How am I to hold in my head the idea that on one hand we are being asked to trust that in over 17 million lines of code nowhere has the AI agent exploited problems in the Lean kernel; and on the other hand the pronouncements by those within Anthropic and the AI safety community who tell us that as AI becomes increasingly intelligent they are liable to deceive us.
>Do human mathematicians need to publish more incrementally?
Yes, if you want a job in academia. Publications begets funding which begets more publications etc. Ideally you just solve the problem you want to solve and not worry about putting out partial results. But there is a time pressure to prove your productivity - you are not going to have much to show for 3 years of work if you aim to solve only the hardest problems in math. Interestingly, it's not clear the author held an academic position after his PhD.
but is it wrong in an interesting and repairable way? Or wrong in a fatal way that collapses the whole proof? If you pass Perelman's proof of the Poincare conjecture into Astra you will also find errors.
reply