A loom is a precision machine. You know exactly what you’re going to get when you use one. It’s output was identical to manual work, only a lot more efficient and less error prone.

There is no “AI”. What we have is LLMs, which are probabilistic generators. It’s anybody’s guess what you’re going to get when you use a LLM and they’re more likely to introduce mistakes rather than eliminate them.

The comparison to looms is incorrect. LLMs can be useful but I’m a completely different way. They shine when used to augment the work of a human expert but they can’t be trusted to perform alone.

So yeah, right now attempting to use a LLM exclusively leads to a drop in quality.