AI Research
AI Just Edited Another AI’s Brain
A tiny change to a small language model reportedly produced a startling leap on math and instruction-following tests. The catch: removing a flaw is not the same as teaching new skills—or building an AI that can improve itself.
Read article→