3 ms·
> These models are trained to be truthful A more accurate statement would be that these models are trained to fit the training data as closely as possible, reg
by lossolo 3mo ago
> These models are trained to be truthful
A more accurate statement would be that these models are trained to fit the training data as closely as possible, regardless of whether the training data reflects the truth.
- altcognito 3mo agoWell, yes, but remember there is the reinforcement learning that is applied after, and the system prompts that will bend the results.
- lossolo 3mo agoYeah, agree on both points. You can embed any bias you want using RL, regardless of training data.
- slibhb 3mo agoNo, they're very clearly trained to be truthful.