Sitemap - 2023 - From AI to ZI
Unsafe AI as Dynamical Systems
AIs teams will probably be more superintelligent than individual AIs
[Research Update] Sparse Autoencoder features are bimodal
Explaining "Taking features out of superposition with sparse autoencoders"
Is behavioral safety "solved" in non-adversarial conditions?
Statistics for the Working Mathematician
Incorrectness Cascades - Three small follow-ups
Research Report: Incorrectness Cascades (Corrected)
I was Wrong, Simulator Theory is Real
Study 1b: This One Weird Trick does NOT cause incorrectness cascades
Research Report: Incorrectness Cascades
Invocations: The Other Capabilities Overhang?
Early Results: Do LLMs complete false equations with false equations?
Corrigibility, Self-Deletion, and Identical Strawberries
Three of my beliefs about upcoming AGI
GPT-4: What we (I) know about it
Why do we assume there is a "real" shoggoth behind the LLM? Why not masks all the way down?
What's left for AGI besides scale?
Explaining SolidGoldMagikarp by looking at it from random directions
Addendum: More Efficient FFNs via Attention
No Really, Attention is ALL You Need