AI Language: How Trained Relationships Shape "Truth

See, the AI is trained on the mass of human language and the interrelationships between words. As they're used in the context of sentences and paragraphs. And along the way, there can be strengthened relationships between words that may not be factually true as we recognize them. So, for example, an actor, an actor's name, say like Brad Pitt, and then a year, like 1973. And for whatever reason, the year 1973 might show up a lot around Brad Pitt or some other year, and then when you ask the engine, what's Brad Pitt's uh birthday, it might conflate the 1973 with Brad, Brad Pitt because statistically those have been more related in the learning. Right. Even though I don't know what year he was born. But probably not 1973. I don't know. Who knows? I think it's more like 1968 or something. Yeah. Anyway, um that that's what hallucinations are. So, remember that an AI engine is not actually doing fact checking by default. You have to you have to tell it to do fact checking, which is what I do. You know, like if you use my engine, bright answers.ai. It actually has the reason it takes so ...