4 ms·
It is getting to the heart of the problem when the claim made is that "no matter how advanced the model" they can't be 'much more than just "really good autocom
by vidarh 1y ago
It is getting to the heart of the problem when the claim made is that "no matter how advanced the model" they can't be 'much more than just "really good autocomplete."'.
Given that they are Turing complete when you put a loop around them, that claim is objectively false.
- jdiff 1y agoI think it'd even be easier to coerce standard autocomplete into demonstrating Turing completeness. And without burning millions of dollars of GPU hours on training it.