Can we find whether models have been backdoored?
This is the first post in a two-part sequence regarding the state of defending from data poisoning attacks. We describe some methods for determining whether a…
Latest
Info dump: Some Important Models for Health and Fitness
This is a synthesis of many facts I've learned over the last few years, mainly about metabolism and exercise,[1] that have helped me become much healthier,…
Claude Code as a Claude Coach
Exercise is hard but it's even harder if you have to use your brain and muscles at the same time. I wish a personal trainer would…
Be less open-minded: Against uncalibrated open-mindedness
People sometimes tell me to be more open-minded. That it’s always a positive trait, all else being equal.I disagree.I think what they mean is that you…
A Review of Anthropic's Global Workspace Paper
The below is a public review Anthropic asked me to write for their new global workspace paper. I recommend at least skimming their paper first.TLDR:I think…
Visioning: Concretely Imagining What You Want
When John told me (Gretta) his practice of “visioning,” I was skeptical at first. I gave it a try, a little bit out of spite, to…
Personhood for digital minds is good
AI agents are sophisticated enough to engage in economic transactions. Some people also grant personhood to AI in the social sense. It’s time to ask whether…
Training AI to be better at correctness than persuasion
Cross-posted from my website. I continue to believe we should pause frontier AI development. Any discussion of alternative strategies should be thought of as planning for…
A global workspace in language models
[This is the blog post for our new paper Verbalizable Representations Form a Global Workspace in Language ModelsReaders might also be interested in: the Public commentary,…
Current views on large-scale longtermist philanthropy
This post summarizes some of my views. It states uncertain things boldly in part to [surface disagreements / help readers point out things that seem wrong…