The news this week reminded me of a discussion I had with Richard Stallman. He was vehemently opposed to calling it AI hallucinations, as that would imply human-like thinking. He was more in favor of calling it "computer-generated bullsh*t."
#AI #hallucination #slop
Discussion
Loading...
Post
@kdkorte I've been pointing this out since the beginning. All the anthropomorphizing language is a problem because it prevents people thinking objectively about this technology and distorts how they relate to it in use, in dangerous ways.
It is big warning to everyone about how bad we are at properly framing something that is designed to appear like us, and is therefore very good at deceiving us, and is controlled by people whose whole way of being is based on manipulation of others.
@happyborg I'd say it's a language problem in general. On the one hand, you want to make it understandable to someone who isn't part of a tech bubble, thus you cannot just invent words. On the other hand, adopting allegories and metaphors risks endowing something with more characteristics and meanings.
@kdkorte It is easy and clearer to say an LLM model produces errors. Using the word hallucination is unnecessary, misleading and inaccurate.