Mark Chen 演講筆記 teaching GPTs to reason

 發表於 2024, November 25

2024-11-25 by Mark Chen

MIT 畢業，去作 quant，然後加入 openai

history of scaling gpt

  • GPT-2, build model to solve one thing
  • GPT-3, inner and outer loop. Can generate news that fools human
  • GPT-4, perform well on exams(高中生). Scales predictably important for safety
[閱讀全文]
2024_Fall  Speech