a cool quote from Jonathan Gorard: 'the distinction between a good model and a bad model is that a good model is one where everything that can be emergent is emergent'

Thanks so much for making this possible 🙏🏻 looking forward to the exciting science!!!

Richard Sutton says AI safety advocates are creating the opposite of what they seek - we should not try to solve the AI control problem because a decentralized system of independent agents is better than everything being aligned under one goal

This rule also applies spatially not just temporarily. If you’re a researcher in tech (or any field) and you’re not reading anything five metres away from you (metaphorically speaking), your own work will be forgotten five metres from you.

Oh yes please

torch compiling both model+loss_fn is giving me significant speed-ups in comparison to compiling just the model. Seems underrated.

It’s 2004. Last year, the Nobel Peace Price was awarded for the solving of all AI alignment and safety issues. Now, in his dorm, young Mark Zuckerberg asks his AGI agent to help him code up a social network. What happens?

Reminder to not take things too seriously

New @bioRxiv #preprint with @chiara_fornetto and T Euler (who left X): biorxiv.org/content/10.110…