I yield to nobody, as a rule, in my general horror at the march of AI. Sure, there are the anxious-making medium-term prospects such as human extinction, the further enrichment of Sam Altman, and the complete impossibility of getting someone from Vodafone to explain what’s wrong with your broadband. But there’s a laundry list of present-tense problems, too. Have you tried to turn off CoPilot lately?
It’s exciting to think that all these high-level proofs buried in squiggles by the historic greats can now be brought to light
We really don’t need to go through the list again; nor, probably, to rehearse the retorts of the AI boosters, which tend to be to mutter about protein-folding and radiology and extrapolate from that to the abolition of work, the dawn of human immortality and a whole lot of other stuff from Iain M Banks’s Culture novels. But I’d like, in the interests of basic fairness, to draw attention here to something that really does look like a big red tick in the proven-positive-effects-of-AI box.
It’s niche, but my goodness it is pleasing, When he died, the brilliant mathematician Kurt Godel (remember him? Bloke who did a great deal to jam a banana in the exhaust-pipe of logical positivism in the first half of the last century) left behind more than 1,000 pages of unpublished mathematical noodlings in 16 notebooks. Not only did these workbooks contain what everyone was pretty sure were some wicked hard sums, but Godel further confounded posterity by putting his notes-to-self down in a form of German shorthand called Gabelsberger, which was obsolete even when he was using it, and which now only six people on earth know how to read.
Enter a Finnish logician called Jan von Plato, who has spent the last ten years trying to make sense of these notebooks, a project – this is also a pleasing detail – that he says he only started in order to avoid being given compulsory retirement by the University of Helsinki. In his words: “What should I do to make myself indispensable? Oh yes, I’ll start reading Gödel’s enigmatic shorthand pages!”
So he beavered away at this with Nordic determination until August this year, when some computer scientists suggest he see if Anthropic’s AI model Claude might be able to help. He fed his work into Claude, and bingo: in two shakes of a lamb’s tail, he had a full rundown of Godel’s unpublished work and discovered that the great man, who barely published between the 1930s and his death in 1978, had tackled 663 theorems and proved 400 of them. He could have had the Fields Medal for his work on the axiom of choice 26 years before somebody else won it.
I think that is just marvellous – and it’s a work of decipherment, synthesis and evaluation that, as von Plato is happy to admit, no human being would have been capable of. If it can happen to Godel’s unpublished work, there’s going to be more gold in them there hills. There are many thousands of pages of unpublished maths by Alexander Grothendieck (remember him? Brilliant mathematician last seen scampering around the Pyrenees raving about the devil) kicking around, for instance.
I can see why some mathematicians are a little uneasy about AI shooting their fox: it seems to have a habit of solving very difficult problems in maths in a way that might put academic mathematicians out of jobs. But it’s exciting to think that all these high-level proofs buried in squiggles by the historic greats can now be brought to light.
That points the way, too, to something that goes back much further than just the 1940s. Most Spectator readers, being well-educated sorts, will be familiar with the Oxyrhynchus papyrus. These were discovered in 1896, when a pair of British classicists were poking around in the rubbish dumps of an unimportant town 100 miles or so from Cairo. Long story short, they hit the absolute motherlode. “The flow of papyri,” wrote one of them, “would soon become a torrent. Merely turning up the soil with one’s boot would frequently disclose a layer.” In the years that followed more than half a million fragments of papyrus (perfectly preserved thanks to the bone-dry conditions in those parts) were gathered up – along, admittedly, with much of the other sort of stuff you’d expect to find at a town dump.
The rubbish heaps of Oxyrhynchus have already yielded lost poem-fragments of Sappho, Pindar and Homer, diagrams from Euclid’s pioneering mathematical work, bits of Plato and an unknown play by Sophocles, Old Testament apocrypha and some of the earliest surviving Greek texts of the New Testament. But the pace of putting them together has been glacial: 98 or 99 per cent of the stash remains unread and unpublished 130 years after its discovery. Why? Imagine hundreds of thousands of thousand-piece jigsaw puzzles all jumbled up, and – what’s more fun – many of these pieces are scrumpled up or wrapped round other pieces and are so fragile that unwrapping them will reduce them instantly to dust.
But you know what’s good at doing jigsaw puzzles (as well as deciphering handwriting in dead languages, and reading x-ray tomography, and digitally “flattening” curved texts, and retrieving ink marks invisible to the human eye, and much else besides)? Yup. AI. So it is not at all too much to hope that AI, in concert with these other technologies, will open a whole body of lost knowledge and art from the ancient world that, like Kurt Godel’s lost proofs, would otherwise have been inaccessible to human posterity. Edmund Richardson, author of this year’s superb new biography Alexander: God, King, Man, has spoken excitedly of this prospect in interviews, and he’s not the only classicist I’ve seen do so.
It’s an extraordinarily thrilling idea, isn’t it, that tech we think of as belonging to the bleeding edge of the future should have such a hand in reconnecting us to our past. And it will be a comfort, at least, to have these texts as p>doom nudges up towards 100 per cent. We can settle in for a wan giggle with the lost book of Aristotle’s Poetics – the one on comedy – even as the buzzing of the drone-swarms grows louder.
The post How AI is unlocking the mysteries of history appeared first on The Spectator Australia.


