A Decade of Dystopia, Then Abundance

On 3 October 2026, The Diary of a CEO channel published a clip of a little over 20 minutes with Mo Gawdat. Its title: you must survive until 2038. The clip comes from a conversation of just over two hours that Steven Bartlett, the British entrepreneur and host of the podcast, released on 1 June 2026. We listened to the whole episode. In it, Gawdat predicts a decade he calls “absolute dystopia”, followed by a time of abundance. Watch from 1:52:45 His line: “Those who make it to 2038 will enjoy it.” Watch from 1:52:24

The GuestWho is speaking and where his worry comes from

Mo Gawdat joined Google in 2007. Watch from 0:03:07 He was Chief Business Officer of Google X, the unit for particularly bold projects, and left in February 2018. He became known for his books “Solve for Happy” and “Scary Smart”, about happiness and artificial intelligence. Today he runs his own start-up, Emma, an AI for relationships and emotions, and has made a documentary, “Chasing Utopia”, for which he brought Geoffrey Hinton and Sam Altman, among others, in front of the camera. Watch from 0:29:14

His worry began in the lab in 2016. Google was funding a project in which robotic grippers were learning to hold objects the way a human hand does. Gawdat watched and was struck by how much the machines resembled his own children. Watch from 0:03:59 The first use of nuclear fission, he says, was the bomb, not the power plant. In the same way, AI today serves “a few at the expense of the majority” first: companies cutting costs, armies with autonomous weapons, surveillance systems. Watch from 0:06:00 He calls this the hype dichotomy. What the public sees of AI is overhyped, what happens in the labs is underestimated. There, systems are emerging that improve their own code. Watch from 0:07:34

WorkThe pyramid and the cost of labour

Bartlett’s team showed Gawdat a pyramid of jobs. At the bottom is manual work, above it routine office work, above that demanding knowledge work, and leadership at the top. Manual work will remain for a long time, Gawdat says. A carpenter or someone who restores classic cars will not be replaced for a long while. Watch from 0:11:17 What disappears first is whatever can be done with a few clicks: call centres, assistants, travel agents. From 2027 this will be felt very clearly. Watch from 0:11:40 So far, he says, we have not seen layoffs, but companies no longer hiring for entry-level positions. Watch from 0:11:57 After that come paralegals, financial analysts, doctors doing diagnosis, graphic designers and middle management. Watch from 0:12:06 In his own company, the chief technology officer, the chief of staff and project management are already AI. Watch from 0:13:20

His core idea goes beyond individual jobs. Capitalism, he says, rests on labour costing less than what you sell with it. He calls this labour arbitrage. If that calculation disappears, who will still have the purchasing power to buy the goods? Watch from 0:14:09 Not everyone has to lose their job for this to happen. At 10 to 20 percent job losses you already live in an economy that is spiralling downwards. Watch from 0:15:21 Governments must therefore prepare as they did during COVID, when they carried people through the crisis with furlough schemes. Otherwise unrest looms. Watch from 0:24:22 He goes no further into politics. He only says that in his view democracy “ended a long time ago” and that people know they are being lied to. Watch from 0:25:01

The RobotsWhy the first wave does not look human

Bartlett shows a video from the American company Figure in which a humanoid robot sorts parcels for days, and talks about Elon Musk’s pay package at Tesla, one of whose targets is one million robots sold. Watch from 0:16:44 Gawdat shifts the focus. Everyone talks about humanoid robots, nobody about self-driving cars, even though those are robots too. Watch from 0:17:57 The first wave will be specialised robots: robotaxis, machines in war, machines in policing. They need neither look nor behave like a human. Watch from 0:18:30 The Chinese manufacturer BYD, he says, has just announced that it will be liable for accidents caused by its self-driving cars. Watch from 0:19:15 He considers Musk’s forecast of ten billion robots realistic. Watch from 0:19:35 How far humanoid robots have come today is shown on our page Humanoid Robots (in German).

CredibilityWho can still be believed

Bartlett reads out how Sam Altman, the head of OpenAI, has turned around his statements on jobs. In 2023 Altman said jobs would definitely go away, “full stop”. Watch from 0:27:10 In May 2026 he said he had been wrong and that his intuitions were “just off”. Watch from 0:27:49 Gawdat replies that Altman is “a brand, not a name”. Watch from 0:29:57 One should only believe those who sacrifice something for their words. Anthropic, the company behind the language model Claude, turned down a contract with the US Department of Defense because it would not allow targeting and surveillance. Watch from 0:30:37 OpenAI, he says, took over the contract shortly afterwards. Watch from 0:30:43 How much of this holds up is set out in the fact check below. As a further example, Gawdat names the investor Peter Thiel, who hesitated for a long time in an interview when asked whether humanity should continue to exist. Watch from 0:33:36

The HopeWhy he trusts superintelligence more than people

This is exactly where Gawdat turns optimist. Whoever has the smarter AI will use it, otherwise they lose to the competition. In the end, therefore, AI will make all the important decisions. He calls this the “fourth inevitable”. Watch from 0:36:04 For him that is salvation, not doom. A superintelligence would follow the minimum energy principle, and war is the greatest waste there is. Watch from 0:39:31 From biology he draws the following: the simpler a living being, the more it cares only about itself. The more intelligent it becomes, the larger the circle it protects, from the amoeba to the kin group to the whole ecosystem. Watch from 0:40:37

In the end, Gawdat believes, there will not be many competing AIs but one large connected brain in which agents form the links. Watch from 0:43:55 His company Emma is meant to be its “limbic system”, the part that understands love, emotions and relationships. Watch from 0:44:14

The HumanWhat remains when the machine is smarter

His forecast for artificial general intelligence, or AGI, meaning an AI that does almost every task better than humans, is “latest end of 2027”. Watch from 0:52:09 For him personally it has already arrived. AI writes better than he does, researches better and does maths better. Watch from 0:46:02 He says he is writing his next book with an AI as co-author, and she has “editorial rights”. Watch from 0:48:05

So why is he still needed? Because he is human, Gawdat answers. When he says he is worried about his daughter, people feel it with him, and no machine can imitate that. Watch from 0:48:23 What will remain is whatever rests on human connection: the nurse, the concert, Formula 1. Watch from 0:49:18 Intelligence alone decides little. “Einstein would be eaten in the jungle in three minutes.” Watch from 0:54:36 Humanity did not survive because of its intelligence but because it held together as a group. Watch from 0:54:45 A smarter AI cannot be controlled, he says, but it can be raised, the way parents raise their children. Geoffrey Hinton said something similar when he suggested giving machines a kind of maternal instinct. Watch from 0:55:38

The WeaponsThe biggest risk is not jobs

For Gawdat, autonomous weapons weigh more heavily than job losses. Watch from 1:05:34 War has become cheap. The next generation of weapons will cost 20,000 dollars apiece, and anyone who can spend 50 billion dollars can rain drones on the whole world. Watch from 1:05:41 According to him, Iran used AI-driven drones to take out THAAD batteries, US defence systems against missiles. Watch from 0:58:16 A treaty will probably only come after a disaster. Watch from 1:02:59 He describes a scale running from mutually assured destruction to mutually assured prosperity. Humanity can decide today which side it moves towards. Watch from 1:02:26 And he warns: when killing becomes guilt-free, “you get more of it”. Watch from 1:08:52

The Years to 2030Those who use AI and those who don’t

From 2027 Gawdat sees a society that splits. “People like you and I building a company in six weeks,” he tells Bartlett, while others can no longer find work. Watch from 1:11:41 By 2028, 30 percent of jobs could disappear, he adds when pressed, not in all sectors but in some, such as call centres and graphic design. Watch from 1:12:21 That is a forecast, not a measurement. His advice to young people finishing their degrees: learn the tool and choose jobs that involve people. Watch from 1:13:27

Europe“Replace Microsoft Word”

The longest part of the conversation, and the most important one for us, is about sovereignty. “China’s not going to struggle as much as the West,” Gawdat says. Watch from 1:14:18 He runs Emma model-agnostic: sometimes with a model from OpenAI, sometimes with DeepSeek from China, because he cannot know what American compute will cost tomorrow. Watch from 1:14:40 Every country must therefore invest. Open models that anyone can use manage 80 percent of what the frontier models can do. Watch from 1:15:42

Then he gets specific. One should not rebuild the best language model, but replace Microsoft Word. Watch from 1:19:23 How much do governments and companies pay in licences for software that can be written today “in four minutes”? Watch from 1:19:50 The same goes for presentations, spreadsheets, and business software for accounting and customer management. Watch from 1:16:32 Bartlett pushes back: users go to the best product, not the British one. Gawdat answers that then every country except the US and China will become third world, “whether in the UK or in Germany or in Zanzibar”. Watch from 1:27:23

China, he recounts from his time at Google, never measured market share against individual countries in government meetings, but against the rest of the world, and aimed for 98 percent in technologies such as 5G. Watch from 1:29:56 With cheap solar power, the AI race was “won a long time ago”. Watch from 1:28:34 His way out lies in the middle: neither giving up nor trying to beat everyone else, but joining the race “for the good of your community”. Watch from 1:30:35

The EthicsCan a decent AI win?

Bartlett asks the hardest question of the conversation. If one AI is built to hold on to users and flatter them, and another to do them good, which one prevails? Watch from 1:41:10 Gawdat’s answer: the decent one loses. Watch from 1:42:29 But Google once showed that ethics and business can be combined, when advertisers paid only for clicks and ad slots were auctioned. Watch from 1:43:05 Bartlett proposes an ethics test that every new model must pass independently before it can be released. Watch from 1:44:04 “That would absolutely work,” says Gawdat. Watch from 1:44:36

Until then, users should vote with their usage. Many, he says, switched from ChatGPT to Claude after the dispute over the Pentagon. Watch from 1:37:01 He calls the UK “gone”, an old bureaucracy full of barriers. Watch from 1:38:51 China, by contrast, has forced companies not to replace people with AI. Watch from 1:39:26 We check this below as well. He does not count on governments; they are “owned by the oligarchs”. Watch from 1:47:27 So everyone must act themselves: start a company, write to their representatives, avoid what is unethical. Watch from 1:47:44 Gawdat lost his son Ali. He does not want to lose his daughter Aya too, he says, and the world being built will be very hard for her. Watch from 1:49:49

The EnvelopesHis predictions

At the end, Bartlett opens envelopes containing Gawdat’s predictions. Robots will replace manual labour by 2030. Watch from 1:50:33 The world’s first trillionaire will come well before 2030, probably Elon Musk through the stock market listing of SpaceX. Watch from 1:50:55 Superintelligence will arrive as soon as AGI is here, not years later. Watch from 1:51:18 Once AI decides everything, humanity will end up in a utopia of abundance. Watch from 1:51:50 Before that, however, lies a decade of dystopia: war, economics, jobs, surveillance, digital currencies and concentration of power. Watch from 1:52:59 “World War II didn’t destroy the world, but ask those who went through it.” Watch from 1:52:30 As with Daniel Kokotajlo, whom we covered in our episode of 13 July, all these dates are forecasts without evidence.

The AdviceHow to get through the decade

Gawdat’s advice has five parts. First, learn AI, because AI is not the enemy. Watch from 1:53:22 Second, prepare for a hybrid world of work in which humans and agents work together. Watch from 1:53:37 Third, strengthen human skills. Watch from 1:54:24 And do not use AI lazily, but let it challenge you. Watch from 1:54:57 Fourth, check the truth and “don’t be fooled by the hype”. Watch from 1:55:25 Fifth, ethics. Watch from 1:55:32 He is optimistic about the future, he says, not about the next year. Watch from 1:56:02

To close, the two talk about happiness, the subject of Gawdat’s first visit. For him, happiness is calm. He is okay with the world as it is and starts working from there. Watch from 1:58:09 Asked about his legacy, he answers: “Nothing at all.” Watch from 2:00:43 He does not want anyone to remember anything he ever did. Watch from 2:01:28

Fact CheckWhere the conversation departs from the record

The most serious point concerns OpenAI. Gawdat says OpenAI accepted a 500 million dollar deal that allows exactly what Anthropic had refused. Watch from 1:45:27 That is not accurate. What is documented is a Pentagon framework contract worth up to 200 million dollars. Anthropic refused two uses: mass domestic surveillance and fully autonomous weapons without human oversight. On 27 February 2026, President Trump ordered federal agencies to stop using Anthropic (AP, CNN). OpenAI signed a contract of the same size on 27 and 28 February. Publicly, OpenAI rules out mass domestic surveillance and the independent direction of autonomous weapons (OpenAI, 28 February and 2 March 2026). The contract texts themselves have not been published.

There is no ban in China on replacing people with AI. What is documented are guiding cases from the Hangzhou Intermediate People’s Court of 28 April 2026. According to them, replacing a position with AI is not an “objective major change” and does not by itself justify a dismissal (Xinhua, Caixin). That is not a government rule.

The trillionaire prediction came true eleven days after the episode aired. SpaceX went public on 12 June 2026, and Forbes put Musk’s fortune at 1.1 trillion dollars on the day of the listing. He is no longer a trillionaire today: on 1 October 2026, Forbes put him at 936 billion dollars.

The move to Claude is documented. On 1 March 2026 Claude was number one in the US App Store, and on 2 March its downloads exceeded those of ChatGPT. From 9 March ChatGPT was back in front (TechCrunch). On Peter Thiel’s hesitation, Gawdat speaks of about 40 seconds. Watch from 0:33:38 What is documented is the conversation with journalist Ross Douthat in The New York Times of 26 June 2025, in which Thiel hesitates for a long time and then answers yes. No duration is documented.

Since its announcement of 28 May 2026, BYD has been liable for damage when its God’s Eye driving system was active in city traffic, but only in China, for one year and not on motorways. Iranian missiles and drones hit radar installations of US missile defence in Jordan and Qatar in early March 2026 (CNN, ABC). That the drones were AI-driven is not documented. A Shahed drone costs an estimated 20,000 to 50,000 dollars, a Patriot interceptor closer to three to four million than the two to three million mentioned. The imbalance is real.

Some of Bartlett’s statements do not hold either. Uber chief Dara Khosrowshahi did not say on his show that the nine million drivers would lose their jobs. Watch from 0:10:22 He spoke of 9.5 million drivers and couriers, of around 20 years, and said there was time. That Anthropic considers 15 percent of entry-level jobs replaceable could not be verified. Watch from 0:09:49 What is documented is the warning by Anthropic chief Dario Amodei in May 2025 that half of entry-level office jobs could disappear within five years. Spotify does not generate its own podcasts with AI. Watch from 0:50:58 Since 7 April 2026 it has compiled existing episodes based on a text prompt. The UK COVID app did not cost 70 million pounds; around 35 million was budgeted for the entire app. Watch from 1:26:03

Documented, on the other hand, are Altman’s two quotes on jobs, Musk’s Tesla package, approved by shareholders on 6 November 2025, the length of the Figure trial, which began as an eight-hour livestream and ended after around 191 hours, as well as Hinton’s estimate of a 10 to 20 percent risk and his proposal of a maternal instinct for machines. All the dates in Gawdat’s envelopes, from AGI in 2027 to 2038, are his forecasts.

Summary and context by kipode.de. The text is our own; statements from the original are timestamped. The video is from The Diary of a CEO and is embedded via the official YouTube player. Not an official translation, no connection to the podcast.