674
Generated "art" does not rule.
(lemmy.world)
Be sure to follow the rule before you head out.
Rule: You must post before you leave.
Im sorry but if your arguments is that "AI is doomed because current LLMs are only good at fuzzy, probabilistic, outcomes", then you do not understand current AI or computer science or why computer scientists are impressed by modern AI.
Discrete concrete logic is what computers have always been good at. That is easy. What has been difficult, is finding a way for computers to address fuzzy, pattern matching, probabilistic problems. The fact that Neural Networks are good at those is precisely what has Computer Scientists excited about AI.
I'm not saying it's doomed! I literally said that it's cool and useful. It's a revolutionary technology in many respects, but not for everything. It cannot replace the things computers have always been good at, but business people don't seem to realize that. They assume that it can fix anything, not understanding that it will only make certain things worse. The trade-off is counterproductive for tasks where you need consistent indexing.
For instance, Google's search AI turns primary sources into secondary or tertiary sources by trying to cut corners. I have zero trust in anything it tries to tell me, while all the problems it had before AI have continued to worsen. They could've used machine learning to better understand search queries, or diversify results to compensate for vagueness in language, or to fucking combat SEO, but they instead clog up the results with even more bullshit! It's a war against curiosity at this point! 😫