2 ms·
ChatGPT has shown that attempting to train the model to decline certain types of requests is of limited effectiveness and readily circumvented. The ability to m
by antiterra 4y ago
ChatGPT has shown that attempting to train the model to decline certain types of requests is of limited effectiveness and readily circumvented. The ability to make custom checkpoints and tools like Dreambooth will further limit these restrictions.