4 ms·
I’m sure the lawyers will eventually figure out a way to train an LLM on them.
by senkora 2y ago
I’m sure the lawyers will eventually figure out a way to train an LLM on them.
- datadrivenangel 2y agoThey probably already have! It seems like an amazing training dataset even if you can't share source data.
- amelius 2y agoHow do you train an LLM such that it is guaranteed to never regurgitate its training data?
- ASalazarMX 2y agoYou punish it if parts of the answer can be found in its training data, and reward it otherwise.
- amelius 2y agoBut the whole point of the training is that you reward it if it correctly reproduces the next token.
- zeroxfe 2y agoThat's not the whole point of the training. It's just (very loosely) a measure of loss used during pre-training. There are many post-training and alignment stages in a typical model that are designed to reward high-quality responses. Technically, yes, it's impossible to guarantee that it won't just regurgitate source material (which is mostly around the tails of the data distribution), but the whole point of training is to build generalized intelligence.
- amelius 2y agoI guess I used the wrong wording but it doesn't change the argument. Yes, the whole point of training is to build generalized intelligence (or at least that's what we __hope__ for). But as far as I understand, we do it __mainly__ by training for the next word in the sequence. PS: you speak of "pre-training" and "post-training", so I'm curious what you think is the main part of the training (?)