They are at 4000 tokens.. or is it 8000 now? That alone without training is enough to handle many specific use cases without further training.
The OpenAI API doesn't have an option to forget a bunch of stuff though so I doubt it's practical to saturate the existing model enough with a specific task that it forgets.
I agree with you but it doesn't seem like people are going to do that.
The OpenAI API doesn't have an option to forget a bunch of stuff though so I doubt it's practical to saturate the existing model enough with a specific task that it forgets.
I agree with you but it doesn't seem like people are going to do that.