I'm curious to open a discussion about whether we've reached a moment of recursive improvements in technology, particularly with tools like Cursor, Windsurf, Claude Code, Codex, and other LLM-driven systems. Recently, there was news about a Google project that developed a new mathematical proof for more efficient matrix multiplications. Are we now able to confidently claim that we've entered a phase of recursive automated self-improvement for AIs?
1 Answer
While it's true that we've made significant strides, we're not quite at the point of self-sustaining recursive self-improvement yet. Human input is still crucial, like a spark in an engine that needs a bit of a push to keep it running. Those occasional ‘ah-ha!’ moments are promising, though, indicating we might get there soon!
Absolutely, humans still play a big role. It feels like we're on a blurry line between human-driven improvements and machines starting to run the show themselves.