Artificial Stupidity, a threat to history?

You might not know much about Chester A Arthur, one of the more obscure presidents of the USA. But if you are a history nut, or you even read a long article about the guy, then you probably know more than the AI does.

A lot of articles about Artificial Intelligence talk about how computer intelligence is a threat to writers (which may include historians). After a test drive of one popular engine, I do not feel the least bit intimidated by the engine’s “intelligence.” The bigger threat might be its popularity, and the possible impact it has on finding the facts.

This is how my journey started. We had a day stuck at home, so I thought it was a perfect opportunity to teach the next generation about Romanian history and current affairs. (We also cover American, British, Indonesian, French, and other histories as the opportunity arises).

Previously, my attempt to teach Napoleon was a bigger failure than the battle of Waterloo. Sure, I can make him funny, but somehow talking about bloody guillotines, angry revolutionaries, and fights in the schoolyard are not the best subjects when talking to impressionable young minds. Some kids like to play it out, while others just don’t like it.

So, I decided to start with people who have less violent reputations. And, we started with a primary source, the newspaper.

Let us find interesting facts about the mayor in the newspaper, I suggested. We found a few instances of him, but not many.

That is when I decided to use Youtube for research. The mayor’s YouTube page was pretty boring, but it had regular, recent, updates. Even at double speed in a language that I struggle to understand, it still seemed too slow. Those updates on your friend’s Facebook page that you like because it says that your friend is still happy and still alive, but about the mayor. Two hours later, I reached the conclusion that the mayor is alive, even if he doesn’t seem to have an interesting life. It is times like this when you notice along the side what the suggested videos are.

There was a suggestion of an AI, but I didn’t know that when I clicked on it. It looked like a tool that I might be able to use for all my projects. As “school” was over, I decided to give it a try.

The AI tool in question could not only have conversations with customers on your site’s webpage (crawling your Frequently Asked Questions or whatever), it could even write essays by crawling the Internet. While it gathered facts, it passed the plagiarism tests. It sounded dangerous, a tool that allowed students to easily cheat on homework and exams.

Some of the essays that the AI “wrote” in the video were good essays about American history. High school level, mind you, it didn’t include any references, and it wasn’t about secure subjects. It was the basic kind of thing that everyone knows, or is supposed to know, like a little story about the signing of the declaration of independence. The facts on these topics looked correct, it was easy to understand, and the AI could get an “A” at many high schools.

So, naturally, I decided to give it a go, to “cheat” on research. There were a few characters from history, as well as our day, who I just couldn’t find much out about. The mayor and Chester A Arthur are two that I had found out about, but I couldn’t remember the exact details.

First, I tried the mayor. I discovered that the poor man died years ago, at a relatively young age. It must be tough to be mayor when you are one of the undead. No wonder he struggles on YouTube.

Then, I came to Chester A Arthur. I asked for a biography, something I could use. Most of what was typed, I already knew. There was the name of the civil service act that was passed during his presidency, and a couple of details about that act that I had not yet memorized. I don’t really memorize birth and death dates, years of Presidency, or the presidents in order. So, it produced a few details there that might be useful if they were true.

But, it seemed to leave a few details out, like about Chester A Arthur’s family.

I asked for details. The AI claimed that he was never married.

Now, Chester’s wife is not incredibly important to the standard way of writing history, but that is an easy-to-find historical fact. (And I would argue that she is important, but that is another story).

I tried a few Romanian politicians who I don’t know much about but I wish I hadn’t. The names of their political parties, even when translated into English, just seemed wrong. Realizing how many false facts it was giving me about the mayor and Chester A Arthur, I tried to stop reading, to block the “facts” from my mind until I could verify them.

But the AI stories were so well told, that I probably learned a few things that I will soon have to unlearn. I need to keep a note about who I researched, so I can find reliable sources that will correct any of the false facts that may be resting in my subconscious.

The problem here is not just a few people learning false history, however. It is that AI is being used to police false information. If the AI is confusing false information with the truth, it may let false information through its net. But that is not the worst of it.

If the AI ends up blocking the truth, the real facts, because they do not agree with the narrative it has somehow assembled, then we end up with a weakened sense of our history. People like Chester A Arthur’s wife get erased from history, you might say so what, people are declared dead ahead of their time, and entire civilizations might be forgotten.

Imagine if you get refused a bank account or rejected from a job because the AI thinks you are dead? Imagine if you are denied citizenship because the AI decides that you and your parents are not related?

I have had a few problems with bank accounts and job applications that I put down to problems with regulations and Artificial Intelligence.

Historians can keep watch on how the AI treats lesser-known stories to see how ordinary people will be affected. If your local mayor or Chester A Arthur is misrepresented by the AI, how much more will it err when researching private individuals?

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *