We know more than we can tell.

  • Our tacit knowledge cannot be explained

AI Models

When AI models think, it can’t be explained succintly to a human without compression of information. When you compress information, it is bad for safety because you miss out on edge cases.