Dan Williams

This account is verified. Learn more

@danwilliamsphil

Philosopher, University of Sussex. Tweets in personal capacity. Interested in: Philosophy, Psychology, Society. Writes at: conspicuouscognition.com

Brighton England danwilliamsphilosophy.com Joined November 2010

3.8K Following 15K Followers 12K Posts

Latest posts 5

Dan Williams @danwilliamsphil 8h

AI right now seems very bad at coming up with its own ideas, but very good at identifying significant flaws in any ideas you give it. (I know people like this, in fairness.) This will probably have interesting effects on creativity.

I wonder how many great ideas throughout history would have been quickly abandoned if their creators had access to a machine that could instantly point out a dozen flaws in them.

Dan Williams @danwilliamsphil 8h

Yeah this is a significant problem that deserves a lot more attention from those interested in using AI to improve collective epistemics.

Thomas Renault @captaineco_fr Nov 3

⚠️ Concentration: A small number of AI Note Writers, including accounts operated by X itself, produce a substantial share of AI-generated notes, raising questions about diversity and independence.
View quoted post on X

Dan Williams @danwilliamsphil 9h

It's not true that AI is "smarter" than us in most ways. Or at least, this rests on a subjective, observer-relative notion of smartness that tracks what we happen to find hard - coding, maths, abstract intellectual tasks - and ignores what we find easy but is in fact much harder from an engineering perspective: the everyday embodied physical and social intelligence that underlies most of what most people do and that AI is still terrible at.

Noah Smith 🐇🇺🇸🇺🇦🇹🇼 Verified account @Noahpinion 10h

The thing is, AI is already smarter than is in most ways, and continuing to get smarter. So even if there's no intelligence explosion, AI will soon be capable of destroying humanity (if it's not already).
View quoted post on X

Dan Williams @danwilliamsphil 9h

Important and worth reading even if, like me, you are very sceptical 👇

Daniel Kokotajlo Verified account @DKokotajlo 11h

Unfortunately, it seems like the intelligence explosion will be happening in 2027 after all. Most of this post analyzes the new METR report (though other sources of evidence are also important). If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario. I am scared, sad, and angry. blog.aifutures.org/p/q35-timeline…
View quoted post on X

Dan Williams @danwilliamsphil 4d

On Amodei's claim as a graduate student that adult deaths are worse than infant deaths, and the polarised discourse about this:

It seems the main thing separating those who are fine with thinking this way, even if they disagree with the specific conclusion (effective altruists, neoclassical economists, some analytic philosophers), from those outraged about it (most other people) concerns the psychology of the sacred and the taboo.

The former group generally thinks this psychology is something to resist in ethics. The latter doesn't.

This isn't a matter of religion. Most secular people have sacred values in the sense in which this term is understood in psychological research. 👇