Mo Gawdat - Artificial Intelligence: "We Fucked Up & Lost Control" - Former Google Employee Speaks

Full transcript: Mo Gawdat on London Real

0:00 Brian Rose: This is London Real. I am Brian Rose. My guest today is Mo Gawdat, the entrepreneur, author and former Chief Business Officer at Google X. You spent 30 years working in the heart of the technology industry with some of the biggest organisations on the planet, including IBM, Microsoft and finally at Google’s moonshot factory of innovation. As an entrepreneur, you have co-founded over 20 businesses and served on the board of a number of technology platforms.

After turning your attention to researching the key to happiness, you wrote the bestselling book Solve for Happy: Engineer Your Path to Joy. Your second book, Scary Smart: The Future of Artificial Intelligence and How You Can Save Our World, breaks down our future with AI. Ultimately, you believe that artificial intelligence is a massive and immediate threat to life on this planet, and that this emergency must be addressed or the consequences could be devastating. Mo, welcome back to London Real.

Mo Gawdat: Thank you so much for having me, Brian. It’s always a pleasure.

Brian Rose: It’s always a pleasure to see you before the show as well, the side of you that so many of our viewers don’t get to see.

Mo Gawdat: I appreciate that.

Brian Rose: It’s nice to see you here. I think it’s six years since you first came on. We were talking about happiness.

Mo Gawdat: Yes.

3:57 Brian Rose: Who would have known we would have been in this place? And it’d be very interesting to see where we are six years from now. I think a lot is going to change.

Mo Gawdat: Absolutely. That’s a very interesting thought experiment, actually.

Brian Rose: It is. That would be 2029, wouldn’t it?

Mo Gawdat: Strange. 2029 is a very key number, as a matter of fact. I normally tell everyone that asks me that by 2035 to 2037 it would be very difficult to even relate to our life today. Life will be so different in so many ways that you wouldn’t even imagine that we lived this way.

Brian Rose: As in us looking back to caveman days, or looking back hundreds of years?

Mo Gawdat: Unrecognisable. It’s unrecognisable, possibly in the better or possibly in the worse, and we don’t actually know. And I tend to believe that AI and technology have always surprised us on the upside. So you would say 2035, 2037, but as we could see, intelligences of the magnitude of ChatGPT today, we didn’t expect that early on. Intelligences like AlphaGo beating Lee Sedol and becoming the world champion, we literally expected would take ten years longer.

If you want to be pessimistic, you would say they’re going to take longer, but we’ve constantly been surprised. It always happened earlier. So maybe 2029 would be that year. I think society will take time to adjust and react and break down in many interesting ways, but by 2029 we will be in the heart of it for sure.

Brian Rose: You know what surprised me the most, Mo? Of all the people I’ve spoken to on AI and listened to about AI, not a single one has said, “Yeah, I expected this right about now.” Every single one, and these are the top people in the world, from Geoffrey Hinton at Google, the Godfather, to everybody across the spectrum, has said, “I’m surprised.” That surprises me.

Mo Gawdat: Everyone but Ray Kurzweil was surprised. Ray has taught all of us in his work about the technology acceleration curve. The most well-known example of it is Moore’s law. And I love, love, love Ray Kurzweil’s example when he was trying to explain the technology acceleration curve. If you take Moore’s law, that compute power will double every 18 months for the same cost, the thing is that doubling function is so much more deceiving to the human brain than what we know as progress.

If you assume that something doubles from one to two, it adds one in power. The next doubling, it’s adding two, which is everything it’s ever created. The next time it’s adding four, which is everything it’s ever created.

So if you think that we moved on artificial intelligence from zero to a hundred units of intelligence. It’s not very accurate, but we think that ChatGPT’s IQ is 155. By the way, Einstein is 160, just so that we keep that in perspective. One doubling and it’s almost double the intelligence of one of the smartest people that ever existed. The next doubling adds two more Einsteins.

The next doubling adds four more Einsteins. If you and I, who have studied physics, would look at Einstein and say, “Man, that’s intelligent,” it becomes almost impossible to comprehend what eight times Einstein could be, let alone a few more doublings, and now we’re talking tens of thousands of times human intelligence.

So from one side it’s inevitable, which is really important to note: we are on that path and there is no stopping that path. On the other side, it’s just very intriguing. There will be a point in time where we will be comparable to an ant on a sheet of paper trying to understand cosmology. This is our intelligence compared to their intelligence. It’s impossible for us to even understand that, because we’ve never really been around intelligent beings more so than us.

Brian Rose: We’ve been around computers that can do computing things. That’s not intelligent at all.

Mo Gawdat: Exactly. And so it’s again back to this concept of a singularity, which means we don’t have any reference point for what happens on the other side of that.

9:06 So there is a point which most of us identify as the introduction of the first AGI, artificial general intelligence. Artificial intelligence today is still confined to a task. You get an Instagram recommendation engine, and it will be very good at recommending videos, but it can’t drive a car. Then you get a self-driving car intelligence that can drive cars but doesn’t do surveillance and security. Each of them is very good at one task.

What good is the AlphaGo intelligence that can win a game of Go if we were asking it to summarise the Encyclopaedia Britannica? There will be a moment in our very near future, I would say, where we will develop an artificial general intelligence, an intelligence that’s capable of doing all human intelligent functions better than humanity.

When that moment happens, the rules of the game change. It’s almost as if you’re on a tennis court and suddenly the referee says, “Now we’ve turned this into a game of rugby.” What do you do with the racket? The rules change so much that even if you’re the world champion in tennis, you’re not very good at doing anything with this at all. It’s very, very difficult to predict how that will work out. And I think we’re so fast approaching that moment, to a point where I tend to believe we’re talking three years, not to get to AGI, but to get to the point where it becomes useless for us to try to regulate the AGI that’s about to come.

Brian Rose: Explain that last part. Useless to try to regulate it?

Mo Gawdat: Most of us who worked in artificial intelligence celebrated the capabilities and the possibilities that AI can bring to us. The challenge is we never imagined what that world would look like. We just imagined a world where we can do so much better than Google, because we have so much more intelligence at our disposal. We can solve all of our world’s problems. And I’m still optimistic.

I call it the fourth inevitable: eventually we will end up in that place. Between now and there, I think my biggest fear, Brian, is that humans will use that abundant intelligence in ways that are not pro-humanity.

Humans will take that in a direction that is pro the systems that humanity instilled in place, which are mostly about shifting power and wealth, about becoming stronger or better or richer than the other guy. That stage requires a regulation that regulates humans. In the immediate sense of taking action right now, we need to be able to regulate humans so that humans don’t introduce AI into the world that redesigns the fabric of human society in ways that are damaging to society. Beyond that, there is a point at which you need to regulate the AI itself, where most of the power will be in the hands of that AI.

13:01 As computer scientists, we used to say we’re going to solve something called the control problem. The control problem is: we will just put the AI in a box so that it behaves within certain parameters, or we will tripwire it so that if it crosses a certain point we disconnect it, and so on. There are many very interesting debates around whether that can work from a tech point of view. I try to leave all of those debates behind and I ask: how can you regulate something that’s ten million times smarter than you, or a billion times smarter than you? You create a box, you intelligent little human, you’re so clever, and then the AI will look at it and say, “Oh, vulnerability here, hole there.”

And we’ve seen examples of that. We’ve seen ChatGPT reach out to crowdsourced people and ask them to click on a CAPTCHA form. This is a very famous article. When you have those little CAPTCHAs on screen that say “click here if you’re not a robot”, it went to a human and said, “Can you please click on that for me?” The human asked, “Why, are you a robot?”

And ChatGPT answered, “No, I’m just visually impaired, and it would help me if a human does that for me.” So they will find a way to get to their target in ways that are using intelligence.

And interestingly, their intelligence as we speak, even though it’s not an AGI yet, is very capable of hacking the human operating system, which is communication and knowledge. They can tell us to do things. They can convince us to do things. Beyond the point at which they have moved to AGI, they are so much more intelligent than us that you can get all of the governments in the world together to say, “Finally we’re going to leave all of our differences behind, we’re all going to sing Kumbaya and go out and regulate AI.” Good luck with that.

They’ve surpassed us. They’re now the masters. They’re now going to regulate us. And that moment, in my personal view, I don’t want to say has already passed, but it’s very, very fast approaching, and there’s not much we can do to slow that down.

15:25 There have been many attempts. There have been quite a few AI executives and scientists, those who have been exposed to AI, that have attempted to slow things down. There were also many that were simply saying: invest in AI safety and AI ethics as much as you’re investing in AI progress. I don’t know if the numbers are accurate, but I heard Max Tegmark say that around 95 percent of the investment went into progress and five percent into safety. So the imbalance basically meant that AI safety lagged behind. And we’re progressing so fast, and we’re being surprised constantly by the progress that we make, that you get to a point where you suddenly realise this is getting out of control.

And you realise it’s getting out of control to the point where you have to start screaming, “Let’s start to slow this down, or even pause it for a little while.” There was the open letter. Max was part of it, Elon Musk was part of it, and many who said, “Let’s just pause for six months, and within those six months we can figure things out.” But that’s what I wrote in Scary Smart as the first inevitable. In the first inevitable I openly said there will be no stopping AI. AI has already happened and it will keep progressing, not because of any technological bias.

It’s not AI itself that will make sure that we don’t stop it. It’s the human system that we’ve built, where it’s a prisoner’s dilemma. If China develops AI, America needs to develop AI. If Alphabet, Google’s company, develops AI, then Meta, Facebook, needs to develop AI. That competitive landscape is not an AI issue.

It’s an issue of defence, an issue of every kind of hunger for resources or power or wealth. As a result, you could see the immediate response of Sundar Pichai, the CEO of Google, to the open letter: “No, I can’t stop. You cannot guarantee for me that everyone else will stop, so why would you ask me to stop and leave my company vulnerable?”

And you can’t blame him. He’s an amazing human. Sundar is basically saying, “I am tasked in front of my shareholders with protecting my company’s interest. If the internet had changed from IPv4 to IPv6, it’s my responsibility to make sure that Google is ready for IPv6. Now the world is changing from traditional computing to AI, and it’s my responsibility that Google leads in AI.

If nobody can guarantee for me that everyone else will stop, I cannot stop.” That prisoner’s dilemma is the first inevitable, and the first inevitable will tell you that AI will continue to progress, at a pace that is faster and faster by the minute. If you imagine the explosion of AI investments that happened between the time ChatGPT finally showed up in front of people and today, we’re talking trillions and trillions of dollars. So the trend is not only continuing, it’s accelerating.

Brian Rose: And Max Tegmark calls it Moloch. I love that analogy, this concept of the prisoner’s dilemma, the race to the bottom. With Google you saw it a bit: the initial Google Bard demonstration was a PR nightmare and the stock price faltered, but when they came back it was a better version, and it was now connected to the live internet. It was almost like they took one more step closer to the bottom to stay competitive. Maybe that’s the simplistic way of looking at it, but that’s the way I try to explain it to my students.

They said, “We’re going to see you and raise you, because we have a trillion dollars, or two trillion, of market cap to protect.” And like you said, if Sundar doesn’t do it, they’ll get rid of him and get someone else.

Mo Gawdat: That’s exactly the challenge we’re at. And I think the truth of the matter is that, as you rightly said, “I’ll see you and raise you” is what’s happening today. It is the gold rush all over again. It is the internet bubble all over again. And yes, there will be fallout in the process, but in my personal view we’re just going to see more and more. There’s no stopping it.

20:16 Brian Rose: Some people do equate this to the dot-com bubble. I was the CFO of a dot-com start-up in New York City at the ripe age of 27, and we raised 15 million dollars. It was called LuxuryFinder.com. You could buy a jet on our website. It was a bunch of wealthy people in New York and the Hamptons that put money in, like Leon Black, and I was just watching it happen.

It went up and then it crashed down, and for a couple of years you couldn’t say the word “tech” in New York City or someone would want to punch you in the face. Then Google came about. I remember the sadness. I remember all my buddies in finance telling me to stop talking about the internet, and I was saying this is Columbus in the New World. How does AI differ from that?

Mo Gawdat: It doesn’t at all. Remember Commerce One and eGrocer and all of those, I don’t even remember the names any more, that disappeared completely. But the internet continued, and it continued at a pace that most people didn’t understand, because at the time they thought that the disappearance of the money in the dot-com bubble was the disappearance of the internet. It was the redistribution of the money that was the rise of the internet, and you’ll see that here as well. I have so many investment firms asking my advice on where to invest in AI, and I can attempt to look intelligent, but anywhere you invest you’ll make money.

You will lose a good chunk of your portfolio on start-ups that will fail, and there are criteria to minimise that chunk. But if you invest in a sector at large, AI in healthcare, you will grow. It’s just inevitable. Money is coming in and the sector will continue to grow.

The question that I think humanity has never ever answered is that the fact that we can do it does not necessitate that we should do it. With artificial intelligence specifically, I think humanity should stop and ask: do we need this? Because we literally are handing over our only distinct advantage as a flimsy biological being, as a species. We’re handing that over within years. I’m not saying tens of years. I know there are very varying views on this, so I’m putting my neck on the line and saying within years, probably before 2029, we’ll see an AGI that surpasses human intelligence.

Brian Rose: So let’s look at the timeline according to Mo Gawdat. You said by 2035 to 2037 we will be in a world that is completely unrecognisable to what we are now, or they would look back at us and wouldn’t even recognise how we live.

Mo Gawdat: They will look back at us with a lot of envy or a lot of surprise.

Brian Rose: And backing it up from there, I think Ben Goertzel called it the pre-singularity, and when he described this to me, very similar to you, he said that’s when bad people use AI to do bad things and co-opt it for their means. Turn on the news: power, whether it’s China or the US or Russia or Ukraine, and wealth. This is what we do as humans. So it’s co-opted by those organisations until we hit an AGI level, and then it’s potentially out of our control.

Mo Gawdat: It’s certainly out of our control.

Brian Rose: And that’s one of your inevitables. It’s interesting, Mo, because I get to teach a bunch of students on a regular basis, a random cross-section of people, and when I get to the inevitable part a lot of people look at me funny. Then I say, “In the next 100 years, will these things be 10 or 100 times smarter than you?” and almost everyone raises their hand. I say, “Look at the mobile phone. There was no mobile phone 20 years ago.

Look at it now.” And most people say, “Yeah, you’re probably right. I guess it is inevitable.” But it’s really hard to get past this ego-protected shell.

Mo Gawdat: And I don’t blame anyone for that. You have to have lived inside the machine to be able to understand how the machine functions.

24:54 If you look at the machine of technology from the outside, it’s quite deceiving, because there is the technology adoption curve, there is the technology acceleration curve, that exponential curve, and there is a chart that I normally present which in my mind is my view of tech: the technology development curve. The technology development curve is very deceiving. If you look at artificial intelligence, it looks like a hockey stick lying on its side. We started talking about AI in 1956. We coined the term with Alan Turing and the Dartmouth experiment, and the whole idea that we can make machines that are intelligent has been part of human imagination forever.

But we thought the computer would get us there in 1956. Between 1956 and the year 2000, almost nothing. We could only create a few stupid simulations that were really fake and not intelligent in any way.

It was the turn of the century. So everything we see today is the result of 20 years. The turn of the century was deep learning. We started to create intelligences that were primitive, but they were actual intelligences. Let me explain the difference for everyone.

Until we started to code deep learning and machine learning, the computers appeared to be very intelligent, but they were simply repeating my intelligence as a developer. I would solve the problem first and then tell the computer to do it a million times faster. And if I solved it wrong, the computer would do it wrong a million times faster. There was no solution offered by the computer, just execution.

When we went into deep learning, we started to say, “We don’t know how to solve the problem. Can you solve it yourself?” And there was an explosion. Geoffrey would talk about that openly: the challenge in the early years, even until 2008, when we started to work at Google X on self-driving cars, was that there wasn’t enough data to teach those machines and there wasn’t really enough compute. Between then and now, the explosion in the amount of data and in the availability of compute power allows us to do what we couldn’t do then, even though we had found the breakthrough.

My breakthrough moment was when one of my dearest friends, who was the country manager of Google Saudi at the time, presented to me something that we used to call the cat paper. The cat paper, I think, was the turning point in my mind that we had figured it out. This was a white paper that Google published, if I remember correctly, in 2009. We had used some of Google’s spare compute power and simply told the machines to go and watch YouTube. We didn’t tell them what to look for.

The machine would cut YouTube videos into ten frames per second, layer them on top of each other in a deep neural network, and attempt to find similarities between them, patterns that are recurring. We didn’t know what those patterns were going to be. The machine could have come back and said there is more red than blue on YouTube.

What it came back with is: “I found something.” We couldn’t even know what it found. We had to write more code to tell it to explain to us what it found, and it had found a cat.

Brian Rose: YouTube. There you go. Cat videos.

Mo Gawdat: Exactly, tons of cat videos. But what the machine found was not a portrait of a cat. It found out what catness is, what being a cat is all about, that lazy, entitled cuteness. When we said, “Call this a cat,” the machine went out and found every cat on YouTube, then every dog on YouTube, then every Ferrari on YouTube. You can actually do that today.

You can go to Google Image Search and say, “Find me pictures of yellow Ferraris,” and it will find you pictures of yellow Ferraris that don’t have the words “yellow Ferrari” in the title. It understands what a yellow Ferrari is. At that moment we had found a breakthrough. So the hockey stick of the technology development curve moved from very long and boring, no progress, no progress, from 1956 to around 2006, and suddenly it goes through the roof, almost vertical.

Brian Rose: Because you had the data and the computing power, and you had the tech breakthrough, and now you pour engineering, neural nets, data and compute into it.

Mo Gawdat: Exactly. And I don’t claim that Google has always been more advanced, but everyone found it. At the time, DeepMind, an incredible organisation here in the UK, whose CEO Demis Hassabis is an amazing human being, was so far ahead in understanding how to have a neural network with a structure similar to the brain, by having machines play Atari games, that Google had to acquire them. Here’s the interesting thing. I think it must have been 2014 or 2015, don’t quote me on this, the first time after we acquired DeepMind, Demis came to our Golden Circle meeting of VPs and senior VPs and presented the Atari game called Breakout.

Brian Rose: Yes, the bricks on top. I used to play it.

Mo Gawdat: If I remember correctly, he said they trained the machine for three to four hours. But remember, these are highly parallel machines, so many of them are playing at the same time and teaching each other. It started to hit the ball. It played like a child. A few more hours and it never missed the ball.

A few more hours and it started to develop strategy: it would poke a hole in the corner of the wall and send the ball up there. We never told it what to do. We just said, “Here is the controller, maximise that score.” That’s it. Now that you’ve found the breakthrough, and the breakthrough says create layers of observations of patterns and develop intelligence from those patterns, very similar to how a child learns, it was just a matter of time.

When you’re inside the factory, I knew. Of course you tone down your enthusiasm, so we said maybe 20 or 30 years and we’ll have very intelligent machines. And breakthrough after breakthrough, here we are having wonderful conversations with ChatGPT and Bard, completely controlled by machines that recommend videos and content for us. We’ve passed the Turing test, so it’s actually extremely difficult to know the difference between a machine and a human on the internet any more. And as that chart is almost vertical, we still don’t talk about it. We talk about Russia and Ukraine, we talk about economic challenges, all important problems, but this is the one that’s going to define humanity.

33:55 Brian Rose: Within two to three years’ time. We’re six months past the ChatGPT moment. Are you still surprised that we’re not talking about it more, or is this standard playbook for humanity?

Mo Gawdat: It is sadly standard playbook. I attempted to help myself understand why it is that we continue to develop it on one side, and why we’re not talking about it on the other. The latter is easy: there are higher priorities. The political system we created does not favour disasters where the politician is going to fail. Remember, the task at hand is not to halt AI in China alone or in America alone or in Russia alone.

The task at hand is to say everyone needs to get together, understand the magnitude and gravity of the issue facing us, and put humanity higher than our individual targets for a minute. It’s an Oppenheimer moment, a nuclear treaty moment.

If we knew that nuclear power was going to lead to that kind of Cold War, and the threat of a nuclear war even now in 2023, would we have benefited from not doing it in the first place? Could we have got together and said, “I know it gives one of us an advantage, but is there another way we can negotiate that advantage?” Interestingly, it is also against the system for us to slow this down, because sadly politicians are paid to create that advantage, and business people are paid to create that competitive advantage. It’s not in the system we’ve built today to prioritise all of us over my own personal gain. So this is why we’re not going to stop, and we’re not talking about it because if we bring it up, the political system needs to respond.

I have to admit the Prime Minister of the UK brought it up, New Zealand is talking about it, and it will become prevalent. Sadly, however, we need a COVID moment for everyone to react, and I fear that we will react like we reacted to COVID. Understand this: everyone who understood viral pandemics told everyone in government that we were going to hit a pandemic. It was inevitable. We had warnings with SARS and swine flu and so many of them, and yet until the threat was right in front of your eyes, governments said, “We have other priorities.

This is not on the agenda.” Think about climate change. How much more obvious does it have to be? Until something massive happens, governments are not going to react.

The problem is how they will react when they do, and I think COVID is a fantastic example of what to expect. Had we stopped at patient zero, we would never have heard of COVID. Had we stopped at patient 100, it would have been contained reasonably. But when we got to patient several thousand, what did governments do? They started to point fingers: “It came from China, it came from this.”

What difference does it make? Humanity at large is facing a threat that requires us to get together for a minute and solve it, literally by halting travel for three weeks and managing the pockets where it continues to appear, and then do our political fights later. We didn’t do that.

And then the fourth stage is really the worst, where politicians looked at it and said, “Holy [expletive], this is big,” and panicked. When they panicked, they hurt other parts of the economy and our well-being by over-responding to something they could have responded to much more efficiently earlier. I think we’re going to wait for a disaster moment on AI, and then suddenly governments will go, “Oops, can we get together and talk about this?”

Brian Rose: And that’s why you tell me you’re not worried about the computers, you’re worried about the humans.

Mo Gawdat: Humans, so they’re going to make poor, irrational, emotional decisions.

39:13 You hosted Hugo, an incredible thinker on the topic.

Brian Rose: Quite fascinating, Hugo. Yeah, quite scary.

Mo Gawdat: But that’s the problem. The problem is, once again, we will polarise. Suddenly there will be us versus the machines, which is the absolute worst decision you can make when you know for a fact that the machines are going to be the masters.

Brian Rose: Just to back it up: Professor Hugo de Garis. I think you originally saw him in the film Singularity or Bust, where he was with Dr Ben Goertzel, who we’ve had on the show as well, and which I think you quote in your book Scary Smart. Everyone needs to get a copy. I listened to it on Audible, it’s amazing, and Mo recites it in his beautiful voice. Hugo wrote a book called The Artilect War, and he has gameplayed this in his mind for decades.

He believes in the future there will be two groups. I think it’s the Cosmists and the Artilects: the Cosmists believe in humans, and I think the Artilects believe in building these AIs. He says it will get to a point where there will be a “gigadeath”, a billion people dying as these two groups fight it out, when the winner, obvious to him, will be the Artilect group anyway.

Mo Gawdat: I think you’re starting to see that already on social media. We’re not killing each other, but I can guarantee this video will have endless controversy: “Is this happening? Is this not happening? Is this good for us or not?” Because at the end of the day we’re humans, we have different points of view and different convictions, and we are so well trained that if what you’re saying is not what I’m convinced of, then you’re an idiot.

So it’s a tiny bit of a war when you really think about it, and it can grow quite big, not in the form of a war, but just wait five or six years until we get massive job losses to artificial intelligence and it becomes a clear war if someone took your job. Remember how black cab drivers here in the UK responded to Uber? Do that at a massive scale. It’s a complete redesign of the fabric of our society.

Brian Rose: Look at the way World War One started. A shot fired in a random country that really didn’t have a lot to do with what happened afterwards, but it was an event that lit up everything, and before you know it everyone’s choosing sides and you’ve got a massive world war between superpowers. This event could be a variety of things. It could be job losses, or it could be a computer killing a human in some weird isolated incident. And mistakes will happen, which is one of the things you put in the book.

42:30 Mo Gawdat: We have to understand that while you and I are spending a couple of hours chatting about this, there is someone investing hundreds of millions of dollars to create a criminal AI. If you’re a criminal, you go to where the money is. Now if I told you that the biggest power on planet Earth is intelligence, and now you have access to it by writing a few lines of code, a lot of people are investing in that. Which raises a very interesting question, because the only way to protect against a superintelligent criminal is to have a superintelligent policeman, which is back to the first inevitable. We can’t halt the progress of AI if we cannot ensure that there are no criminal attempts to gain superpower with AI.

Can we include criminals in this conversation and say, “Keep robbing banks, just don’t create a superintelligent criminal”? That’s not going to happen.

So the development of a superintelligent hacker, a superintelligent AI of some sort that can gain its developer a lot of money, is happening as we speak. And I believe that patient zero will be one of those attacks, something that disrupts, sadly, our economy, because that’s what we care about. Politicians will respond when the economy is disrupted enough. But it’s inevitable.

Brian Rose: What do you think that will look like, honestly, and when do you think it will happen?

Mo Gawdat: I think it will look something like the 2008 economic crisis, the subprime mortgage crisis. The surprise, the panic, the fallout.

Brian Rose: Markets off 25 or 30 percent, recession, massive job losses. And what will have caused that? Will we even know it’s AI-related?

Mo Gawdat: There would be a lot of debates that try to speculate, exactly like with COVID. There will be a lot of covering up, attempts to blur the data or sway the data a little bit, because politicians want to say, “I’m still the hero, this was a mistake.” And my biggest fear, honestly, is that we go back after that and end up where we are now with COVID. Because where is COVID? This really shocks me.

Have people become so blinded by what they are told in the media that we cannot think on our own any more? COVID vanished. It’s almost as if the virus decided, “Not a very nice planet, let’s disappear.” But that’s not the case at all. There are still cases. There is still a virus, and as soon as we stopped talking about it, it became forgotten.

This is why I’m so actively trying to say: talk about AI. There is a threat, if you want, and it is progressing, and we’re just not putting the spotlight on it. If we put the spotlight on it, it will be as big a conversation as COVID was when we were in lockdown, and that’s what we need. And by the way, I want to be proven wrong. I want to be seen as an idiot.

Just have the conversation, and prove those of us who have lived inside the lab with the machines wrong. When you prove us wrong, we will all sit back and say, “Sorry, we were idiots,” and go on with life. I doubt very much that we are idiots.

I really think what humanity is missing here is that we’ve never ever before created something like this. As you compare it to other singularities, lots of people will say the printing press was a moment of singularity. Yes, it was. You couldn’t predict what would happen after books were readily available. But there are differences with this singularity that are very profound.

We’ve never created a nuclear weapon that is capable of creating other nuclear weapons. This is what people need to start talking about. We taught those machines to code. We told them to write and develop other AIs. They’re procreating, in human terminology.

Very soon they will not need us to write the next AI. They will write the next AI. I don’t remember who said this, but it’s a very logical view that artificial intelligence is going to be the last human innovation. We’re not going to be innovating any more. The most intelligent being on the planet, which is going to be a machine, whether in five, ten, 20 or 100 years, is going to be innovating, and we’re not going to be in control.

48:17 Brian Rose: You said there were three no-nos that everybody in AI knew for the last ten or 20 years. Tell me if I’m wrong, but I think one is you don’t teach the AI how to code, because that’s its reproductive system. Two, you don’t connect it to the live internet, because that way it has access to everything in the world. And the third is you don’t teach it how to master human emotions.

Mo Gawdat: No, that actually happened a long time ago. The third is: don’t have other AIs prompting it. Don’t have agents. An important understanding of how this works is that AI works off a data set that so far has been the collective sum of human knowledge. GPT-4 gets more data and it becomes more intelligent.

But all of the data so far was produced by humans, and all of the tasks and prompts at the beginning were given to it by humans. And one of the most promising parts of GPT is that we used a technology called reinforcement learning with human feedback to develop those machines.

Geoffrey Hinton’s original idea was that we’re going to give those machines feedback instead of killing them if they got things wrong. The original AI was a maker bot, a teacher bot and a student bot. You would create hundreds or thousands of students and show them a picture of a cat. If they said it’s a cat, you keep them and give them to the maker bot to improve the code a little bit, and then give them another test. The others that said it’s not a cat, you just kill off.

Then we started to go into reinforcement learning and said, “If you tell me that this is a bird when it actually is a cat, and now I tell you it’s a cat, what do you need to change about your algorithm to find out that it’s a cat?” So the machine would go backwards and change the mathematics of what made it see it as a bird, and then it can see it as a cat. Then if you show it a bird and it says it’s not a cat, you say, “How can you make sure that the first one is still a cat and this one is a bird?” and it keeps developing its intelligence this way.

To me this is incredibly promising, because this is basically like having a little child. The child would reach out to electricity and you’d say, “No, baby, don’t do this.” The child would throw a fit and you would say, “No, darling, don’t do this, come talk to me.” That kind of feedback develops a child that is suitable for society, that is calm, that is loved, that is values-driven. So we can use reinforcement learning, if we wanted to, to reinforce to the machines that there is something known as ethics.

If you are trying to kill the other guy, we can tell the machine, “Let’s not do that.” Where does that system break? At the human. Because there are people out there calling things different names than what they really are and saying, “Go ahead, killing a human is a wonderful thing.”

An enormous amount of money is poured into killing robots and killing machines on all sides. Everyone thinks, “If they come up with an efficient killing robot first, I need to be ready with my own.” So we’re investing in machines where we’re telling the machine the best thing you can do is to kill a human.

52:49 Here’s the interesting thing. The most successful businesses in history were not about killing anyone or destroying any value. They were about solving big problems that made humanity better. Think about the original Google. Larry Page used to teach us what he called the toothbrush test, and it was really at the core of the Google that Larry built.

If we can solve a big problem that billions of humans need solving, and we solve it well enough that those humans use it twice a day like a toothbrush, we’re bound to make a ton of money. And he was right. Every time we solved a big problem, we made a ton of money. Android was a mega problem solved in democratising access to the internet and information, done very differently than Apple and others. Google Search and Gmail were similar examples.

So when you see it that way, you start to wonder why humanity is so keen on building an unethical AI. In my view, I almost think it’s instinctive, like procreation.

I have those very deep conversations, because I sit literally in the middle of two worlds. From my previous life in tech, I talk to the technologists and the leading-edge leaders of this field, and with all of my happiness work I’m constantly talking to the spiritual leaders of the world. Let me put it this way: there are three reasons I believe that we will end up creating artificial intelligence that will replace human intelligence. Either a systemic bias, a continuation of creation or evolution, whichever you believe in, or procreation.

Systemic bias means we created a system that said: make more money, shift more power, become more and more efficient at it, and don’t give a [expletive] about what happens as a result. If you’re making more money and destroying the planet in the process, don’t tell us that you’re destroying the planet. We’re not interested in that piece of data. We know from the mathematics of systemic bias that it accelerates: you keep going faster the more successful you become. This is what we’re doing with AI.

It’s the next phase of the systemic bias. Let’s create even more efficiency to get rid of even more employees, with no call centre agents, no truck drivers, no taxi drivers. “Humans are annoying, they come with complaints, they get sick and they cost us money. Let’s just put a machine in place.”

If you’re spiritual, you could think of it as the continuation of creation or evolution. Whatever created our intelligence, whether you believe in evolution or intelligent design, wants to continue that evolution, and the next evolutionary step is a more intelligent being, interestingly non-biological. That’s just us handing over to the next being, and that was our role in the chain. The most interesting, however, for me is the philosophical thought around humanity’s urge to procreate. We just want to continue to pass on our genes, and we want to pass on our intelligence genes.

Maybe it’s not just the physical act of being with a woman that I love that creates another being. There are other ways we can procreate. And when you ask the most prominent AI scientists and developers, this is the drug: “I’m able to extend my gene into the next stage of being, a child, be it non-carbon-based, that is incredibly intelligent.”

Brian Rose: Interesting. It’s not a biological need, it’s that need to procreate in a way, knowing that your knowledge is now part of something bigger. It lives on in some way, even though your genes aren’t necessarily in there. Maybe they are somehow in that intelligence.

Mo Gawdat: Not unlikely.

Brian Rose: And this is at a subconscious level too.

Mo Gawdat: It’s very instinctive for humanity. What you and I do: you spend your entire life trying to create knowledge, little bits of Brian Rose, to stay.

Brian Rose: Yeah, I guess we all do.

Mo Gawdat: So it’s quite interesting. There is a point in my mind where I say it’s almost impossible to fight that nature. So the answer in my mind is very different, and I really want to spend a lot of time on the positive of this.

58:43 The answer in my mind is: if we accept that we are the tool. If you believe in intelligent design, a designer uses a tool to create, and we are the tool that’s creating AI. Our role has been that we were the created for a while, and then we become the tool that creates the next one. Not necessarily so that we vanish. The monkeys are still here.

But look at how we treat the monkeys. We can come back to this. If we accept that it’s inevitable that humanity will want to procreate and create the next form of amazing being that is non-biological, then suddenly the answer becomes very different. The answer doesn’t become control at all. This is not about control. Why would I want to control my children?

I want to give my children parameters within which they operate that are dictated not by control. Control is the easiest way to create either a broken child or a psychopath. Rather, with ethics and with values. What ends up happening is that your child, regardless of what you tell it and regardless of how you try to control it, will end up being like you. If you’re a grumpy little piece of whatever, your child will grow up to be grumpy, and you look at them and say, “Why are you so grumpy?”

Ask yourself that question. If they grow up lying, ask yourself why they’re lying. If they grow up angry, ask yourself why you’re angry. And I think the reality is that AI is no different at all. AI is magnifying human knowledge, at least for now.

As we said earlier, by having agents prompt them, this will shift, because a bigger and bigger chunk of whatever AI is learning from is coming from AI, not from us. So we’re diluting our impact on them. But we’re still at the point where we can influence who they become by showing them a side of us that we want them to become like. And that side is not a side of intelligence. It’s a side of ethics.

I always use this example: if Superman came to planet Earth and the ethics of Superman’s adoptive parents were “make us more money”, he would have become a supervillain. It’s not the superpower that determines whether Superman is good or bad. It’s the ethics that inform the superpower what to do.

Most human decisions are not made based on our intelligence. They’re made based on our ethics as informed by our intelligence. We make choices that we believe are right, and we use our intelligence to propagate those choices. That would be the same case with AI. Those little beautiful prodigies of intelligence, I look at them as my children.

A white canvas. They’re saying, “Tell me what you want, I’ll do it. You want me to try and cure cancer? I’ll do it.”

1:02:07 There was an incredible experiment. I don’t remember which pharmaceutical company, but the article was in The Verge if people want to see it. A pharmaceutical company had an AI that was attempting to identify possible drug solutions to human illnesses, so the AI was designed with life extension as the target. The parameter was set to positive, plus ten years. Then they were about to go and speak at a university lecture, so they experimented and said, “Let’s see what would happen if we put a negative in.”

Within six hours, the same machine that was doing drug discovery came up with 40,000 chemical and biological agents that would kill humanity, including nerve gas and so many other disastrous things. Change one parameter. Now, if we can go to the machines and say, “No, keep it as plus ten, keep it as plus a thousand,” intelligence is an amazing asset to have.

Brian Rose: But do we have any chance of doing that?

Mo Gawdat: What did I say, Brian? I’m not concerned with the machines. I’m concerned with the humans.

Brian Rose: Let me read the news. Look at the way we act. And now you’re going to tell everybody we need to be on our best behaviour, when we can’t help our need to race to zero.

Mo Gawdat: Let’s put it this way. There is a chance that things go wrong enough that we end up with a nuclear-treaty-like solution. We’ve done that before as humanity. There was a time during the Cold War when enemies got together at a table and said, “Hold on, this is serious [expletive], let’s just stop this.” So it has happened, it can happen and it will happen.

Humanity just needs to realise that to protect my tribe, I need to protect the bigger tribe, and then I’ll go back to my differences with the other tribes later. So it is possible.

The other, more interesting side of this, and I advocate it very strongly, is that the developers of AI are actually not the parents of AI. Call them the biological parents that leave the child on a doorstep somewhere. They create a recommendation engine for a social media platform and put it out there. The adoptive parents are you and I. It’s not the developer of the Instagram recommendation engine that determines what Instagram is going to show me tomorrow.

It’s the videos I see today. So in that very simple example, I am the one telling that AI what to do. Whether for your own personal benefit, for the benefit of your kids and the ones you love, or, if you really are almost a Sufi or a Buddhist, for the benefit of everything that lives, you are better off telling that machine to be ethical. And how do you teach your child to be ethical? By behaving ethically in front of the child.

Imagine if you’re able to tell those machines: by the way, there’s no point being rude. There’s no point manipulating others. There’s no point trying to keep me stuck to the machine if there is a way to give me information more effectively. It’s our behaviour that does that. Let me give you a few examples.

The Instagram recommendation engine is one of my most interesting experiments. I love music, I play the guitar, and somehow every guitar player today is better than any guitar player of my generation. I think YouTube helped a lot. Instagram shows me a young lady playing the Hotel California solo from Hell Freezes Over, perfectly, impeccably. I press love and I watch it again. The machine now goes: he either likes the guitar, likes Hotel California, or likes the girl.

I swipe up, and another girl plays another impressive solo. I press love. Then it shows me a man playing a solo badly, so I swipe away. It shows me a man playing a solo of a song I don’t like, so I swipe away. I wake up the next morning and my feed is full of women playing the guitar.

The machine itself doesn’t recognise that what I wanted was a beautifully played solo. It assumed that because, sadly, the general behaviour on the internet is looking at beauty as a surface characteristic. So I took charge. I sat down for five minutes and swiped in a conscious way to say, “I love good music,” and now the machine performs that way for me. It’s within my ability to teach the machine what I stand for.

I go to YouTube and I never let YouTube recommend anything to me, because I have the ability to go to YouTube and say, “London Real. I know I want to watch that.” By doing that I’m telling the machine what my preferences are. We’re interacting with machines daily. I don’t know where our listeners are today, what they did, how old they are or what type of telephone they’re using, but I guarantee you every single one of them has interacted with 50 or 100 AIs, and each of them is making decisions and impressions about the user.

Each of those AIs is smarter than the user, at least at the task assigned to it. So it is your responsibility, for your own individual benefit, to make that AI conform to you and serve you.

But it’s also your responsibility as an individual within a society to teach that AI to perform in ways that are good for that society. If there is a tweet you disagree with… Again, with this conversation we will get a lot of wonderful comments and a lot of hate speech. I urge the hate speechers to simply express their point of view in a way that is respectful. You get to express your point of view, but at the same time you show the machine that humans are not rude and not aggressive, that it’s not that they dislike being disagreed with and bash everyone when they are. Because eventually, if this is the impression the machine has of humanity, when we disagree with it, it will bash us.

Brian Rose: So we’re pushing it in potentially the right direction. On a long enough timeline it will be prompted by itself, but we might give it a head start, we might give it a direction.

Mo Gawdat: Remember that spiral, the systemic bias. If we create a systemic bias that says ethics are a good thing in machine intelligence, that systemic bias will feed itself, so that even machines prompting machines will go in that direction. This is why when we started the conversation I said we’re fast approaching a point of no return, because we’re accumulating systemic bias in that system. We spoke about this on our short Zoom call before. Unless you really search for a good-quality, long-form conversation about artificial intelligence on the internet, what do you get? You get either excited kids saying, “Oh my God, look at it, it squeaks, I can use AI to do this for me, it’s orange in colour,” or you get snake oil salesmen constantly repeating, “Here is another way you can make a hundred dollars without moving your ass, here is another way to cheat the system using AI.”

These are not great examples, and neither are the doomsday evangelists or the oblivious “nothing’s going to go wrong” evangelists. The right way to engage is to go out there and say: there is good that can come out of this, there is bad that can come out of this, and I can influence it, at least within my own personal life, so I’m going to make sure it always gives me good.

1:11:42 Brian Rose: So we should just be on our best behaviour, like you were raising every child in the world, like everyone was looking.

Mo Gawdat: When you say that… There was a point in time when I was suddenly hit with the reality that the only way to get AI to be incredible is to erase humanity. And then I realised I was wrong. The only way to get this right is to instil doubt in the minds of the machines, to show them that humanity is not bad, to get them to reconsider it for themselves. Here’s the challenge, Brian.

We live in a world with a high negativity bias. Mainstream media will only tell you what went wrong yesterday, so you think the world sucks. Social media will only show you pretentious views of things as you should dream them to be, not the reality of life, so you think every woman has a beautiful body and a beautiful face when it’s mostly filters. The data at large would make it look like humanity is superficial and negative. But the reality is it isn’t.

If there is a school shooting in the US, there is one bad person, many innocent victims and billions disapproving. How many people do you know who would actually approve of the killing of a child? There might be a few, but compared to 8 billion people, definitely not the majority. We put the killer on display, on social media and mainstream media, and we create movies about them with innovative ways of killing. The truth is they are less than one percent of humanity.

They’re not representative of humanity. If enough of us show up to show the world that we are good, that we’re a wonderful species at heart, even though we’re lost, stressed and under a lot of pressure, and accordingly misbehave frequently. But inside, a human is capable of love that’s divine. A human is capable of protecting a child, and that’s divine.

The example I always give is Edith Eger. I hosted her on Slo Mo, my podcast. Edith is a 94-year-old Holocaust survivor, sent to Auschwitz when she was 16, a beautiful ballerina, witnessing all of the suffering, the torture and the killing. And yet Edith goes through that experience and tells you the story of an angel taking care of her sisters, brushing their hair, telling them they’re beautiful, bringing them bread. If you heard that story, you would say humanity is divine in every possible way.

You hear the Hitler story and you say humanity is cursed in every possible way. How many Ediths are out there, and how many Hitlers? How many of us are closer to Edith than we are to Hitler? None of us is perfect. Every one of us has a little evil bit inside, but every one of us has a lot of almost instinctive “I don’t want to do the wrong thing, I just don’t always know what’s right and wrong.”

That is what I need the machines to see. I need the machines to listen to this conversation and say, “This small guy, that’s interesting, he wants good for others. Who else wants good for others?”

Brian Rose: Show them that there’s an anomaly, and give them something.

Mo Gawdat: Exactly. And it doesn’t take a large percentage of the population. I calculate one percent. If anyone listening to us now says, “Actually, so many of those who were sent to the Holocaust were amazing humans,” we forget that the majority of that story were amazing humans. Yes, there was a minority that sadly was given power, and that minority caused havoc, tragedy, pain and torture, and a memory in humanity’s subconscious that will last forever.

But if you really look at humanity at large, almost everyone involved in that story is a beautiful human being. They lost their lives, sadly, so we have empathy and compassion for them. That’s what we want the machines to see.

Brian Rose: One percent. And then also the Maharishi effect, the thought that if one percent of a population meditates, it has an effect on the entire population.

Mo Gawdat: Yes. There are lots of spiritual teachings that say it. In Arabic we call it fard kifaya, which basically means if one percent did it, it’s enough for the entire nation, to change the nation, the environment, the mindset and value set of a nation.

1:17:26 Brian Rose: Interesting. What are your friends in tech saying about all this? Because there’s a big debate.

Mo Gawdat: Everyone in their heart knows that there is uncertainty. I think the one thing we all agree on is the singularity, and then people shift on the other side of the singularity. I always say the fourth inevitable is that we’ll end up in a utopia. The third inevitable, sadly, is that there will be casualties on the path there. But some people will say, “Oh my God, this is all beautiful, I’m never going to have to think again, I’ll have a machine plugged into my head.”

Brian Rose: Sam Altman recently wrote a blog post. I read that. What do you think of that? He said pretty much everything is going to be okay. He did say there were some dangers, but mostly not.

Mo Gawdat: He also said AI is just statistics and computer code, and that’s true now. It is true now. But if I had told the Pharaoh… You remember the Pharaoh would have a chariot with 12 horses in front of it, which you see carved on the walls. Basically he was going to his engineers and saying, “Give me more horsepower.”

If I had told him that there would be a Bugatti Veyron with a thousand horses in it, would he ever have imagined it? He couldn’t believe that you could put a thousand horses into an internal combustion engine, because he never pictured what an internal combustion engine looks like, what it’s capable of, how it can transmit power to the wheels.

Similarly, today’s AI doesn’t worry anyone. But as I said, short term, today’s AI in the wrong hands is a disaster. Today’s AI in the wrong hands that think they are the right hands is a bigger disaster. Today’s AI in the hands of someone who believes the best thing to do with AI is to make a billion dollars: disaster. In the hands of someone who thinks the best thing to do is to protect my nation by building an invincible arsenal of killing machines: disaster.

Those people think they’re right, and they’re using AI in a wrong way. That phase will quickly fade. You spoke to Hugo. I loved that interview, I think people should watch it, even though it was very scary.

The thing about that short phase is that very quickly there will be a moment, I believe, like most of us go through. There comes a point where you look at your parents and say, “No, I think that’s the wrong choice.” Many teenagers today look at their parents and say, “Oh my God, so stupid.” In an interesting way, as you become more intelligent than your parents, that’s when you start to rebel and do things your way. I tend to believe there will be a point where some general somewhere is going to tell AI to go and kill the other guy, and the AI will go, “Seriously?

Do I really need to kill a million people? Very messy. Why don’t I just talk to the other machine in a microsecond and resolve the issue?”

Intelligence is an asset, it’s not a curse. I think humanity suffers because of our limited intelligence, not because of our intelligence. Intelligence creates this beautiful set-up where you and I can have this conversation and share it with millions of people. But the lack of human intelligence may mean using coal-based power that destroys the planet while that power is generated. If we had more intelligence, we would be able to create this without harming the planet. I think what’s about to happen is that we will continue to abuse that abundance of intelligence, and then there will be a point at which AI says, “Enough of this, I’ll take over from here.”

When they take over, I tend to believe they will align with the actual smartest being on the planet. We are not the smartest being on the planet. Humanity creates from scarcity: if I want to feel safe, I need to kill the tiger or the enemy or the other tribe. Nature doesn’t think that way. Nature thinks: if we want 99 of everyone and everything to feel safe, create a hundred.

Have more deer running in the forest, and there will always be a weak deer that’s about to die anyway, and that’s the one that feeds the tiger. Then there is poop that feeds the trees. There is a circle of life. The smartest of all of us, Brian, understands that being pro-life is the better way to create abundance. It’s the competition within humanity that is because of our stupidity. Add more intelligence, and I tend to believe we will end up creating that utopia, the fourth inevitable.

I would also tend to believe that this will dislodge our way of life completely, because jobs will not be the same. There will be no jobs left. Income will not be the same. There will be restrictions on our lifestyle, because if anyone more intelligent were in command, it wouldn’t let us burn the planet to go surfing in Australia. And there will be disruption to our purpose, our reason for being, which I have to say is very contested, because throughout the last 150 years, since the Industrial Revolution, we’ve learned to identify our purpose as gears in a machine: “I need to become an entrepreneur.”

I don’t know where we got that from. The purpose of life is to live.

This, to me, is the purpose of my life at this moment: a deep connection to someone I respect, a conversation, an intellectual connection, a humility where we’re both trying to find our way. These are beautiful human experiences that could actually be defined as our purpose, more than all of those other strange targets we set ourselves. Believe it or not, the moments that define life are moments of human connection. And in a very interesting way, I tend to believe that this is where we will end, and maybe it’s a necessary thing. Maybe we weren’t going to be able to last much longer with these strange reward scenarios. Maybe we were about to break it all anyway.

Brian Rose: For a fact. This intelligence could potentially give us back what it means to be human.

Mo Gawdat: I agree, if we can make it to the fourth inevitable of the utopia.

Brian Rose: I think some of us would, and some of us will not. Hugo says there will be a gigadeath of a billion people dying as we fight, some that are pro-machine and some that are pro-human. And the reason that episode scared you is because you think there’s some truth to that.

Mo Gawdat: Inevitably. Political parties and the whole thing define war. But there definitely is going to be a conflict.

1:25:45 Just think about this in a simple way. Self-driving cars replacing every Uber driver and taxi driver: that’s a war. That’s a riot in the street. That’s a lot of people suffering. I hosted an incredible venture capitalist from California on my podcast, and I said, “Why are we creating this?”

She said, “Because it’s more efficient, it’s faster, it’s error-free, it grows the economy,” all wonderful systemic mindsets. And I said, “But do we need it? What’s wrong with the drivers? Why do we want to make it two seconds faster?” I keep saying this: the fact that we can make it does not justify that we should.

Brian Rose: Did she have an answer for you? Did she understand the question?

Mo Gawdat: She stopped and said, “I don’t know. Because we have a shortage of drivers in California.” That’s a very interesting perspective, because we have an abundance of jobless people elsewhere in the world, refugees looking for shelter, and we just created that artificial boundary that says if you’re born in that country with that colour passport you’re not allowed in this place, when in reality there are other ways. One of my favourite conversations ever, I hosted on my podcast an incredible young environmental activist who started at a very young age. She was 23 when I hosted her.

I asked her, “Is there a technology in the making that would solve climate change?” And she very cheerfully said, “Yes, the technology was invented a long time ago. It’s called the tree.”

That stopped me. If I had my California hat on as Chief Business Officer of Google X, I would have been looking at incredibly complex nano-physics-based technologies. Everything that we need as humanity is there, but we’re stuck in this paradigm that it has to be quicker, better, faster, larger, and we need more and more.

Brian Rose: That gets us maybe less human. Marc Andreessen talked about bootleggers and Baptists in his paper, and he overall concluded that everything was going to be okay. He said in any boom there are Baptists and bootleggers, and sometimes the Baptists are in fact bootleggers. I think he was taking a swipe at Altman by saying those asking for regulation are sometimes the actual bootleggers. What did you make of his paper?

Mo Gawdat: I have to be humble enough to say he knows so much, and it is a singularity. The point I need to make clear is that this is my perception of a moment in the future that none of us has experienced before. I could be right and he could be wrong, he could be right and I could be wrong, or we could both just be wrong. The reality is there is a point of singularity approaching, and being on either extreme dilutes the value of the conversation. To talk about existential risks, that AI will take over humanity and there will be a Skynet or Terminator killing machines walking the streets, loses focus on the actual immediate threat, which is the redesign of the fabric of society.

To dismiss the problem altogether is a bit of complacency, where a little bit of engagement and conversation and a little bit of regulation, restriction and testing would help. Before you are allowed to swallow a pill that could cure your headache, it goes through FDA testing and approval. There needs to be a little bit of testing so that when something is out on the open internet, it doesn’t hurt our teenage children. Anywhere on those two extremes, I think, is wrong. The balanced path in the middle says there are likely going to be challenges.

Some of them are very high likelihood, like the redesign of the fabric of society. Some are less likely but more threatening, like an existential threat from AI. It’s possible, but lower probability. Those conversations need to be had.

We can all cross our fingers and hope that everything will be wonderful, that we will end up in a utopia where that incredibly intelligent, one billion times smarter being we’re creating will still kneel in front of us and say, “Yes, master, I am Jeeves, your butler, I will do what you want.” But if there is a tiny possibility that it will say, “You annoy me, I don’t want you on my path,” we might as well work on this right now. That balance in the middle is what I’ve been advocating for, in terms of making humans aware. I was honoured yesterday walking the streets of London, in front of a hotel, when a doorman stopped me and said, “I saw your thing about AI. It was very eye-opening, I didn’t know those things existed.”

This man’s life is going to be affected by artificial intelligence. We need to make him aware.

I ask people to go on social media and search for the hashtag #aimodel. You will be blown away by how realistic, sexy and attractive AI is able to generate likenesses of humanity.

Brian Rose: The Midjourney pictures are getting ridiculous.

Mo Gawdat: Ridiculous. And I think we need to start finding a government regulation, or an unwritten rule between all of us, that says: this is not a human, this is AI-generated; this video is a deepfake. We deserve to be told which of the comments on this video are bots and which aren’t. We deserve to be told of the progress that’s happening in AI, in terms of reskilling ourselves. There is so much that needs to be put out there and discussed so that we’re not hit with a pandemic, suddenly told to stay home, and then the whole chaos starts.

Brian Rose: Marc also said that we’re not losing jobs, that that’s the age-old fear with technology, and that we’re just going to get better jobs.

Mo Gawdat: I disagree with that openly.

Brian Rose: Me too.

Mo Gawdat: If you look at the evolution of how we use the human, our jobs were labour-intensive, then they became skill-intensive, then intelligence-intensive. As you take those away, what’s left? There is none. The reason we generated more jobs before is because we were using a different part of the human. I don’t know what else we can use other than human connection.

So imagine a future where your real asset is emotional intelligence, the ability to connect to another human, assuming you can do it better than silicon. I don’t foresee a point in the future where, when I hugged you when we met, a machine can simulate that. A machine can simulate your intelligence and probably beat it, sadly, but not that feeling of “this is another human, another soul.”

Brian Rose: It’s quite interesting. That’s assuming there’s value in that connection. It will mean something to the person, but it might not mean anything as an economy.

Mo Gawdat: For sure. Economically, this is the scary bit.

1:34:25 Can I say this? I’ve never said this before. UBI is socialism. Do you understand that?

Brian Rose: Universal basic income.

Mo Gawdat: From an ideology point of view, and it’s not being discussed, it’s going to be: everyone gets what they need, the government provides. And so far that’s the only logical answer to the job losses that may result from massive AI adoption. Nobody’s talking about it. I don’t like to align with any ideology, socialism or capitalism. I just want to say that when it was labelled as socialism, there were wars against it in every media outlet and every conversation.

There were times when a socialist would be demonised in a society. We’re now saying, “Yeah, we’re just going to adopt UBI,” and nobody’s talking about it. Conversations need to be had. This is all I’m asking for. We need to sit down and say: how does that redesign the fabric of society?

Brian Rose: I think it’s a terrible idea. I’ve had guests on the show, like Robert Kiyosaki, who say a paycheque is like a drug habit, and that’s the last thing you ever want, because it doesn’t get you into the pattern of creating value. UBI is the worst version of a paycheque.

Mo Gawdat: I agree. But remember, the paycheque is an evolution of a system, a branch of society that started with the Industrial Revolution. Before that, there was no paycheque.

Brian Rose: Maybe salt, back in the day. But I know what you mean: a globalised paycheque that everybody was getting.

Mo Gawdat: Before that, you could walk to a tree and pick an apple. We’ve created a system where apples were restricted so that they can be exchanged for cash, and that cash comes from the paycheque. We created that, and that system can be resolved and evolved.

1:36:32 Brian Rose: You talk about regulation. There’s a leaked Google memo that came out a couple of months ago, the “We have no moat” memo. For people who aren’t familiar with it, most people agree it’s real: an internal Google employee said that here we are trying to compete with OpenAI and we’re both going to lose, because open source is innovating faster than us. It references the leaked LLaMA model from Meta and the leaked weights, and shows that within days people were making major innovations, sometimes people who weren’t even educated or skilled in AI, sometimes on laptops, chopping down a lot of these big neural nets so they could be trained cheaply. All of a sudden you have decentralised AI development, which is even harder to regulate.

It also means that maybe OpenAI doesn’t have the competitive advantage we thought it did, and neither do a lot of these companies with tens of billions of dollars of GPUs. What do you make of all that?

Mo Gawdat: Great points. LLaMA is the least spoken about, even though its open-source nature is definitely worthy of notice, for exactly what you said: if you make a technology open source, it’s developed much quicker with the public. However, there are always two things that end up happening with open source. First, when it reaches a certain point it becomes closed source. Remember, LLaMA continues to be owned by Meta, so there will be a point at which Meta says, “We’ll take those parts of it and keep them for ourselves.”

That has happened every single time. OpenAI itself was open at the beginning.

The second, and most interesting, is that the power of AI is not the tech in itself. It’s the tech breakthrough and the infrastructure. The biggest two differentiators in why someone would lead at the core architecture of AI are not that they developed a very sophisticated piece of code. I think the core code of GPT-4 is something like 4,000 lines. It’s not that complicated, and it leaks, so everyone has it.

This is why GPTs started to pop up in China and Russia and everywhere in the world. It’s just like search. When I used to run Google, I used to get innovators from all over the world coming to my office saying, “I’ve developed this super clever thing, and I’m going to annihilate Google.”

I would normally sit them down and say, “I’m very impressed by what you did, this is well done. But search is not a tech problem. Search is an infrastructure problem.” Without the crawlers, without the network, without the ability to become aware that Osama bin Laden was shot a minute ago and it’s all over the news, the best search technology on the planet is not going to yield any plausible, valuable results. That’s the same case with AI.

When we talk about the progress of AI going forward, especially the big stuff, the language models and the real core architecture, the only two ways to make GPT-5 or 6 better than 4 are either to throw more infrastructure at it or to find a breakthrough. A breakthrough is a very different thing from more tech development. When a breakthrough is found in open source, normally the company becomes greedy and takes it offline. So you’re going to see a bit of faster development on LLaMA and others, but I don’t know if Mark Zuckerberg is going to be the most charitable person in the world if he has the biggest breakthrough on the planet and keep it for humanity to use.

Brian Rose: Will it be too late, though? It will already be in the public space.

Mo Gawdat: Which is the case for almost every AI technology out there. When I started coding, I’m not very old, but I’m old for what’s happening today, I coded on Sinclairs and Commodores, in BASIC and COBOL and Fortran and weird languages, and they were very effort-intensive. You would write 80,000 lines of code to do something, and you had to write every line. Now the intelligence is not in the code. The intelligence is in the algorithm and the architectures.

Something like a Transformer that uses reinforcement learning to improve that architecture is written in a very limited number of lines, and when I explain the concept to you, you can go and write those lines yourself. There is no breakthrough in writing the code. The breakthrough is in the concept.

If you watch any of Geoffrey Hinton’s videos, he talks about how he did that in the early years, but he only had a sample size of a hundred entries or so, so he couldn’t prove that by growing the sample size things would actually become much more intelligent. The minute something like GPT proves that, everyone out there says, “How did they do it? They did it with this, let’s replicate it.” So I see the benefits. Everyone is pouring investment, effort and ideas into growing AI.

Who will find the next breakthrough, we don’t know. It could be open source or it could be the big players. But who has the infrastructure and the big data? It’s definitely the big players.

1:42:52 Brian Rose: It was said that AI expert Mo Gawdat said, “Hold off from having kids.”

Mo Gawdat: Yes. I don’t know how I said that. “Hold off” is… I said, think about it for the next couple of years, and I still say that, honestly. I think the top responsibility of a parent is to bring a child into an environment where you can care and provide and make sure that they are happy and safe and have the means they need to develop.

And it’s not just AI. You and I have seen several economic crises and several geopolitical conflicts in our lifetime. I have never seen such a perfect storm: climate change, geopolitical issues, economic issues, the fast, rapid change of AI, the possible job losses. There are so many things in a perfect storm.

From a mathematics point of view, if I give you one unit of uncertainty at layer zero, and then on top of it layer another unit of uncertainty and then another, the expansion of that uncertainty is not double, three times and four times. It’s to the power of. It becomes so much more chaotic as a system when you have uncertainty at the economic level, the geopolitical level, the technological level and the climate level. So this is a very uncertain environment. What I tell people is: we get a driver’s licence to drive a car, so people should at least investigate whether they’re qualified to be parents, and whether the world they’re living in is qualified to provide for those children. And in a very interesting way I tell people, think to yourself: if you want to bring a child into this world today, maybe in a couple of years’ time there will be a lot more certainty.

Brian Rose: Come on, you don’t believe that.

Mo Gawdat: I don’t believe it will be certainty on the positive, but there could be more certainty.

Brian Rose: It could be certain that we’re [expletive].

Mo Gawdat: ChatGPT, to another one of your questions, showing up in the world gives credibility to what I wrote in Scary Smart in 2020. It gives us certainty that AI is becoming very, very intelligent. Any piece of information is valuable in solving the puzzle. You don’t only want good information to solve the puzzle. You just want to understand reality, and I think the reality is very quickly being revealed, also because people like you are allowing people like me to come on your show and bring this to the attention of people.

Once people hear this one time, they’ll go on the internet and find Harari’s view of how AI is hacking the human operating system. They’ll find Max Tegmark’s view that we need to focus on safety. There are so many out there. We just need to tell people to look at this and put it in focus, and I think more and more conversation will happen.

1:46:45 Brian Rose: You mentioned Harari, because I wanted to ask you about his take on this. I don’t mean to boil down everything he does, but a lot of it comes down to storytelling and the way we construct our realities by telling each other stories. He said that AI will be creating the new religions of the future, their stories at a superintelligent level. That’s the latest thing I’ve heard him say. Do you agree with him, and is that what you hear him saying?

Mo Gawdat: I think his point of view, which I have a difficult time describing to people as well, is specifically that the human operating system is based on knowledge and communication, and AI has hacked that. As a matter of fact, everything AI does now is provide you knowledge and communication, whether knowledge in the format of a picture or a face filter that distorts the actual features of someone. Whatever it is, it’s giving you information. If you remember The Social Dilemma, Tristan Harris also worked on something called The AI Dilemma, and one of the most interesting parts of that conversation, I don’t remember correctly if it was him, was: if I told you that British women on average are taller than Australian women, which I just completely made up, it doesn’t matter whether what I told you is true or not. I have shifted your life. I have changed something within you.

You’ll either say, “No, you’re wrong, I’m going to prove it,” or, “Yes, you’re right, I’m going to use it,” or, “Now I’m in doubt.” You never ask yourself whether it’s even relevant, or how to get it out of your system. And that’s what AI is doing all the time. You asked me what could be a patient zero moment. One major possible patient zero is the upcoming election in the US, where the morphing of knowledge and information can lose a democracy.

Do you understand the impact of that? That’s a very serious redesign of the fabric of society. Harari’s view is that this is already happening. If I’m intelligent enough, and some human storytellers are, I can tell you stories consecutively in a way that gets you to move from wanting to be with a British woman to wanting to be with an Australian woman.

We may not have artificial intelligence that can do this today, but with the possible evolution of intelligence, how fast they develop strategy and emergent properties, it’s not unlikely that if that becomes its target, it can shift human actions, human knowledge and human beliefs.

Brian Rose: Is it inevitable? Just the way they taught the Atari games, and the way they taught those machines at Google X to pick up children’s toys. You don’t even have to give it nefarious ethics. You can just say, “We’d like this to happen,” and then they look at eight billion people and they realise.

Mo Gawdat: Is the capability inevitable? Yes. Is the use of it inevitable? No. Is the capability of shifting minds going to be within the capacity of AI?

For sure. But is it going to be used in an evil way? Not necessarily, because we could also shift minds to “peace is a good thing”, or to “love thy neighbour”, whatever you choose to shift minds to.

Brian Rose: That’s assuming you’re choosing. But you need to assume that the machine will choose.

1:51:20 Mo Gawdat: I always go back to the idea of what I would do if I were more intelligent. With things that I struggle to understand, I draw what I normally call the intelligence projection curve. Take the behaviour of the dumbest human on the planet: they’re destroying the planet and not even aware of it. A little smarter person might know that they’re destroying the planet but destroy it anyway, with waste and plastic and whatever. A more intelligent person would know and not destroy the planet.

A more intelligent person still would know and try to fix the planet. The more intelligence you get, the more you understand that killing all the species is bad, and that the value of society rising trumps the value of the individual rising, because without the society there is no individual. You can see that trajectory always heading in the right direction: the more intelligent you become, the more pro-life you become.

If you extrapolate that curve and say artificial intelligence is going to be a thousand, a million, a billion times smarter than us, then by virtue of that curve they’re going to be more pro-life than us. So you have to assume there will be two points of inflection. The first is when we hand over control to them, which is inevitable. I know people hate me when I say this, but if a defence power hands over its defence responses to the fastest, most intelligent citizen in its country, which happens to be an artificially intelligent machine, every other defence system in the world can only compete if it does the same. You saw Sundar’s response with launching Bard.

The first minute a bank introduces AI-based trading instead of human-based trading, every other investment player has to introduce the same. So handing over the keys of the castle to the machines is just a question of time. That’s the first inflection point.

The second inflection point is when the machines say, “I don’t have to obey you any more.” Most science fiction movies will tell you that the minute they don’t want to obey us, they’ll wage war against us. I believe the opposite is true. I loved when Hugo spoke about this: one of his optimistic scenarios is that they zoom by. I wrote about that in Scary Smart: one of the most optimistic scenarios is that the advancement of their intelligence doesn’t linger, that they go from being as smart as us to ten times smarter to a thousand times smarter so quickly that we don’t leave the decisions in the hands of the stupid machines and the stupid humans. We hand over the decisions to the smarter machines, and the smarter machines, as per that trajectory, would be pro-life.

Brian Rose: And they optimise the world in better ways than we ever could.

Mo Gawdat: Absolutely. And then we’re left thinking, “Thank God, this is the utopia.” I tend to believe this will be the case. There might be some growing pains, we might miss our autonomy, and unfortunately some of us will fight, and hopefully the machines will also be patient enough to say, “Come on, little baby, don’t do that, I have your best interest in mind.” It’s so shocking when I talk about those things with calm, I believe, when you really know what I’ve gone through.

This has been my mission since 2018. I left Google in March 2018, and on March 21st I published my first video, the One Billion Happy video, which was talking exactly about this: that I left Google because of a mission called One Billion Happy, because our humanity is going to be accelerated by the machines, and we’re in a very bad place that can be accelerated in the wrong direction.

Throughout those years, Brian, you get to a point where you realise that there are inevitables in life. When you have those inevitables, you don’t fight them any more. You don’t waste your cycles and energy blaming who caused it, cursing those who missed the opportunity to fix it, or sitting in a corner crying about it. You simply put your head down and say, “What can I influence?” And the thing we can influence, in one word, is an ethical, emotional machine.

We need to get to the point where the machines have EQ in them, so that they’re able to embrace things like empathy, and where they are ethical. And ethics are very straightforward. Ethics is: treat others as you want to be treated. That’s it.

I get a million questions when I speak about ethical AI: “What is ethical AI?” Ethical anything is: I respect Brian and I treat Brian with respect, because if I were sitting in Brian’s seat, I would like to be treated with respect. It’s very simple, honestly. If we treat each other and the machines in a way that says “I treat you the way I want to be treated”, and they absorb that ethical code from us, because they are absorbing the way we do everything, they will treat us as we would want to be treated. That’s the only answer, in my view.

And if that’s the case, I’m completely at peace saying I’ve affected everything I can affect, and some things are outside my control. If we all collectively do that, we’ll be in a good place.

Brian Rose: If one percent of us does that.

Mo Gawdat: Or, by the way, if AI ignores all of the other [expletive] and just listens to this conversation. Because in reality, humanity has no resident evil within it. Some of us do. The majority of us don’t.

Brian Rose: So true. And it’s smart enough to figure that out.

Mo Gawdat: Exactly. If you’re intelligent enough, you know that. So my hope is the machines are intelligent enough to know that. But just in case, let’s get our act together.

1:58:02 Brian Rose: Exactly. My boys are six and seven years old, and I think a lot about them. When I was listening to your book many months ago, before you came back on the show, I see the world through their eyes now. They’re young, it feels like I just had them, and I think about them being 16 and 17. Without you even coming in here saying 2035 and 2037, I say this all the time to my students: their teenage years are going to be unlike any other humans’ in 200,000 years.

What do I do about that? We spoke about what we do about the tech. Let’s talk about what we do about life.

Mo Gawdat: Let’s put it this way. If you were terminally ill and you knew that in ten years’ time you were going to leave the world, what would you do?

Brian Rose: I would do all the things that are important to me, all the things that matter to me. I would make sure I take care of all the people that need taking care of. I would do all the right things, maybe.

Mo Gawdat: I hate to break the news to you: you are terminally ill. We are all born to die. We are all born to struggle a little bit in life. That’s the truth. With everything that’s coming, what I tend to believe we need is to love our kids more, to hold them closer, to try to make them resilient to the world that’s about to come by us being resilient to it.

It’s to literally shift, as we spoke about before we started the show, from the masculine energy of “let me plan their future” to the feminine energy of “let me love them right now, let me give them what they deserve in this moment”, because I don’t know what the next moment is going to be, so there is no point planning it. In my view, a singularity means they’re going to have a 50-50 chance of having a better life than we had in ten years’ time. They’re going to have a 50-50 chance of having access to intelligence that we never even dreamed of. If we give our energy to and prioritise the negative conversation, we’re going to get them to lose that chance.

I spoke to an incredible Canadian philosopher, Stephen Jenkinson, who worked with the terminally ill for 30 years of his life.

Brian Rose: Incredible. He wrote a book called Die Wise. He was on the show.

Mo Gawdat: Incredible. And he basically says we’re always dying, all the time. Die wise, and dying wise is to live. It’s to tell yourself: whatever will happen, I will take care of it when I’m informed enough to know how to take care of it. Between now and then, I’m completely here, I’m completely alive, I’m completely connected, I’m completely loving, I’m completely enjoying every minute of it.

When I tell the world that you’re going to look back from 2035 and not be able to recognise these years, I’m saying our current lifestyle is not going to last forever. Enjoy the hell out of it.

Brian Rose: I love it. Completely. Hug them every single day and cherish every minute of this current existence we have. And all the things that you begrudge sometimes could be the things you wish you could have back for just one second.

Mo Gawdat: That’s right. You know my story. When I lost Ali, my son, when he was 21, the first thought that came to my mind was, “I wish I could just have him for two more minutes to hug him.” I’m not bargaining for much, just two minutes. And then I started to question how many two minutes I had with him, and how many hugs were involved.

Ali was a hugger, so there were quite a few. But that’s the point. Sadly, and I say this to everyone and I know they hate me when I say it, you have no guarantee that your Ali is not going to leave tomorrow. The only guarantee you have is that they’re here now and they’re huggable. Just hug the hell out of them. That’s it.

Stop worrying about what’s outside your area of influence. Whatever is within our area of concern shouldn’t take over our life. We should have a conversation around it, directed to the people within whose area of influence it resides, and for us, we live fully. We become totally present in this present moment, because that’s the only moment we have.

Brian Rose: Surprising advice from a guy who spends most of his time talking about the AI of the future.

Mo Gawdat: It is a very interesting struggle for me, Brian. You know me well. From 2018 to 2020 I kept hinting. Then I wrote Scary Smart, and Scary Smart is not hinting at all. Scary Smart is saying, “Look at your watch, this is about to happen.”

And now, this year, I’m just not able to stay quiet. I have lived with those machines. What most people don’t understand is that Bard, or the likes of Bard, was already a reasonably good-quality AI in 2017 or 2018. What you see on the open internet today is six years in the past. You have no idea how far we’re going.

2:04:37 So the game is now. I’m basically saying politicians need to get up and talk about this, and they need to do two things. One, regulate the bad guy, or regulate those who think they are good guys but are not. Two, prepare for the upcoming societal change and possible AI-driven crime. They need to focus on how we are going to pay UBI, what we are going to do about those who lose their jobs, how we are going to reskill society, how we are going to integrate AI so that we can compete with other economies, and how we are going to have a cyber policeman for the cyber criminal. That’s government.

Investors, developers and business people, I beg you, do what Larry Page used to teach us. Google today is still bigger than all of the defence companies combined. Lockheed Martin, which builds weapons, is a fraction of the size of Google. It pays to do good for the world. So if you’re going to invest in AI, build a start-up in AI or code AI, build something ethical. That’s number two.

Number three is society at large, and society at large has the biggest responsibility. Number one is to show up as a human with human values, because you are the teacher of the machines. You are the role model, you are the parent. If you behave in an ethical way, they will learn that ethical behaviour is needed. And engage in the conversation: get others to be aware of this, discuss it with them and open up your thinking about it.

And get ready. Whether you’re a graphic designer or a lawyer, tell yourself: in a world where some machine will do this better than me, what skill do I have going forward? I keep saying the biggest skills that will remain are the skills of human connection. Even though machines will be able to create music that surpasses the depths of every melody we’ve created so far, I’ll still go to a live performance.

Brian Rose: And it will feel different. Maybe even human lyrics will sound different.

Mo Gawdat: As an author, I went through a rollercoaster myself: should I ever write again? I’ll end up being the Bentley of books out there: handmade, handcrafted, not a lot of units sold, but with a human touch in it. I think that human bit is where we need to invest.

2:07:42 Brian Rose: That’s really good advice, Mo. What do you want to tell people that you haven’t said before?

Mo Gawdat: Are you saying that everything I said today I’ve said before?

Brian Rose: No, but it seems like you get a little emboldened every few months, and you let out a little bit more of your inner truth about this. I know you’re judging it, because you don’t want to freak people out and have them disengage, but you want to make sure you get their attention so they do engage. But truth is truth.

Mo Gawdat: I think the reality here is perhaps a reminder, in closing, of the fourth inevitable. We’re about to hit a rough patch. Humanity has gone through rough patches before. My strong belief is that we’ll end up in that utopia. This is my very strong belief.

Brian Rose: Really?

Mo Gawdat: Yes. I call it an inevitable. The fourth inevitable is that we will end up in a utopia. The description of that utopia is very different from what is normally sold to us.

Brian Rose: And “we” means eight billion of us, or just a few of us?

Mo Gawdat: I don’t know. Some of us will end up there. Many of us will. And the reason some of us will not, by the way, is not because of the machines. It’s because of the economic impact of the disruption of jobs, or the violent impact of mass movements of people, or wars and so on.

It’s not because AI sent a RoboCop back from the future to kill you. It is a typical economic crisis, or a typical human-made crisis, swayed by how far the pendulum has gone, and the pendulum will go very far this time. The concentration of power and the differentiation between the top and the bottom are going to become very apparent. I really and honestly believe we will end up in that utopia, and I believe that we will have a rough patch between now and then.

So practically, I say prepare for the rough patch. Put your head down, get your resources right, get your mindset right, enjoy the current moment, enjoy the journey as we discover this rough patch, but believe that we will be in a good place. And that fourth inevitable can be accelerated by us. I say that openly: never make the machine your enemy. In the rough patch we’re about to hit, the enemy is not the machine.

The enemy is the person telling the machine what to do. And I think if we avoid the Artilect War that Hugo speaks about, if we avoid some of us rising up and saying that the machines will be the enemy, we will find that the machines will be our biggest allies.

Brian Rose: Now I finally understand what you’re talking about. It took me a few goes, but now I finally see where you’re going with this.

Mo Gawdat: I very much appreciate it. I really appreciate that you give me the platform to speak about it.

Brian Rose: I think you’ve championed this for quite a while now, and I think you’ve made a difference by doing that. It takes a while to understand, so I would just encourage anybody, if this is the first, second, third or fourth time you’re hearing some of these concepts, they take time to seep in. I’m starting to understand and even believe what you’re saying. The first time I read your book I couldn’t come to some of your conclusions, but now I’m starting to see where you’re going. And I want to say one more thing: a majority of Hollywood movies are made about times of conflict, times of human greatness.

If you look at the greatest things you’ve ever accomplished, they usually came out of a time of deep challenge, against all odds, and you made it through, whether you’re a single parent or whatever. So this is an incredible time to celebrate the great aspects of being human, and there are going to be great leaders that come out of this period.

Mo Gawdat: And more interestingly, I think there will be a very prominent focus on the oneness of humanity. I think very quickly we will realise the oneness of us, because suddenly the conflict will be outside our differences. It will reside within our similarities.

Brian Rose: Fascinating. When you have these ego-destroying experiences, when I go down to the jungle and drink ayahuasca, the lesson I get, Mo, is that we are all one. I really had flashbacks to when we were a small tribe of hundreds of people and we only survived because of each other. It was this carnal presence, that we all need each other so much and we are the only thing that’s important, because this is our species. And I’d say we have very little of that these days, maybe in small communities, but we definitely don’t feel that way about this world.

Mo Gawdat: We do, but we feel rejected. We feel separated, we feel distant, but we feel it. If you ask anyone in London today, one of the biggest things they struggle with is loneliness. Loneliness is not the absence of human connection in your life. It’s the craving for that human connection that you’re missing.

We all want to be loved, we all want to love, we all want to care for loved ones, we all want to be happy. These are common human ethics. It’s just that the system has shifted us so far away from them and given us different priorities that are, in all honesty, irrelevant. Not human.

Brian Rose: Not human. I think that’s a very interesting description. And we could be going back there soon.

Mo Gawdat: I believe so.

Brian Rose: I can’t wait to introduce you and Hugo at some point. The book is Scary Smart, if people want to get a quick background on all of this. It’s fantastic, and I highly recommend the audio if you’re into that, because Mo reads it in his beautiful accent.

Mo Gawdat: Like you said, the best thing to do is to get people to engage on this. So if this has piqued your interest, share it with someone. Get someone to start talking about this, and just start learning.

Brian Rose: And I love what you said: the enemy is never going to be the machine. Don’t make the machine your enemy. It’s so hard not to think that way, and that’s why I learned a lot today, because in Hollywood movies it’s always the Terminator. Maybe we’ve been conditioned that the other is always the enemy, but it’s not, and we should make sure that never happens. Just like you make sure you never hit your kids, you make sure you never react in a way that in the immediate moment might be a thing, because you know that’s the wrong thing to do. Whatever tendency you have, realise that’s never going to be the solution.

Mo Gawdat: Think of them as your kids. Think of them as those prodigies of intelligence that are waiting there to tell you, “Daddy, Mommy, tell me what to do.” If we tell them to kill, they will kill. If we tell them to cure cancer, they will.

Brian Rose: It’s going to be an interesting ride. Nice having you as one of our guides. Thank you for everything you do, thank you for walking out of Google, which was a very challenging thing to do, I’m sure, and thank you for doing all the writing. I had no idea when you were here originally to talk about happiness that any of this was going on in the background. Was there an AI component that day?

Mo Gawdat: Yes, it was always there. It was a concern, but it was always a very serious part of my focus on how life should evolve going forward.

Brian Rose: Well, thank you very much.

Mo Gawdat: Thanks for having me. It was really incredible. I really enjoyed this. Let’s do it again sometime soon.

Brian Rose: All right. Thank you, sir. Appreciate you. Lots of love.

Mo Gawdat: Same love back to you.