 |
| Chess-GPT's Internal World Model (adamkarvonen.github.io) | ||
| recommended by 1 user | 2024 | 120 points by homarp | llm internals | original thread | | ||
| One user recommended this thread. What they said: |
| Miraltar May 15, 2024 | on: Ilya Sutskever to leave OpenAI | context |
This paper and other similar works changed my opinion on that quite a bit. It shows that to perform text prediction, LLMs build complex internal models.
https://news.ycombinator.com/item?id=38893456