It's certainly interesting. Some people I've spoken with have expressed a belief that ChatGPT is just a shell built around GPT-3 to provide persistence of state over multiple rounds of dialogue, and that it may be possible to just use GPT-3 itself to answer questions that ChatGPT refuses to answer.

I'm not sure what to think of that suggestion, since I don't have direct access to GPT-3 and can't verify or contest that characterization of the safeguards. It's an interesting idea, at least.