Daniel Tan

@DanielCHTan97

alignment research lead @arcadiaimpact. prefer email to DM (unless mutuals) ai labs should slow down / pause

London, UK

Daniel Tan
@DanielCHTan97
Nov 4
@jankulveit has a much better analogy IMO, in that being in an in-group is like microdosing a psychedelic substance resulting in “vigorous debate in a synchronised attention field” lesswrong.com/posts/oL7aFoRm…
Daniel Tan
@DanielCHTan97
Nov 3
whoever said the rationalists are a cult clearly failed to appreciate the extent of internal disagreement
Daniel Tan
@DanielCHTan97
Nov 3
sounds like an easy followup is to just run the same experiment(s) on tons of older models!
Tyler Tracy
@tylertracy321
Nov 3
This is the type of thing we could have gone a long time not knowing because no one tried it. Probably it sounds bonkers but this mad lad tried it and it worked.
2
Daniel Tan
@DanielCHTan97
Nov 3
whoever said the rationalists are a cult clearly failed to appreciate the extent of internal disagreement
18
Daniel Tan
@DanielCHTan97
Oct 30
people with adhd forget who they are every day and need to reinvent large aspects of their life from first principles. this probably gives them an accurate felt sense of what it’s like to be an llm
1
Daniel Tan
@DanielCHTan97
Oct 27
So tl;dr 1. OpenAI is power-seeking, and pretty transparent about this 2. Anthropic is power-seeking but tries to appear as if they are good This "overt" and "covert" misalignment is reflected in GPT and Claude respectively lesswrong.com/posts/nLaQmJf4…
Mikhail Samin
@Mihonarium
Oct 26
From an ex-Anthropic employee:
7

Daniel Tan’s posts