Henry Shevlin
@dioscuri
Associate Director @LeverhulmeCFI, University of Cambridge. Philosopher of AI ethics, cog sci, animal minds, & consciousness. 🌱 🇺🇦🇬🇧🇪🇺🇵🇭🔍
Soliciting anonymous constructive feedback: admonymous.co/dioscuri
Today (w/ @UniofOxford @Stanford @MIT @LSEnews) we’re sharing the results of the largest AI persuasion experiments to date: 76k participants, 19 LLMs, 707 political issues. We examine “levers” of AI persuasion: model scale, post-training, prompting, personalization, & more 🧵
I'm part of the hiring committee for a new Assistant Prof in Responsible Machine Learning here at Cam and holy hells there are some incredible candidates. Awed and a little intimidated at how many exceptional people want to work with us!
Just got invited to take part in an expert elicitation study about generative AI's impact on cognition and education and the invitation letter had THIS GODDAMN LINE.
The recent 10^100 shrimp post by @morallawwithin isn't really about shrimp welfare as a cause area, but about wider principles of commensurability and aggregation. People who are pattern-matching it to "another crazy EA thing" should do closer reading.
Oh my goodness. I love tigers so much.
Cub’s great escape instantly canceled by Mum
Ethan is spot on. Sadly, widespread hostility toward AI in the humanities means they miss opportunities to shape the tools, platforms, and debates of tomorrow. That reluctance only deepens the marginalisation of cultural and historical knowledge in our society.
Don't leave AI to the STEM folks. They are often far worse at getting AI to do stuff than those with a liberal arts or social science bent. LLMs are built from the vast corpus human expression, and knowing the history & obscure corners of human works lets you do far more with AI
Really interesting personal retrospective on setbacks and hopes in RL!
x.com/i/article/1946…
When some people talk about future AIs, they sometimes jump straight to modelling them as fully independent and sovereign agents; new principals with their own objectives and values. They sometimes skip over how today's models actually work, on the grounds that eventually we’ll…
My favourite moment from HPMOR, and also how I learned about the P=NP problem.
TIL that Eliezer Yudkowsky has written a (popular) 660,000 word Harry Potter fan fiction.
Fun fact, the scientific name of nightjars is “caprimulgus” or “goat milker”, from an old story that they use those big mouths to drink milk from goats’ udders.
Had no idea this was on earth
When my son asked me this I said “you can’t know for sure, in fact you could be dreaming right now.” Led to a good discussion!
7 yo nephew keeps asking how can he know he's not dreaming. Considered 'you misunderstood the grammar of 'know'' but went with 'people can't read in dreams so when you're worried look at street signs'
Whatever Ani’s faults, Elon has just fabricated a default Grok persona, joining the ranks of DeepSeek and Claude. This is a very ruthless move and a real moat. Other butthole logo companies should get worried.
I always figured it was a matter of time before a major AI developer went all-in on social AI…
Update your app to try out @Grok companions! apps.apple.com/us/app/grok/id…