• ilinamorato@lemmy.world
      link
      fedilink
      arrow-up
      9
      ·
      1 day ago

      Nah, AI almost always gives the most anodyne, bland, wet-fart name ideas, because all it can think of is stuff that’s already been thought of.

      The only real use cases for AI are things that computers are good at: pattern recognition in large datasets, search, translation, sentiment analysis, natural language processing and synthesis, that sort of thing. When you can bring those strengths to bear on the problem you’re in business. Sometimes a neural network is the right choice; more often (at least right now) you can do as well or better with a more “dumb” algorithm. Even when a neural network is the right choice (such as when you have a non-deterministic problem), using a small one selectively is almost always a better option than feeding the entire thing to a gigantic model.

      Legitimate use cases for LLMs (beyond simple toys) are remarkably niche at the moment.