3 ms·

Hmm yeah I kinda like the concept that it's "hallucinating" 100% of the time, and it just so happens that x% of those hallucinations accurately describe the rea

• empath75 2y ago
  That x% is far higher than people think it is because there's a tremendous amount of information about the world that ai models need to "understand" that people just kind of take for granted and don't even think about. A couple of years ago, AI's routinely got "the basics" wrong, but now so often get most things right that people don't even think it's worth commenting on that they do. In any case, human consciousness is also a hallucination.
  • notnullorvoid 2y ago
    It really depends on the set of prompts you present the LLM. If it's anything requiring reasoning, you'll often get nonsense that sounds like sense. It has a higher chance of being accurate with knowledge queries. LLMs are impressive, a very lossy search engine in a small package, capable of outputting convincing natural language responses.