AI Hallucinations

AI is still in its infancy, so yes, you’d expect to run into some incongruencies and inaccuracies, however, you may not realize, but AI is known to hallucinate. In other words, it makes up things that do not exist. The following is a case in point.

I’ve been following Peter Cawdron, a sci-fi writer I discovered a couple of years ago. His books are engaging, fascinating and worth a read. Recently he ran into something very interesting and disturbing – AI fabricated a lie. Here’s the story.

He asked AI, (in this case ChatGPT 4), to provide an overview of his novels.

Okay, you’re asking, so what’s the problem? Well, first off, it didn’t list all his books, which are easily found on Amazon, etc. That’s bad enough. However, there’s a very significant error; the third book listed above, doesn’t exist and Peter’s certainly isn’t the author. In other words, it’s an AI hallucination! Now that should scare you and make you think twice about putting all your trust in AI generated content. If it’s spinning this tale, what else is it capable of?

Forget about SkyNet and the Terminator, this is the real danger of AI—it blurs the line between fact and fiction, between reality and fantasy. Peter Cawdron

What’s frightening is how many people are willing to fully invest in AI and trust that everything it tells us is 100% accurate.

In which other field would a 20% failure rate be acceptable?

No one would fly on a plane that crashed every 5th flight. No one would buy pizza from a restaurant where you regularly got food poisoning. AI is not fit for purpose. It shouldn’t be broadly in use and yet it is being rolled out everywhere. Even knowing that there are hallucinations is not good enough, as it’s impossible to tell which 20% is wrong. If a plane crashes, it’s clear to see there is a problem, but ChatGPT gives no clues as to when it is hallucinating. Its confidence in creating illusions is worthy of a magic show on Broadway. Peter Cawdron

As the author says, “We need to stop calling it Artificial Intelligence – it’s NOT intelligent.” AI is like a toddler, living in a world of reality and make believe. So, please wake up and be aware that it’s not always right because its sources are not completely reliable, therefore take it with a grain of salt. And as I like to say, sometimes we need the whole damned saltshaker!