Predictions Over What?
Accuracy is only half the battle
Crossing the Rubicon
Posts about AI alignment. Victory or death.
Subscribe:.rss.atom.json.md.m3u.pls
Live Last read · last published · next check
Latest episodes
Catching Misalignment vs. Fixing It
What happens if evaluations, interpretability, and monitoring succeed?
Emergent Misalignment, Inoculation Prompting, and Unsolved Problems
The AI Alignment ecosystem working as intended?
Solutions to Corrigibility, Deceptive Alignment, Reward Tampering, Monitorability, and Reward Hacking
I am not a crackpot
Probably Safe AI
A Case for Marginal Progress on Superalignment Theory
Myopic Goals without Myopic Capabilities
How we can get AI to set up a good future without caring about it
Intelligence Explosion Macroeconomics
Economists believe in an intelligence explosion, even if they don’t realize it yet
Safe Predictive Agents with Joint Scoring Rules
It's time to talk about what I've been working on
Myopia Is All You Need (For Alignment)
That's not strictly true, but myopia is still incredibly useful
Simplifying Corrigibility – Subagent Corrigibility Is Not Anti-Natural
Ensuring corrigibility in an AI and ensuring corrigibility in the agents it creates are two completely distinct problems.