Thinking that now is pretty much insane...
And even then, if LLMs only improve with more compute, you'd have to be crazy to bet that they won't do it.
Thinking that now is pretty much insane...
And even then, if LLMs only improve with more compute, you'd have to be crazy to bet that they won't do it.
www.anthropic.com/institute/ec...
www.anthropic.com/institute/ec...
how the fuck will we get through curtailment for misaligned groups training models aligned to their intent
how the fuck will we get through curtailment for misaligned groups training models aligned to their intent
On the other versions I assume the comparison is a bit too dark
On the other versions I assume the comparison is a bit too dark