Ahem… yikes.

Ars Technica, Benj Edwards: OpenAI’s GPT-4 exhibits “human-level performance” on professional benchmarks