Usually when there’s something we need to know, we go looking for an expert. These days, that expert is often on-line, and we often find him or her via Google, Wikipedia, or YouTube. There are a lot of experts on-line these days, and most of us have learned to be a bit cautious before deciding who we’re going to listen.

There’s an old maxim that says an expert is just an idiot from out of town. He’s supposed to know his business, and you don’t know anything about him, and you aren’t an expert in his field, so you assume he’s got the real goods. (This is probably a corollary of the Gell-Mann Amnesia Effect.)

But does he? Maybe not.

There’s a whole flock of new experts in town, these days. ChatGPT is the most famous, but there are a bunch of them based on the same Large Language Model (LLM) technology. Folks are expecting great things of these programs, and they are getting built into all manner of applications and websites.

But is ChatGPT actually an expert, or just an idiot from out of town? It’s called an AI, an Artificial Intelligence, and it’s pretty amazing; but it isn’t actually intelligent. It has no judgement; and as many folks have seen, it’s quite willing to make stuff up when it doesn’t know the answer.

And where does it get its answers? Well, from the data it’s been trained on. And where did that data come from? The Internet.

The Internet.

Think about that.

There are many fine pearls on the Internet, but those pearls are surrounded by cheap costume jewelry, all of which is embedded in hearsay, nonsense, and manure. How well did ChatGPT’s trainers do at weeding out the trash? We don’t know. And how did their biases affect their choices? We don’t know.

I’m not saying that apps like ChatGPT are without value. I expect that the day will come that they make fine research assistants—provided you’re willing to add the human judgement to their search results.

But experts? No more so than any other opinionated lore-master on the Internet.

____

Photo by Joshua Sortino on Unsplash