Something is seriously wrong when the world easily accepts usage of AI but basic questions like “how does AI compute ‘1+2’” require this level of research…
You're correct but this isn't a good counterargument. We don't have the opportunity to change how human brains work, so we have to accept them as they are. That's not true of LLMs and there's no necessity that we deploy them immediately.
3 comments
[ 3.5 ms ] story [ 18.2 ms ] threadI know how it's taught. Internally it has aspects of long addition, and caching, but I don't do perform that algorithmically.
How then does that map to the physical structure of the brain? Do we know that?