremely interesting job in Silicon Valley. But would I actually be running Mi Six or would I just be sitting in a whole load of stupid meetings? And similarly, for whether it’s health reform or anything else, could you get some of these very able people to do it? Yes, you could.
And we saw in summer 2020, when the whole system opened up amid the chaos, there was actually incredible interest from all over of very able people wanting to come in and get involved to help, which I think is a sign of what I’m talking about. It’s not just hope, like, we could actually see it happening in 2020.
But what also happened in 2020 is another sign. Once the system started to close back in, in quarter 3, those people went back to their old jobs. If you’re sitting on some very successful thing, making a lot of money in your walled garden with your fish ponds, you’ll drop it if you can actually have a real impact. But if you can’t have a real impact, then you’re not there to just send stupid emails for trying to get authorizations for two years. And that’s the fundamental thing.
Dwarkesh Patel 01:46:24
So in the UK you’ll have labor for the next four or five years. So if you have a new startup party and let’s say it takes control after that, and you have another Brexit like political machine. Is that too late for some of the big crises you’re worried about? Whether there’s going to be a financial crisis. Or one big thing that people who have been listening to my podcast, my previous guest Dario Amodei, the CEO of Anthropic said that something that could emulate a human was two to three years. So you have a government in five years that starts working on it in the correct way. Is that too long to solve the crises that are coming up?
Dominic Cummings 01:47:06
Could well be. Maybe. As you see from the news, there’s all sorts of things that could happen. I mean, the people in charge could easily create a nuclear war in the next twelve months, right? Their behavior almost only makes sense if they actually wanted to do that. [Laughter]
Pessimism of the intellect, optimism of the will. My overall assumption is failure. That the Western world will go the same way that everyone else goes in history. In failure, collapse, and bloodshed. It’s probably what will happen, but still, you got to try.
Dwarkesh Patel 01:47:37
What will you personally be doing in these four to five years before this thing can ramp up? More reading and thinking?
Dominic Cummings 01:47:47
There are things in education I’m interested in doing in terms of building things outside the government system that don’t rely on the government. I’m thinking about the idea of creating a new political party. Someone I think is going to do it at some point. My thought has been that the normal path of history is collapse and then people create something new out of that. It’s extremely hard to avert crises, but if we’re going to, then there needs to be some new political force here. There needs to be a new political force in America that can replace existing elites. The fact that they’re all getting so old and that they’re so visibly failing gives some hope that it could happen.
But it comes back to this Catch-22 problem. The worse the system gets, the harder it’s becoming to get able people to get involved with politics, not easier. And the whole idea of a new party and the whole idea of doing politics differently, everything fundamentally rests on whether or not you can get able people to step forward and do it. And history suggests that’s phenomenally hard, that normally that only happens after the crisis comes.
Dwarkesh Patel 01:49:14
Speaking of which, how would you get these political talents to hop onto your party’s label? And it’s sort of a Catch-22 because you probably need voters to convince the political talent to join your party but you need the political talent to track the voters.
Dominic Cummings 01:49:33
Dwarkesh Patel 01:49:34
What is the trick to getting these people to ditch the existing system?
Dominic Cummings 01:49:38
Well, it’s extremely hard in the first past the post system. When I did some research in America recently, I asked about the third party thing, and you see this problem come straight through in focus groups. People say straight away, I hate both the old parties. They’re both obviously rotten. They both obviously can’t do a good job. We obviously need something new. America’s got so many wonderful people. Why the fuck have we got all of these ancient, old, useless people in charge of absolutely everything? It’s all a nightmare.
“All right, okay. So imagine the following party comes along, and it’s got the following people, and it says the following things. What do you think?”
“Oh my God. That would be so great, but I’m not sure if I would vote for it.”
“Well, because my vote might be wasted, and it might let the crazy people from the other side in.”
So even if you imagine this wonderful new entity run by the best and brightest in America with an agenda you actually agree with, even then you might not vote for it because you fear that it might let you. You see, there’s a fundamental structural problem with the first past the post system. Now, I’m in favor of first past the post system. I’m not suggesting getting rid of it, but it does make it structurally very hard to replace one of the old entities.
I mean, it’s why the whole Brexit thing was such an incredible missed opportunity. Winning the referendum. The chaos that that created. The collapse of both the old parties in 2019. One of the parties coming to us, begging for us to save them. Us going into Number 10. Having the majority. We kind of maneuvered the system into such a place that we could actually then transform one of the old parties into the new thing that we wanted to build. That’s very hard, but it’s easier than creating something new and then taking over in a first past the post system.
Dwarkesh Patel 01:51:30
On Brexit. If it’s possible to build a new political machine that’s relatively cheap to do and is much more effective than the current political parties at winning elections, what is it that you would do? Whether it’s political modeling, whether it’s the use of new technologies, or I don’t know, testing things out with social media. What are the tricks that you would use to win?
Dominic Cummings 01:52:05
Obviously all the AI stuff means that there’s huge opportunities there. But I would stress Colonel Boyd: People, ideas, machines, in that order. Why was it that we built tools to help win an election that no one else built? The primary question is that we really wanted to win, and we really were focused on the voters, and that’s why we actually built tools. The old parties aren’t focused on the voters and aren’t really interested in what they’ve got to say. Therefore, that’s just not what they are thinking about every day. So I don’t think the most important question is the technical side. I think it’s — can you create something with very able people with Jeff Bezos like obsession on the customer, that really, really wants to obsess on the voter. If you can build that culture, then that culture will then build the technology and will exploit the tools 100 times better than the old parties will. But it’s that that comes first, not the technology.
Dwarkesh Patel 01:53:09
With Brexit. Afterwards on your blog, you’re writing about why you did Brexit and how the government is dysfunctional in these ways and it’s similar to the things we’re talking about. We’ve got these crises that can happen in minutes.
How similar, in retrospect, do you think that is to the message that Vote Leave, the organization you chaired, advertises the reason for Brexit? One way you could say is like, “Oh, it’s customer obsession. What is the customer here interested in?”
Is it important for the two to be the same? The reason you wanted to do Brexit versus the reason that you might advertise Brexit should be done.
Dominic Cummings 01:53:39
Any kind of mass communication always involves huge simplification and huge focus. It’s just an unavoidable aspect of things. Forget me and Brexit, anybody doing politics, has to deal with it. Same with Bezos and Amazon. Yeah, it’s customer obsessed but there’s also a whole bunch of stuff that he’s not talking to the customers about.
So I think that’s just intrinsic to politics. The vast majority of what you’re thinking about, what you’re talking about are not going to be the focus of public communications.
Dwarkesh Patel 01:54:12
How do you think about the bargain with voters? Is it like I have these long term priorities and the voters want this thing, which I don’t really care about, but it’s almost like a trade where — Help me achieve these long term priorities or give me the power to achieve these long term priorities and these other sort of, like, smaller things, which I don’t think are the most important thing, will also accomplish those.
Is that the way you think about it?
Dominic Cummings 01:54:35
To some extent, that’s always what politics is. Because as I said before, you could only ever talk about a relatively small fraction of all the things that you think are important.
Take civil service reform. Civil service reform isn’t even a subject of interest to the insiders who have to live with it all day. I actually have more conversations with people outside politics about that than I do inside politics. Seems very weird but the people inside the system just so completely accept the existing system that they don’t even really talk about that.
So is a new party going to spend a huge amount of time discussing the intricacies of HR for how some kind of new civil service should work? How to create red teams inside Whitehall? How to create a startup for a drone army to replace parts of the MoD? No, because most people are not interested in all of that. But neither are the insiders. Neither are the people. The thousand most powerful people in the country don’t think about that at all.
Dwarkesh Patel 01:55:39
While we’re talking about talent, you obviously put out that famous post saying we want weirdos and misfits to come into government.
There’s lots of weirdos and misfits out there. What is a specific kind that is really effective in government?
Dominic Cummings 01:55:52
There are lots of different kinds and It depends very much on what you’re doing. Some of the people we brought in were very technical. For example, something that you’re interested in in the whole AI world, there were some very technical AI people. It massively depends on the particular roles.
There’s another set of talent which is rare inside the government, which is just a kind of entrepreneurial project management type E, actually getting things done character. No surprise that we ended up in a crisis having to use British special forces for a lot of those kinds of things because they’re pretty much one of the few elements of the British state that is still extremely able and has that kind of operational, punchy, can get things done fast vibe. So it’s just highly dependent.
(01:56:43) - Taiwan
Dwarkesh Patel 01:56:43
Is a war over Taiwan with China worth it for the US and the UK and the West together?
Dominic Cummings 01:56:48
Obviously not. It’s a completely insane idea. It’s an island that you can see from the Chinese coast. It’s full of Chinese people. The people in the army are cousins of the people in the opposite army. For decades, one China policy was accepted by Democrats, Republicans across the Western world. It’s going to be unified, but it should be done peacefully, not violently. That was the right approach.
It’s completely crazy for us to be ditching that approach and to be willfully antagonizing China over a place filled with Chinese. It’s fucking stupid.
Dwarkesh Patel 01:57:26
But it has worked in the past where the Soviets had the huge Red army in Europe and with nuclear deterrence, we were able to keep them from conquering the rest of Europe. Why not just have this sort of thing again?
Dominic Cummings 01:57:39
Us threatening the Soviets to say you shouldn’t invade Germany and France and Britain is a parallel to say that Taiwan shouldn’t be part of China? That’s a really bad analogy.
Dwarkesh Patel 01:57:51
What’s wrong with a strategic ambiguity where we haven’t committed to defending Taiwan, but just making China think twice about invading Taiwan?
Dominic Cummings 01:58:00
Britain gave strategic ambiguity a great run out in summer 1914. How did that work out? So you want to repeat that with nuclear weapons where the crisis can kill a 1000 times more people and happen a hundred times faster than what happened in summer 1914 over an island filled with Chinese people.
Also, it’s just not credible, right? Do you think the Chinese are sitting there thinking, “Okay, we know that Taiwan is completely existential for us, and America also thinks it’s existential for them?”
No, of course not. Of course that’s not what they’re thinking, because it isn’t. So you can’t go around threatening nuclear war over something where it’s not actually a credible threat. Where people think either it’s complete bullshit or they are actually insane.
Dwarkesh Patel 01:58:58
But that is the entire strategy behind mutually assured destruction as well, right? And that worked. To actually retaliate is kind of insane.
Dominic Cummings 01:59:07
But there it was credible for America. There the argument for America was: we are defending the Western world. We have NATO. There is a NATO agreement on mutual self defense. And if you breach that and you attack one of us, then it means war against everyone.
That was a lot more credible than there’s a little island off the Chinese coast. It’s got nothing to do with us historically at all. It’s full of Chinese people. But we’re going to threaten potential nuclear war over it. It’s totally non-credible. And doing it makes you sound mad.
Dwarkesh Patel 01:59:46
I was talking to a friend about this, and he was defending extending the nuclear umbrella to Taiwan. And even if it works, it’s kind of like having a girlfriend who says, if we break up, I’m going to burn down your house. So we better not break up.
Dominic Cummings 01:59:58
Or I’m going to kill myself. Maybe you decide that day to do what she wants, but you’re also thinking she’s mad and this is not sustainable.
Dwarkesh Patel 02:00:08
Yeah, and not responsible.
(02:00:10) - Russia
Dwarkesh Patel 02:00:10
You spent time in Russia when you were a younger man. You were trying to start an airline there, right?
Dominic Cummings 02:00:15
Dwarkesh Patel 02:00:16
What did you learn about Russia, and do you think the regime there is stable in the long run?
Dominic Cummings 02:00:20
I learned a lot of things doing it. I learned that Russia is a mafia state. I learned that London was treated back in the mid 90s, as the Moscow taxi drivers called it, the laundry for the mafia estate. That is something which people here have not wanted to face because there’s a lot of money to be made out of the whole enterprise, even in 2021. When I said that it’s just crazy that we’re still doing this. There was a lot of pushback. Literally months before the Ukraine war started, it was completely normal in London to defend the way in which we’ve handled the whole thing for the last 20 years.
I also learned about incentives a lot. So everything I had to do with the airline was a total fiasco. And one of the things I figured I learned was — I, like many naive Westerners, got involved with it, thinking, well, the people obviously want to make the airline work, right? That’s the whole point of having an airline. It’s a successful airline. Wrong. Actually, what most Russians wanted to do was steal from the airline and move the money offshore. And I never realized that. And a lot of the people involved with it didn’t realize that.
Russia was full of Harvard MBA types flying in from JFK and arriving in Russia thinking that it was some kind of vaguely normal country. And they, like me, learned the hard way that it is not.
Dwarkesh Patel 02:02:00
That lesson with incentives, for example, is isomorphic to other institutions you’ve seen, or was that specific to Russia?
Dominic Cummings 02:02:08
Everything is crazier and more insane in Russia, but I think it definitely helped me deal with other failures. You always learn a lot, I think, from total failures and having to confront the fact that you just totally misunderstood what on earth you were getting into. And I completely misunderstood what I was getting into there on every level.
So I think it was very useful for me, actually. When I came back to Britain and then got involved with politics, I was much more careful about trying not to fool myself and trying to figure out what’s real and question my own assumptions about things.
Dwarkesh Patel 02:02:50
Speaking of which, a character you’ve highlighted who exemplifies that sort of epistemic humility is Bismarck. You’ve done a deep dive on him so I just want to ask you a bunch of questions, not only about him, but maybe just a study of history in general.
First of all, where in the world do you think a Lee Kuan Yew or Bismarck type figure is most likely to emerge as the leader of that country? Where is it most likely today?
Dominic Cummings 02:03:13
I can imagine someone like that emerging in China or Russia. It’s much harder to imagine someone like that emerging in Britain or America or the Western world.
Dwarkesh Patel 02:03:29
Dominic Cummings 02:03:30
They’d be crushed by the system. You could imagine in China, some Machiavellian character coming through. You could say the CCP has done that to some extent. A Machiavellian character comes and exploits the system in a kind of somewhat mafia way, manages to take over, and then Stalin style starts purging people and manipulating the whole thing so that they acquire more and more power.
The good thing about the British and American system is that it’s proved quite resistant to that sort of thing. But that hits upon a fundamental paradox. That our systems have been good at preventing the kind of catastrophe of Stalin-type person taking over but the very characteristics that make it very hard for a Stalin to take over also make it chronically incompetent as well.
Dwarkesh Patel 02:04:29
Yeah, or hard for a good statesman to take over as well.
You are somebody who has actually lived through and participated at a high level in a hinge point in history, with masterminding Brexit and then as chief advisor doing COVID.
What is it that you now understand about how history is written? When you were looking back at previous periods, how do you think about what is written versus what actually happened? Knowing how your time has been described in the press and in contemporary accounts.
Dominic Cummings 02:05:02
Close to 100% of the things written about me, what I was trying to do, what I thought, what I said, what I wanted to do, what I actually did, is complete garbage. So it obviously affects how you see everything else. It can’t have any other effect. It was true in the Department of Education. It was true in the referendum itself. It was true in Number 10.
It definitely makes me much more skeptical about all kinds of details. I think whenever you read history, it definitely makes me think all the time. “Yeah, but that might easily be wrong. And what else was actually happening behind the scenes also?” Also just like, how much is lost to it.
If I think about my own time in doing the referendum or Number 10, close to 100% of the reality of conversations and motivations and calculations is basically lost. It’s history. None of it is really recorded anywhere. I mean, there might be some sketchy WhatsApp conversations here and there. The odd person keeps a diary, which can be useful for sure, but it was pretty rare, certainly amongst top people. Even if they keep it, I think they destroy.
Dwarkesh Patel 02:06:28
Especially in this age where so much is recorded even if you don’t try to. Whereas if you look back at Bismarck’s time, there’s not any emails and so forth that you can dig up.
Dominic Cummings 02:06:39
Yeah, but it’s a bit of paradoxical though because those guys were all constantly sending letters to each other and a lot of those letters do end up surviving because they are just filed in some paper archive somewhere and then 60 years later there’s a war and people get their hands on them. For us, though, how much of key people’s WhatsApps between Jared and Trump or me and Boris, or all these sorts of things, how much of that digital archive actually will be saved and accessible is arguably less than in the old days.
(02:07:12) - Bismarck’s career as an example of AI (mis)alignment
Dwarkesh Patel 02:07:12
You were giving a talk or participating in a panel at the Orwell Festival of Political Writing about Bismarck, and you had a quote there that I thought was so interesting and I want you to explain it.
Bismarck’s partly a story of how intelligence tries very hard to escape all constraints, and all attempts by less intelligent people to force higher intelligence to align with certain goals and values are at best highly uncertain and dangerous.
What did you mean by that?
Dominic Cummings 02:07:38
So it was prompted by studying Bismarck at the same time as having a lot of conversations with people about the whole AI alignment problem and whatnot. And it just occurred to me one day that it’s an interesting exercise to consider Bismarck’s career through the prism of the AI alignment arguments. So here are all the safety features that we’re going to try and create. We can constrain it in this way, we can constrain it in that way. If all else fails, we can try and kill it. Can we switch it off? Etc.
But if you look at his career, you basically see the AI alignment problem actually just living. You see something that’s much more able than its competitors and it defines success fundamentally as expansion of its own power, which also means its own freedom of maneuver. And it treats all attempts to align its goals with broader goals as enemy action to be destroyed. And it’s highly effective at preemptively destroying them. People try and send the intelligence services to close it down, it ends up taking over intelligence services and using them to blackmail its opponents and forcing them into suicide and exile.
At every stage, every attempt to create safety features is defeated. And the ultimate thing is to just switch it off. As people start writing letters all across Europe essentially saying, “Switch it off. Just switch it off.” But they can’t switch it off.
Dwarkesh Patel 02:09:14
It kind of works at the end, right?
Dominic Cummings 02:09:16
Well, it’s human, so it gets really old.
Dwarkesh Patel 02:09:19
But who was the Kaiser that kicked him out?
Dominic Cummings 02:09:21
And in the end the Kaiser says “I think I can do without you.” The Kaiser was wrong, of course, and everything immediately started going downhill for Germany thereafter. And in some weird ways, also parallel to the AI argument, a lot of the key people around Europe started sort of wistfully saying, “Well, you know, it was a nightmare to deal with this terrible, superhuman AI, but it wasn’t a pirate. It wasn’t insane. You could at least negotiate surrender. It wasn’t just a berserk pirate.” Some of them almost felt nostalgic for its rationality after it had gone.
Dwarkesh Patel 02:10:04
Another analogy here is it set up all these things that nobody understands, these systems of alliances and relationships in Europe. And you can imagine, like an AI advisor, and you’re not sure what’s going on, but it’s making, I don’t know, maybe it’s like a hedge fund manager that’s AI and it’s like making a lot of money. And then you’re like, but I don’t trust it. I’m going to shut it off. And then a few years later, there’s like a huge financial collapse because it was doing something you didn’t realize that was super important. You shut it off. You didn’t understand what it was doing that was super important. You shut it off. Suddenly all the shit that it tried to do goes to shit. And you have World War I.
Dominic Cummings 02:10:37
Your interpretability program that’s running it doesn’t really work. It can’t really explain to you why it’s doing these things. And then you think, “Oh, well, it all looks a bit murky, I’ll close it down.” And then you find out that actually the structure that it built was super valuable.
Dwarkesh Patel 02:10:55
Yeah, that’s such a great analogy.
Why is there not a definitive biography of Bismarck out there?
Dominic Cummings 02:11:00
Well, I would say there is really. There’s a three volume one by a guy called Oto Pflanze done in the 80sm, I think. Not very well known, but it’s by far the best.
Dwarkesh Patel 02:11:11
Okay, big picture. What is the cause of Britain no longer being the global superpower it once was? Is it decolonization? Post war socialism? Is it the war itself?
The productivity stuff started happening after 2005, but people have been talking about England declining for a long time. What is the big picture cause of that?
Dominic Cummings 02:11:33
Well, the most important thing, obviously, is World War I. At the end of that the naval dominance was gone. Huge financial reserves poured into the mud of the trenches. I mean, Europe’s never recovered from it, never mind just Britain. So I think that was the single biggest element to it.
Dwarkesh Patel 02:12:02
Yeah. Wasn’t it like one in eight British men between the ages of 20 and 40 died. Or it was maybe even higher than that.
Dominic Cummings 02:12:08
The years afterwards saw the biggest transfer of property in Britain in basically 500 years since Henry VI separated us from the Catholic Church, the dissolution of the monasteries, and that huge shift in the 16th century.
Immediately after World War I there was a similar kind of epic shift. As a consequence of the scale of death. And also, you also see that echoed in the blogs I did on Alan Brookes diaries through World War II. And it’s interesting he keeps referring back to it as well, when he’s constantly bemoaning another military disaster and why the British army is not working so well. His answer is, “Well, of course, like a whole generation of great leaders was destroyed.”
Dwarkesh Patel 02:12:53
I’m curious why the opposite thing isn’t more common, where you have all these generals who have seen World War I, maybe they’re better at fighting World War II as a result? I mean, Hitler was a World War I officer. The World War II generation did seem special. Did the World War I experience help them in any way?
Dominic Cummings 02:13:11
I don’t know, and I’m not an expert but it’s striking that the view of contemporaries was that lots of the best young people were killed, and that’s why now a lot of the senior echelons are not up to the job.
Dwarkesh Patel 02:13:24
One thing you’ve emphasized a lot over your writing and throughout your entire career is the need for basic research as a way to move England forward and maybe move the west forward. But when you look at England’s GDP per capita, it’s far behind the US. And one thing I was wondering about is maybe it makes sense for the US to be doing a lot of basic research to expand the frontier. But Britain doesn’t even seem to be on the frontier.
If it’s just about economic growth, at this point shouldn’t it just be copying the things that already work in the US instead of doing a whole bunch of basic research and hoping in like 20 years or something, it contributes to productivity growth or something.
Dominic Cummings 02:14:00
The answer is not as simple. Britain needs to do a whole bunch of things. Some of it is frontier research, but that’s only part of it. Part of the reason for that also is not just economic growth, but also security and national independence, etc. How much of Europe in 20 years time is actually going to have meaningfully independent capabilities in lots of areas and how much of its choice will only be — do we buy American or do we buy Chinese?
If Britain wants to have its own choices, then it’s going to have to build things. The science and technology staff is important for us to develop, but not just for economic reasons. But it’s also very, very far from a magic bullet. There’s a whole bunch of stuff that we have to do and that we started work on in 2020. The zoning laws are a crucial part of it. The whole ecosystem around startups. The applications of the research, that whole ecosystem is a nightmare in all sorts of ways and it’s one of the reasons why constantly great startups here end up selling out early to American companies. All of that needs dealing with. There’s just massive regulation of area after area after area, which throttles growth. The whole housing market is a complete shit show. I mean, everywhere you look is bad.
Dwarkesh Patel 02:15:30
Given that most attempts at finding NIMBY and red tape and over regulation have failed. Even the one you attempted to fix planning, aka Zoning in the UK, when you were chief advisor. What is the correct strategy to fighting NIMBY? Because it just seems like really hard to fight it head on.
Dominic Cummings 02:15:49
I think a lot of people look for a magic bullet of communication whereby we come up with some equivalent of take back control, but for planning, and then we persuade everyone and the public will shift and start cheering house building. I don’t think that’s realistic and I don’t think that’s the answer.
In fact, in 2020, we did actually make some changes to the planning laws, but I think it’s instructive how we did it. We didn’t talk about it at all. I actually basically had a communications blackout on the whole thing and that’s because talking about it is extremely hard to do, even if you’re very good at communications. So it was completely pointless trying to do it with that Number 10 and that Prime Minister and that Conservative party. So my approach to it then was just do it, don’t talk about it, and talk about other things. And so we did actually manage to get various things done then, but without most people even noticing.
I think overall, though, if you’re really trying to do at a big scale, the actual answer is that you just do it. You focus public communications largely on other things. When people actually see the effects of it with growth and houses they can afford and businesses starting and growing and thriving local areas and whatnot, then you can point to that. But you’re not doing it in a theoretical way. It’s not politicians promising things. It’s being able to point to real change in their areas and say, “Look at what’s happened. Now, do you want to go back? Do you want the people who don’t like this to take over?” You see what I mean? I think the idea of coming up with some kind of agenda on it and then trying to persuade the public about mass deregulation and mass housing is a fool’s errand.
(02:17:37) - Odyssean education
Dwarkesh Patel 02:17:37
Let’s talk about education for a bit. You mentioned earlier that that’s one of the things you’re working on now. But also you have a big history here. You were at the Ministry of Education, and while there you wrote a very interesting report on an Odyssean education and this was one of the quotes from that that I thought was pretty compelling.
“We need leaders with an understanding of Thucydides and statistical modelling, who have read The Brothers Karamazov and The Quark and the Jaguar, who can feel Kipling’s Kim and succeed in Tetlock’s Good Judgement Project. An Odyssean education would focus on humans’ biggest and most important problems and explain connections between them to train synthesisers.”
This is a very interesting idea. Do you want to explain a bit about what the idea behind the Odysseyan education was?
Dominic Cummings 02:18:26
Politics is the hardest thing. It’s the most complicated. There’s a reason why it’s just constant failures. It’s much harder than the hardest startup. The people who can do well at the top level are able to look at lots of different things and have some kind of sense for them.
Generally speaking, the people who are doing it have done history degrees or politics, philosophy, economics degrees at Oxford type stuff. And then they’ve gone into the world of politics and government. And it’s just an incredibly narrow way of looking at the world. There’s huge skills that you don’t have. Intellectual skills you don’t have. Practical skills you don’t have. You don’t understand how things actually get done. You don’t understand how regulation actually has its effect, because you don’t see at the coalface what it’s like starting a business or scaling a business.
My view is much stronger, having been in Number 10. That what I wrote ten years ago was correct, that especially having gone through COVID and the horrors around that. The way to deal with that is to have different kinds of people with different kinds of skills. But you also need some people who can look across multiple domains at the same time. So I think it’s not a surprise, for example, that one of the most useful people in COVID was someone who’d done a Physics PhD, was now working in AI, but had also built a business and had worked with that business in Whitehall. So they could see and understand a lot of the difficult science around COVID. They knew what it’s like actually running a business. They understood the interface between the business world and the government. They were familiar with all of those different worlds, and that meant that they were much better able at looking at the overall problems than the average MP, the average senior official, the average special advisor, etc.
Dwarkesh Patel 02:20:42
Is this ability to synthesize across fields something you can actually train in something analogous to Plato’s academy? Or is this just that you have to hire people who have it? The reason I ask is: One thing I was thinking while reading this, you have a lot of famous examples in history of scientists and mathematicians. There’s scientists who understand very complicated things, but they don’t understand the very basic insight in Hayek’s knowledge problem and they are socialists. Or they don’t understand the basic game theory, and they’re pacifists. They can understand complicated things, but having that sort of ability to transfer between — I understand power laws, therefore I understand the power of scaling in AI or, I understand exponentials, therefore I understand how bad a pandemic could be. That’s like an additional thing that’s hard to train.
Dominic Cummings 02:21:31
It totally depends on temperament and personality. Two of the most interesting intellectual people, 20th century are John von Neumann and Kurt Gödel, who are both friends, fled Europe because of the Nazis and ended up in Princeton. Both extraordinary minds. Both did things of stunning originality. Von Neumann was completely at home in the world of Washington and could also go and deal with bureaucrats and navigate the corridors of the White House. He could give presentations to politicians and understood — Let’s have a pre meeting about this in order to make sure that so and so understands that. And so then the question is presented to the president in the right order so that this guy doesn’t say that before this person says that, etc. All the kind of political skills you need to actually get something done and avoid your meeting turning into a shit show.
Kurt Gödel is like the exact opposite. If you’re going to put people on a graph, Kurt Girdle would be like the exact opposite end of that. Famously totally useless in every practical way, a menace to himself. So, yes you want people with great intellectual ability, but as you say, some of those people can be highly functional in a political environment and some of them are completely catastrophic.
Dwarkesh Patel 02:22:46
Yeah, but it seems like I don’t know how you train a John von Neumann, because then he also assisted the US in thinking about Cold War game theory and things like that. Whereas is it just easier to hire those people?
Actually, maybe a different way to ask this question is, does the UK train these technical people who can synthesize across fields but it’s just that they’re trained in the UK, but they go to the private sector or emigrate? Or is it just that the UK doesn’t train them to the same extent as the US does?
Dominic Cummings 02:23:19
I think elite universities now definitely don’t train this sort of thing. The way that I think about it is this. It’s easy to imagine creating new courses for 17 to 23 year olds where they do a much broader mix of intellectual subjects, where they study a broader mix of intellectual subjects. They read some Thucydides but they also understand some basics about statistics and Tetlock type stuff. But also, instead of them just sort of going off on summer holidays and wasting time, they also go off and work at SpaceX or some kind of startup or with the military, or in a hospital ward, ER room. And they’re moving between these worlds of the theoretical and the intensely practical constantly.
Now, that won’t be to everyone’s temperament. Some people have a temperament where they find that interesting and they’ll get a lot of value out of it. A lot of people won’t. So you have to think of it in multiple legs. So one thing is to create those courses and that will find you a set of people. But there’s another set of people, let’s say Tim Gowers, Field’s medalist and mathematician at Cambridge. He’s gone through one whole thing now.
Are you going to send him off on some course? No. But could people like that be brought into government to help in all sorts of ways? Yes. If they spent time working in government on some practical problems, would they develop a much better understanding of how government actually works and therefore be able to come up with new insights about things? Of course they will. I don’t mean they all will, because some of them would be like, “Good, all right, and just be completely hopeless in that environment” and it would be unfair on everyone and them to involve them in it. But Tim Gowers would be a good example of someone who personally could cope with the environment and I think would bring great value to.
I don’t think there’s just one thing. What we need is multiple experiments with different pathways. Some of it is bringing in older, established, elite talent and then mixing them up in the system. Some of it is training younger, elite talent differently. But there’s no magic bullet to it. You have to try a bunch of things at the same time. Some people will respond very well to an odd mix of theoretical intellectual pursuits and intensely practical things. And that is definitely something worth trying to train and definitely can be trained.
Dwarkesh Patel 02:25:56
How does PPE (Politics, Philosophy, Economics) fuck up the people who go through it? Because basically, almost every Prime Minister has gone through it, and even US presidents? Clinton was a Rhodes Scholar and he went through it. What is the way of thinking that it instills that you find problematic?
Dominic Cummings 02:26:15
I think it encourages this whole kind of wordcel bluffing. That everything just becomes that you spend a week, you skim through a few books, you come up with some vaguely plausible stuff, and if you seem with good manners and a bit of social polish, then suddenly it all seems plausible. And you’re all surrounded by people who operate in exactly the same way.So you have a room full of polished social wordcels. It’s a very bad set of people to throw at a problem like COVID at them.And also it encourages them to think that the plausible sounding few sentences to get through this conversation is actually what’s important rather than— What’s the truth? What’s the actual answer to the problem? How are you actually going to implement this over many years? Everything like that is completely disdained.
It points to a general problem with our system. One of the most important things is that everything to do with operations and management and actually getting things done is like the lowest status thing in Whitehall. The highest status thing is A) bullshit about “political strategy” and media and giving interviews. Everyone wants the word Strategy in their job title, and it’s practically always bullshit. It shouldn’t be in the job title, and the job will be improved by removing the word from their job title. And nobody wants to be on logistics, operations, and actually making sure that something happens. And that’s at the core of why so many things work the way they do.
And you can even just see it in the whole policy process. The high status thing is writing the policy and then spinning it to the media. The totally low status thing is what are the actual implementation details? Has someone already tried to do this in eight different countries in the last four years and each time it’s been a disaster? No one cares about that. That’s left to much lower status jobs and none of the senior people will pay attention to it and they’re actually just formally separated.
So one of the things that we did in 2020 was we said that one of the ways in which the Number 10 system should change is the policy people should be brought together and actually physically sit with the management people, which is completely revolutionary idea, because that way when you’re thinking up the ideas in the first place, you’re immediately from the beginning talking to the people whose job it is ask “Will this actually work?”
Rather than what happens now is that these people sit around, they go to bullshit seminars, they talk to journalists, they talk to MPs, they publish a paper. It’s actually full of holes. It doesn’t work. That paper then goes off to the implementation people, the policy people then move on to whatever their next thing is. The management, implementation people look at it and go, “Well, this is not going to work for all the following reasons.” But it’s their job to make it work. They then have to go back and then start arguing.
“But we already decided that we were going to do this like nine months ago.”
“Yeah, but it can’t work.”
So our idea was if you put them together, then you actually find out on day one or day four. — “Oh, actually, this is a stupid idea and it won’t work for the following reasons. We have to do it this way instead.”
No one cares about that. It goes back to, what do people care about? No one actually cares if it doesn’t work.
Dwarkesh Patel 02:29:52
Yeah, the debug loop is broken. It’s like you’re debugging a program, but you can only do it a month after you write the program or something.
Dominic Cummings 02:29:58
Yeah, exactly. Imagine if you could only debug the program like 18 months after you wrote the program.
Dwarkesh Patel 02:30:02
What do effective altruist and rationalist types get most wrong about politics?
Dominic Cummings 02:30:07
They don’t understand how low the caliber of the people are in politics. They don’t understand the fundamental golden rule, which is that the people in politics are almost never actually trying to solve the problem and don’t care about solving the problem. I have constant conversations with E/R types where they think that the political people they’re going to talk to really care about solving Problem X, which is fundamentally incorrect. So it’s not surprising that a lot of their plans go wrong.
I think that they are also susceptible to the idea that elites are more rational and the voters are more dumb and easily manipulated by emotions, whereas in fact, the truth is the exact opposite. In my opinion, educated elites are by far the easiest to manipulate with emotional propaganda, and the voters are much tougher to fool. The rationalists think it’s the other way around.
Dwarkesh Patel 02:31:07
Conventionally, you are seen as, or thought of as combative. I mean, in this interview you’ve been very polite and pleasant, but does the fact that you’re seen as combative, does that help you maybe get your way when you’re in government or something? And is it calculated?
Dominic Cummings 02:31:23
It’s not calculated. I don’t know if combative is the right word. I think the fact that people inside the system know that I actually do have priorities and I do actually care about them and I don’t care what the media says, and I have an extremely high tolerance for everyone going crazy in order to actually get what I want definitely, obviously, has an effect, because people will think, “Well, maybe we should do a deal with him. He cares a lot about blah, maybe we should just let him do blah and see if we can persuade him.”
We’ll help you with this, and maybe you could help us with something else.
Having priorities, actual priorities, is incredibly rare in politics. Almost no MPs have them all. Practically by definition. Not worrying what the media says and not worrying about everyone hating you, I think, is a very powerful advantage.
All the way through the referendum and all the way through Number 10, the insiders constantly said — These people are just completely crazy. They don’t know what they’re doing. It’s all going to fall down on their heads. That definitely was an advantage because it just means that they constantly fooled themselves and undermined what they were doing. And they didn’t learn from 2016. The mistakes they made in 2016, they just totally repeated in 2019.
Dwarkesh Patel 02:32:49
And in a system where people don’t have priorities, don’t actually care about policy, somebody who does has got to seem strange to them in their world.
Okay, final question. What does pessimism of the intellect and optimism of the will mean to you?
Dominic Cummings 02:33:06
If you look at history, then you have to expect that what happens to everyone else is likely to happen to us, i.e. disaster. But that’s not a reason for giving up. You have to try, even though you don’t think it’s going to work.
Dwarkesh Patel 02:33:19
Dominic, this is a huge pleasure. I honestly learned so much. You really can’t get this anywhere else. Having somebody who is as thoughtful as you inside the government, so that you can have both a sort of higher level intellectual picture, but also the level of detail and knowledge of what actually happened and the strategy and information of it. I couldn’t get it from any other interviewee. So this is a huge pleasure. I really enjoyed this.
Dominic Cummings 02:33:43
Thank you so much. I really enjoyed it.