@_akhaliq @mhutter42 haven't checked it but mmh it smells a lot like Solomonoff's reference UTM

@nouhadziri @DimitrisPapail It also makes me think about universality: is it really the right concept? Couldn't we get, say, a proof of the RH in a few years, with a model that can't sum and add two numbers?

GPU hours is the new bitcoin

RT @francoisfleuret: Here you know you should pull up the logit of the class you proposed when you get +1 and push it down with -1, but tha…

Does anyone know/can guess how they made fp8 training actually work? That's probably the most impressive to me. This would mean sparse training with gradients is indeed possible.

RT @gushalls: @bryan_johnson People overestimate AI in the short run & underestimate AI in the long run.

@yacineMTB So you know! Interesting. Will read your tweets more carefully.

RT @mhutter42: Santa Arrived! The PDF (of a colorful Xmas version) of the "Introduction to Universal AI" book is now freely available onlin…

@prajdabre1 Negative log-loss, wow! You have probabilities higher than one my friend 😅

The "online" part is where it gets dangerous. Without it, it's only the humans that are dangerous, not the AI.

@torchcompiled @neurallambda

@fouriergalois Then yes that's exactly the system I think we need!

If the Riemann hypothesis or smthg of the same importance is solved by a deep neural network, what will happen? Will there be even more investments and a wake-up of general society? Or will it just be the same "yes but it's just another task"?

AI benchmarks are just falling like dominos... Looks to me that this is not playing anymore. The benchmark now is the real world.

So, how much is the Riemann hypothesis worth in test-time compute dollars?