We adapt, we always have, and we'll adapt to AI too

Humanity is very good at adapting. We adapted to the periods of war and to the periods of peace, and some of us adapted to Covid restrictions so well that we wanted them to continue (in July 2021 about a quarter of British adults told Ipsos they wanted nightclubs closed for good, and almost one in five wanted a permanent 10pm curfew) ^[1]. We adapt, we always have and we always will, and with AI, as it was with computers and with digitising every record we had, we'll adapt and eventually thrive.

I'm writing this because I read AI 2040, the plan the AI Futures Project (the group behind AI 2027) published as their Plan A ^[2], and I don't think it will work. It's ninety pages with the appendices and a serious attempt at writing down how the next fifteen years could go well, so it deserves a proper answer.

What AI 2040 says

Their starting point is that the AI companies are racing to build AIs smarter than humans in every way, and that the race ends either in extinction or with a tiny group of people in control of everything ^[2]. Plan A is the US and China making a deal in 2029, pausing new training runs while they set up inspections, making almost all AI research public, and putting each side's new datacenters in another country so either can destroy them if the deal breaks down (they call it mutually assured compute destruction). AI is held at roughly the level of top human experts from 2035 and only allowed past it in 2040 ^[2].

The reason for all the slowing down is speed. In their scenario only 26% of Americans still have a job in 2036, the world economy grows around 200 times in a decade, and in an appendix they compare the 2030s to living through five hundred years of English history in five ^[2]. A footnote also gives their team's estimates of AI wiping out humanity as between 10% and 30% ^[2]. I'm not going to argue that number here. Where I disagree is on the pace, and on the idea that the way through is two governments setting it for everyone for fifteen years.

We've had change this fast before

I'll concede it first. Change is too quick right now, and it could get a lot quicker. But we've lived through stretches of change just as fast, and the fastest ones came with a war attached.

When the First World War started in August 1914, less than eleven years after the Wright brothers first flew, the Royal Flying Corps had around 100 aircraft. By the end of the war in November 1918 the RAF had more than 22,000 ^[3]. In March 1942 American drug companies had made enough penicillin to treat a single patient ^[4], and two years later 2.3 million doses were ready for the D-Day landings ^[5]. The Second World War also gave us radar, jet engines, the first electronic computers and the atom bomb, all in six years.

The cost was horrific. Between 15 and 22 million people died in the First World War ^[6], and between 60 and 75 million in the Second, about 3% of everyone alive in 1940 ^[7]. We came through both and adapted to the world they left behind, at a price nobody should ever pay again. So what if we can get that pace of change without the loss of life?

Shipping is the peacetime version, and it happened inside one working lifetime. Before containers, every sack, barrel and crate was carried on and off ships by hand, and New York alone once had more than 60,000 longshoremen ^[8]. In 1956 Malcom McLean sent the Ideal X from New Jersey to Houston with 58 metal boxes on board, and loading cargo went from about $5.86 a ton by hand to 16 cents a ton in a container ^[9]. Today a ship carrying what would once have needed thousands of people to load, unload and move runs with a crew of around twenty or fewer, and Maersk's Triple E ships carry more than 18,000 containers with a crew of 13 ^[10].

It wasn't easy. Most of those dock jobs went, the work left the old waterfronts for bigger ports ^[8], and the families and neighbourhoods that lived off the docks had a hard time of it. But most people adapted to it.

Most of the world isn't there yet

It feels incredibly fast for coding, but I'm not sure about the other parts of language. All of social media is full of AI slop text, and communication isn't as structured as coding. And most people haven't used it at all yet. Microsoft estimates that in the first three months of 2026, 17.8% of working-age adults in the world used a generative AI tool at all, even once ^[11]. In the UK it was 42%, in the US 31%, in Romania, where I grew up, 18%, and in Nigeria and Ethiopia around 10% and 8%.

And 2.2 billion people were still offline in 2025, with 94% of people in high-income countries online against 23% in low-income ones ^[12]. For those of us at the edge of this, building agents and reading ninety-page scenarios about superintelligence, it's easy to blow it out of proportion. Technology takes time to reach critical mass, and that time is when people and institutions do the adapting.

It doesn't have to be a tiny group

I don't think a tiny group of people controlling everything is where we have to end up. The internet didn't start out as a tiny group of people. Tim Berners-Lee invented the web at CERN in 1989, and on 30 April 1993 CERN put the web software in the public domain so anyone could use it and build on it ^[13]. The innovators and dreamers dream for society, not for themselves. The rent seekers come later, and slowly erode it.

Plan A wants a lot of the same openness, with almost all AI research made public and dozens of companies around the world allowed to catch up ^[2], but it gets there through a deal between two governments. The people who built the web built it because it was fun and exciting, not because they were thinking of the billions they'd make, and that's why most of it is open protocols and open source. The internet wasn't going in a great direction before AI, with a lot of it ending up behind a few platforms that decide what you see, and with AI we can maybe turn it around, if enough of the people building it build it to be open.

Where I land

It won't be easy, and the docks and the wars weren't either. We've adapted to every big change so far, including ones that came as fast as this one, and I don't think AI is the one where we stop.

Adapting and learning, alongside imagination, is what makes us human. AI 2040 is long, but the main scenario and the five centuries appendix are worth an evening, and then you can decide for yourself whether you think people will cope ^[2].

References

Read on 20 and 21 October 2026.

1. [1]Ipsos MORI for The Economist, fieldwork 2 and 3 July 2021, 1,025 adults in Great Britain aged 16 to 75, as reported by DJ Mag on 14 July 2021. Source for 26% wanting nightclubs to stay closed permanently and 19% wanting a permanent 10pm curfew. djmag.com
2. [2]Thomas Larsen, Romeo Dean, Brendan Halstead, Eli Lifland, Ryan Greenblatt and Daniel Kokotajlo, AI 2040: Plan A, AI Futures Project, 2026, the full edition as a PDF. Source for the scenario's summary and timeline, the deal and mutually assured compute destruction, making AI research public and letting other companies catch up, the team's 10% to 30% estimates of AI-driven human extinction (footnote 4), 26% of Americans in work in 2036, world GDP growing about 200 times in the 2030s, and Appendix N, Five centuries in five years. ai-2040.com
3. [3]Historic England, 100 years of the Royal Air Force, Historic England Blog, 28 March 2018. Source for the Royal Flying Corps having around 100 aircraft in August 1914, and the RAF having more than 22,000 aircraft in November 1918. heritagecalling.com
4. [4]USDA Agricultural Research Service, The Rescue of Penicillin, ARS timeline. Source for drug companies having produced only enough penicillin to treat a single case by March 1942. ars.usda.gov
5. [5]Richard Conniff, How World War II put penicillin into every pharmacy, National Geographic, August 2020 issue, published 14 July 2020. Source for 2.3 million doses of penicillin made in preparation for the D-Day landings. nationalgeographic.com
6. [6]Wikipedia, World War I casualties. Source for an estimated 15 to 22 million military and civilian deaths. en.wikipedia.org
7. [7]Wikipedia, World War II casualties. Source for an estimated 60 to 75 million deaths, about 3% of the world's estimated 1940 population of 2.3 billion. en.wikipedia.org
8. [8]Jordan Fraade, Next City, 26 May 2015, There's Money Hiding in New York City's Waterfronts. Source for New York once having more than 60,000 longshoremen, and the jobs moving to New Jersey and Staten Island with larger ships and containerisation. nextcity.org
9. [9]Human Progress, Heroes of Progress, Part 17, Malcom McLean. Source for the Ideal X sailing from New Jersey to Houston in 1956 with 58 containers, and loading costs falling from $5.86 a ton by hand to 16 cents a ton. humanprogress.org
10. [10]Wikipedia, Triple E-class container ship. Source for the first Triple E ship entering service on 2 July 2013, a capacity of 18,270 TEU for the first generation, and a crew of 13. en.wikipedia.org
11. [11]Our World in Data, Estimated share of working-age adults who use generative AI, from Microsoft's AI Diffusion Report (modelled estimates from usage of 19 AI tools, ages 15 to 64), retrieved 21 October 2026. Source for 17.8% worldwide in January to March 2026 and the country figures (United Kingdom 42.2%, United States 31.3%, Germany 31.1%, India 17.6%, Romania 17.5%, Nigeria 10.1%, Ethiopia 7.6%). ourworldindata.org
12. [12]ITU, Facts and Figures 2025, press release of 17 November 2025. Source for 2.2 billion people offline in 2025, and 94% internet use in high-income countries against 23% in low-income countries. itu.int
13. [13]CERN, The birth of the Web. Source for Tim Berners-Lee inventing the World Wide Web at CERN in 1989, and CERN putting the World Wide Web software in the public domain on 30 April 1993. home.cern