| The risk of AI can often seem abstract. Here is a way to see them in two tables.
The first is from OpenAI & shows what GPT-4 was capable of before alignment cdn.openai.com/papers/gpt-4.p…
The second is a new paper from the major AI companies on long-term risks arxiv.org/pdf/2305.15324… pic.twitter.com/syOPtBTMG4 |
Thornton Neuroethics 