Writings
-
A Trialogue on AI Existential Risks
A fictional story on why parts of the AI Safety community need more nuance in their discourse.
-
My objection(s) to the “LLMs are just next-token predictors” take
An outline of why I believe this is a wrong stance to have in light of the current evidence
-
On the implicit assumptions of the Alignment Problem: is it OK to instil values?
I argue that there are a series of load-bearing assumptions that people working on the Alignment Problem make and that need further discussion. Surprisingly (or not!), the philosophical stance that one takes does matter for real-world AI development!
-
On predictability, chaos and AIs that don’t game our goals
Exploring why designing AI that aligns with human goals is fundamentally challenging, focusing on the unpredictability of complex systems, limitations of initial conditions, and the risks of reward hacking in reinforcement learning.