Help! We need a nuanced discussion. Can ai help?

I find this quite weird. So often the deception research outcomes appear so “human-like” eg the models “lie” or whatever in similar ways (or at least matching my intuition) aka the mirroring of human behavior. But this seems so much more “unfaithful” than I would have expected compared to humans

I heard that they all chat to Claude, rated the best at calming their neural net nerves

That’s so cool #hopeforthefuture

You’re pointing to the important shift people need to get their heads around - more false positives than false negatives. Govt moving from play-not-to-lose to play-to-win mindset and ideology. The question - what error are we prepared to live with?

“genome as a generative model” 🤔

I am consistently fascinated by how every new benchmark reveals something else we don’t know or creates the need for a new word or breaks an existing word.

States rights

He’s in our 2024 @artificiality.bsky.social book awards! A few other MIT Press pubs too! www.artificiality.world/artificialit...

You can check out any time you like but you can never leave

Me too! Cant wait to read it

Invented by ChatGPT? 😵‍💫. Prompt: “give me ten ideas for a convenience food for gen z-ers who need to stay hands free and slightly dehydrated. Be creative and work harder!”

The next word Silicon Valley will destroy. After “intelligence,” “ethics,” and “agency.”

The question is at what point is it self defeating? I wish SV didn’t believe its own bs so much they could see that it’s not really worth destroying the very resource they need. And lots of good voices in there - need to keep them talking