Post by @lurker@awful.systems

2026-10-19 00:17 UTC

Might as well start this off so we all remember this graph from AI 2027 right? the graph that the whole paper relies on? well its wrong. not just normal wrong, it’s ‘the authors literally disowned it’ wrong (ignore the claude mythos purple x i took this from a reddit screenshot) but they completely changed their predictions in december 2025, several months after the original paper was published

Replies (2)

@rhamph@awful.systems 2026-10-19 05:54

That ‘superexponential’ thing annoys me so much. Just the most blatant ‘the line goes up because I really really want it to’, while all their rationalisations for it are just describing exponential behaviour. We talk about the importance of humanities but it’d be nice if they learnt a bit of maths too.

Open ##5128920
@scruiser@awful.systems 2026-10-20 02:29

Is this the thing where the baked an asymptote to infinity into their “model”? It was both bad modeling and also made all of their other input “data” completely irrelevant. I remember someone pointing this out at the time AI 2027 dropped, but there was so many other problems with it in the sneer spaces we didn’t talk about it too much among all the other problems. And someone on lesswrong actually pointed it out, but somehow, “we accidentally baked an asymptote into our model that is supposed to be detailed and carefully thought out” wasn’t a breaking point for lesswrongers.

Open ##5152488