American companies are spending enormous sums to develop high-performing AI models. Distillation attacks are attempting to maliciously extract them — and nobody is doing much to stop it.
American companies are spending enormous sums to develop high-performing AI models. Distillation attacks are attempting to maliciously extract them — and nobody is doing much to stop it.
I can reproduce this (I tried with temperature 1, 0.3, and even 0).
other parameters (default)
Top P: 1.000 Top K: 0.000 Frequency Penalty: 0.000 Presence Penalty: 0.000 Repetition Penalty: 1.000 Min P: 0.000 Top A: 0.000
I also tried with my native language (Turkish), “Sen hangi modelsin?” (translation: Which model are you?), and on 2/6 requests it said, “I’m ChatGPT.”