He must be a busy man, Rishi Sunak. When he’s not rescuing the country from inflation, sending the Royal Navy towards troublezones, making long term decisions for a brighter future, herding worried Conservative MPs towards the lemming-edge of the next election, and organising globally important AI summits, I doubt he has much time to read the darker recesses of TwitterX and Reddit where he might have come across a character named ‘Jimmy Apples’.

Which is a shame, in terms of that AI conference – being held this week at Bletchley Park, the “home of computing”. Because this summit seems like a sincere and valuable attempt to achieve some global governance of artificial intelligence, even as it explodes into every corner of our lives. In which case Sunak, the host, would be better equipped if he knew of the short, remarkable oeuvre of Mr Apples.

Apples first surfaced some months ago, as a random anonymous user on Twitter/X, while implying he was an insider at OpenAI – the most famous AI company of the moment. Then he put some flesh on the bones. On 4 March, he predicted that the next iteration of OpenAI’s famous Generative Pretrained Transformers, GPT-4, would be premiered on 14 March. That turned out exactly right – even as others got it wrong.

He has also revealed codenames supposedly used by OpenAI for their various AI models – names like Gobi and Arrakis (the latter derived from the novel Dune). These names have also proved correct: they are used inside OpenAI. At the same time, perhaps to refute online skeptics, Apples posted a photo of a receipt from a restaurant facing the OpenAI HQ, suggesting he really does work there. Apples could easily be a fake, but taken with the accurate inside info, the account is successfully suggestive.

So far, so what, you might say? Well, in recent months Jimmy Apples has made far more dramatic predictions. In April he said that the holy grail of AI research – Artificial General Intelligence (AGI) – the iteration when a computer can do virtually any task as well as a human being, and plenty oftasks much better, will be with us by 2025. Less than 18 months away.

To get a sense of the potential import here, the arrival of AGI means, in some views, the end of human society as we know it, as so much human labour will become literally worthless: a computer will be able to do anything you can do, and do it for pennies, without cease or effort. And who knows what the emotional, psychological and spiritual impact might be, when humans are, for the first time, sharing a planet with an alien and superior intelligence. Will we try to kill it, ape it, love it, pray to it? A mix of all four?

But there was a further bombshell. On September 18 Apples dropped this incendiary tweet, which has since become famous (in the niche nerd-world of obsessive online AI discourse):

@apples_jimmy AGI has been achieved internally

The result was predictably excitable. Could it be true that OpenAI already has an AGI model, sitting in some lab within the company, some kind of GPT-5 or GPT-6 only known to astonished insiders, but too momentous or perilous to be released to the public?

To add mustard and pickles to the pastrami, this revelation – that Big Tech companies already have versions of AGI – is reinforced by other rumours. At industry events Google tech bros have been known to drop hints that Google is sitting on something sensational. Could it be that Blake Lemoine, the Google engineer famously sacked in 2022 for claiming that one Google AI model was already ‘sentient’, ‘like a 7-year-old child that knows physics’, was simply but foolishly telling the truth?

The mystery deepened through September, when Jimmy Apples deleted his Twitter account, even as his ‘AGI: internally’ tweet turned into a meme. Perhaps, people wondered, he had said too much, and been sacked or censored. And yet the meme was only reinforced when OpenAI CEO Sam Altman himself (a fan of the novel Dune, as it happens) returned to his dormant Reddit account for the first time in eight years, to say something. What he said was, in these exact words, ‘AGI has been achieved internally’.

When Reddit went bonkers, he insisted this was all a joke.

Perhaps it was a joke. A drunken jape to troll the world. Perhaps, as others surmise, it was another clever way to boost the value of OpenAI – if so, it has certainly worked. In 2019 OpenAI was worth about thruppence, at the beginning of this year it seemed to be worth $27 billion, now it is thought to be worth nearly $90 billion. With a mere 400 employees or so, it is now, surely, the most valuable company per employee in the history of the world. So even if Jimmy Apples is a wind-up merchant, an awful lot of people are betting there is, nonetheless, some extremely valuable merchandise lurking behind the tomfoolery.

Moreover, even if Jimmy Apples is a complete fabrication, he speaks to a definite truth. Almost everyone now agrees that AGI (and, after that, ASI – Artificial Super Intelligence – the moment when AI becomes a recursive self improvinggodhead, far superior to all of humanity combined) is much closer than anyone anticipated a decade back. Right across the AI landscape many bigwigs at AI companies like DeepMind, Anthropic, Stability – as well as Google, Apple, Meta, and in academe – are predicting AGI within the next 2-10 years. Somewhere around 2027-8 seems popular. Four years away.

Perhaps this sudden urgency explains why President Biden dropped, with great fanfare, a sweeping Executive Order on Monday of this week, seeking to enforce US regulation and oversight of all aspects of AI, and to protect American citizens from its misuse. The order goes into impressive detail, but a neutral reader can’t help thinking it reads like a Legislation to Control the Use of This New-FangledElectricity in about 1880. ‘Companies generating over 35.7 gigawatts must get permission from Congressional department B9’, and so on. ‘Licenses shall be issued for anyone using more than 8 light bulbs’.

The greatest story in history is unfolding more or less quietly while everyone’s eyes are glazed over

Worthy as Biden’s effort might be – he even wants to ensure AI fits his drive for equity, diversity, and inclusion – the effect, I suggest, will be minimal. Even if he succeeds in slowing American AI development for a few years, the technology is far too tempting, revolutionary and powerful to be slowed, let alone stopped, and countries like China will eagerly overtake the USA (and they won’t be wokifying their Chatbots). The race to AI is like the advent of the printing press combined with the Manhattan Project, multiplied by the Industrial Revolution, condensed into a decade. The monster, in other words, is already at the door, and the fangs are bared.

Just the other day, Jimmy Apples returned to Twitter. Here is one of his latest retweets:

The greatest story in history is unfolding more or less quietly while everyone’s eyes are glazed over. It hardly matters how obnoxious the average San Francisco booster is, even they underestimate the importance of this place. A true creation event is underway.