this post was submitted on 28 Jun 2025
227 points (91.9% liked)

Fuck AI

3358 readers
1192 users here now

"We did it, Patrick! We made a technological breakthrough!"

A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.

founded 1 year ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] pinball_wizard@lemmy.zip 3 points 1 week ago* (last edited 1 week ago)

If there was an art search engine that indexed every piece of human-made art that could be found and allowed you to search for it in natural language, would you use it? This would be different from a search engine in that natural language allows for more clarity of context and emphasis.

Like all other LLMs, I would use it if I lacked the skill to use a better, more accurate, less halucination-prone search engine.

There's a great place for LLMs - as training wheels for non-experts.

  • Hard coded solutions designed thoughtfully are the best solutions.
  • Dedicated query languages that take skill to learn are the best solutions.
  • People who use the command line get better results from their computers.

But not everyone needs to use a carefully engineered command line interface, but those who do so get better results.

For about 80% of all problems, a halucination-prone human language interface will get the job done. This is lovely and opens up exciting opportunities for non-experts.

The problem is when people believe the lie that the human language interface to the computer gets better results for an expert. It still doesn't. It never will.

A rock that we taught to think will always be most powerful when operated directly using special "rock-that-we-taught-to-think" codes. Computers are still computers, and they are operated most powerfully and accurately by using computer languages. No abstraction has ever changed this.

Usually that doesn't matter anymore. But sometimes it still does.

And abstractions are still great. They make computers useful to more people.

Most domains already have specialized human computer interface languages. Experts in those domains already know those languages. Experts may find LLMs convenient, but LLMs do not meaningfully increase the capabilities of an expert.

But LLMs can help the rest of us, and the interface you purpose would be cool and useful, to the rest of us.