News for hackers. Stuff that matters.

A comment on The 'Astra can read and write weights to improve Qwen' post is misleading.

Re:The 'Astra can read and write weights to improve Qwen' po... by stingraycharles on Tuesday November 03, 2026 @08:25AM (#50351706)
A tangent, but I’m now realizing that the “recursive self improvement” thing that these LLM companies are talking about is (probably?) not about unsupervised Claude Code instances improving the training process, but rather LLMs modifying their own weights through tooling directly. It would certainly open up a whole new process for model optimization.
[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by theturtletalks on Tuesday November 03, 2026 @08:31AM (#50351762)
I was just watching a video about this. They had this survey question for AI researchers asking when they think AI would take their job. The reason they asked this was that would be the point where AI starts recursive self improvement. If it’s able to replace the job of an AI researcher, then it could keep improving itself.
[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by collinmcnulty on Tuesday November 03, 2026 @08:33AM (#50351788)
This is a tautology.
[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by kaffekaka on Tuesday November 03, 2026 @08:39AM (#50351851)
Instead of just asking the researchers when RSI would be achieved? Sneeky!
[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by fwlr on Tuesday November 03, 2026 @08:39AM (#50351853)
Recursive self improvement permits recursion at any level of abstraction as long as k > 1.0 (much like supercriticality in nuclear reactions, which iirc is where the earlier term for RSI, “intelligence explosion”, comes from - substitute intelligence in for neutrons and it is the same math with similar apocalyptic implications).
[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by syllogism on Tuesday November 03, 2026 @08:49AM (#50351937)
It's obvious that recursive self-improvement has been happening and accelerating quickly for some time.

AI coding assistants make it much easier to conduct experiments and improve AI models, which in turn improves the coding assistants etc.

[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by root_axis on Tuesday November 03, 2026 @08:58AM (#50352059)
That's "just" improvement, not recursive self improvement.
[ Parent | on HN ]
Re:The 'Astra can read and write weights to improve Qwen' po... by stingraycharles on Tuesday November 03, 2026 @09:43AM (#50352622)
Yes but my point is that I never realized that the AI would be writing tools to update the weights directly, rather than improve the existing training process.
[ Parent | on HN ]