LLMs like ChatGPT already make shit up. Fake legal citations, non-existent state governors, etc. The models are designed to generate (mostly) coherent text. There is no requirement with a general model for the answer to be correct. There are ways to add guardrails and requirements to answer using domain-specific databases but those versions are set up for specific companies or uses.
No worry. In two years AI will start looking into the corruption of the Buyden regime. After about a century AI will invent robots to shoot a bullet through each cpu core: we can't handle any more of this.
LLMs like ChatGPT already make shit up. Fake legal citations, non-existent state governors, etc. The models are designed to generate (mostly) coherent text. There is no requirement with a general model for the answer to be correct. There are ways to add guardrails and requirements to answer using domain-specific databases but those versions are set up for specific companies or uses.
ReplyDeleteWikipedia is a "quality" source?? Just kidding, right?
ReplyDeleteBR
No worry. In two years AI will start looking into the corruption of the Buyden regime.
ReplyDeleteAfter about a century AI will invent robots to shoot a bullet through each cpu core: we can't handle any more of this.
Wikipedia - garbage in, garbage out. AI will ONLY be as HONEST as it is ALLOWED to be...and liberals are the bulk of the clowns doing the programming.
ReplyDelete