>LLM’s etc can’t do that under current methodology
I hate to give a smarmy result, but are you sure you know what RLHF is? Because this is one way to correct said data.
>LLM’s etc can’t do that under current methodology
I hate to give a smarmy result, but are you sure you know what RLHF is? Because this is one way to correct said data.
I am aware of RLHF, and no it doesn’t solve this problem.
There’s a great deal of lesions to be learned from X PB of training data that wouldn’t be covered.