Comment on OpenAI's reported 'superintelligence' breakthrough is so big it nearly destroyed the company, and ChatGPT
I think the takeaway is that they’re trying to create a LLM that can answer questions that it wasn’t trained on.