3 ms·
There's a _lot_ of evidence that LLM's _do_ generalize, though.
by empath-nirvana 3y ago
There's a _lot_ of evidence that LLM's _do_ generalize, though.
- mjburgess 3y agoThere's many notions of "prediction" and "generalisation" -- the relevant ones here, which apply to NNs, are extremely limited. That's the problem with all this deceptive language -- it invites people to think NNs predict in the sense of simulate, and generalise in the sense of "apply across different effect domains". NNs cannot apply a 'concept' across different 'effect' domains, because they have only one effect domain: the training data. They are just models of how the effect shows itself in that data. This is why they do not have world models: they are not generalising data by building an effect-neutral model of something; theyre just modelling its effects. Compare having a model of 3D vs. a model of shadows of a fixed number of 3D objects. NNs generalise in the sense that they can still predict for shadows similar to their training set. They cannot predict 3d; and with sufficiently novel objects, fail catastrophically.
- lossolo 3y agoThere is also a lot of evidence lately that they do not generalize. https://arxiv.org/abs/2311.00871 https://arxiv.org/abs/2311.00871 https://arxiv.org/abs/2309.13638 https://arxiv.org/abs/2309.13638 https://arxiv.org/abs/2311.09247 https://arxiv.org/abs/2311.09247 https://arxiv.org/abs/2305.18654 https://arxiv.org/abs/2305.18654