AI and the question forced open

est 17.5h · logged 0h · started — · touched —

LLMs, machine consciousness, and what the easy problems are teaching us: large language models got surprisingly capable around 2017–2023, forcing the field to answer questions it had been deferring. The most current and most unsettled unit.

next action

done when you can

State why LLM consciousness is a question that can't be cleanly answered from either direction.Explain one concrete test that's been proposed (Chalmers, Butlin et al. 2023) and why it's hard.Distinguish the philosophical question ("is this system conscious?") from the functional question ("does this system behave as if it understands?").Articulate your own position provisionally, knowing the field will look different in 5 years.

resources

○Could a Large Language Model Be Conscious?— David Chalmersanchor · paper
○Talking About Large Language Models— Murray Shanahancompanion · paper
○Climbing towards NLU— Bender & Kollercompanion · paper
○Geoffrey Hinton and Scott Aaronson TOE episodesoptional · podcast series

sessions

unlocks Everything. This unit is the one most in motion right now, so finishing it means you're roughly current on a question that's being argued out in real time.

open in notion ↗