10 ms·
There was a recent paper about a way to censor LLMs by just deleting the connections to any bad outputs, rather than training it to refuse them. I think this te
by astrange 2y ago
There was a recent paper about a way to censor LLMs by just deleting the connections to any bad outputs, rather than training it to refuse them. I think this technique wouldn't work.
Obviously you could train any bad outputs back into them if you have the model weights.
- stainablesteel 2y agointeresting, there's going to be an arms race over censoring and uncensoring future powerful llms a lot like the getting a cracked version of photoshop back in the day