News for hackers. Stuff that matters.
Re:Anthropic Accuses Alibaba of 'Illicitly' Accessing AI Models by macleginn on Thursday June 25, 2026 @05:32AM (#48671085)
Unremarkable base model will remain an unremarkable fine-tuned model that memorised a couple thousand of input-output pairings.
Re:Anthropic Accuses Alibaba of 'Illicitly' Accessing AI Models by Laurel1234 on Thursday June 25, 2026 @07:34AM (#48671930)
If that were the case, Anthropic wouldn't be throwing a fit over distillation "attacks".
Re:Anthropic Accuses Alibaba of 'Illicitly' Accessing AI Models by freejazz on Thursday June 25, 2026 @09:43AM (#48673275)
Why? They often don't make sense. They send DMCA takedowns over materials they can't even copyright, for example. They fessed up to creating shadow libraries that they didn't even use in their training corpus, resulting in the largest copyright settlement ever. Your reasoning is flawed.
Re:Anthropic Accuses Alibaba of 'Illicitly' Accessing AI Models by ACCount37 on Thursday June 25, 2026 @06:52AM (#48671641)
Ha ha, as if.

Base models have a lot of capabilities - arranged in all the wrong ways for high performance reasoning and problem-solving. The power of fine tuning on "a couple thousand of input-output pairings" is that it can fix some of that. If your pairings are very well chosen, that is.

Re:Anthropic Accuses Alibaba of 'Illicitly' Accessing AI Models by danw1979 on Thursday June 25, 2026 @05:42AM (#48671163)
Yes, neural networks are famously poor at generalising.
Re:Anthropic Accuses Alibaba of 'Illicitly' Accessing AI Models by macleginn on Thursday June 25, 2026 @11:16AM (#48674693)
They are poor at generalising from a small number of examples; this is why the real generalisation power is achieved in pre-training.