Re:Astra can read and write weights to improve Qwen
by jaynetics on Monday November 02, 2026 @04:05PM
(#50343746)
That's like saying the moon landing wasn't impressive because the US were in a space race with the soviets.
Re:Astra can read and write weights to improve Qwen
by tintor on Monday November 02, 2026 @06:44PM
(#50345876)
That is not what they are saying. Frontier labs are incentivized to train their models on all ML knowledge and techniques that people at those companies posses, so that current frontier models can assist them with developing the next frontier models.
Re:Astra can read and write weights to improve Qwen
by im3w1l on Monday November 02, 2026 @03:34PM
(#50343330)
Of course they try train their models on these techniques. Finding success is still news.
Re:Astra can read and write weights to improve Qwen
by fmbb on Monday November 02, 2026 @03:52PM
(#50343543)
Why?
How is this different from other things LLMs are trained to manipulate?
Re:Astra can read and write weights to improve Qwen
by zardo on Monday November 02, 2026 @04:15PM
(#50343869)
If it's new it's news.
Re:Astra can read and write weights to improve Qwen
by im3w1l on Monday November 02, 2026 @05:08PM
(#50344576)
We aren't fully at AGI yet, even though it's starting to get damn close. So every new capability is in some sense unexpected.
Re:Astra can read and write weights to improve Qwen
by senordevnyc on Monday November 02, 2026 @04:38PM
(#50344171)
Why would they train the models they release with this though?
OpenAI, Anthropic, etc are locked in a race to create better models. They all have interpretability teams. Of course they train their models on every technique under the sun to understand and improve their models, including raw weights. The primary function of these models is to build their successor.