zlacker

>>modele+(OP)
AI is going to continue to have incremental progress, particularly now in hardware gains. No one can even define what AGI is or what it will look like, let alone be something that OpenAI would own? Features progress is too incremental to suddenly pop out with "AGI". Fighting about it seems a distraction.

>>okhuma+3k
There's also no reason to believe that incremental progress in transformer models will eventually lead to "AGI".

>>root_a+kO
Yes, but I think everyone would agree that the chance isn't 0%

>>snapca+qQ
I don't agree, I think many people would argue the chance is 0%.

>>root_a+sR
Are you one of those people? how can you be so confident? I think everyone should have updated their priors after how surprising the emergent behavior in GPT3+ are

>>snapca+bS
Perhaps you should update your priors about "emergent behavior" in GPT3+: https://arxiv.org/abs/2304.15004

>>nickle+rU
This is like saying that nothing special happens to water at 100 degrees because if you look at the total thermal energy, it's a smooth increase.

>>famous+OX1
Please read the paper. The authors are using more precise and specific metrics that qualitatively measure the same thing. Instead of having exact string match being 1 if 100% correct, 0 if there is any failure, they use per-token error. The crux of their argument is that per-token error is a better choice of metric anyway, and the fact that "emergent abilities" do not occur when using this metric is a strong argument that those abilities don't really exist.

However thermal energy does not more precisely or specifically measure a phase transition. They are only indirectly linked - nobody would say that thermal energy is a better measure of state-of-matter than solid/liquid/gas. Your argument makes absolutely zero sense. Frankly it seems intentionally ignorant.