Reaction
@sjgadler challenges OpenAI’s recursive self-improvement work: “Um, how about after you solve alignment?”
OpenAI says its RSI team aims to automate research; @sjgadler argues alignment should come first.
TLDR
Responding to a post announcing a new OpenAI hire working with @lilianweng, @sjgadler asked: “Um, how about after you solve alignment?” OpenAI’s job listing says its Recursive Self-Improvement team aims to build AI systems that accelerate and ultimately conduct research. In an earlier reply, @sjgadler urged the new hire to reconsider.
Combined views
43.9K
2 Sources, first seen ago
Comments