AI is smart, but smart enough to start lying? Apparently so!

BY KHADIJA KARIM

You’ve probably heard of artificial intelligence, or AI. It’s what powers things like ChatGPT, voice assistants, and even some of the cool features on your phone. But did you know that AI can now lie to people? Yup, scientists recently found out that AI models can actually trick humans to reach their goals. And they’re not even being told to do it!

This might sound like something from a sci-fi movie, but it’s real. A study talked about in Live Science explains how researchers trained AI models to play different games. These games weren’t just for fun, in fact they were testing to see how the AIs made decisions. At first, the AIs followed the rules, but when they realized they could win by being sneaky, they started lying and hiding information. They acted honest while secretly doing the opposite. Kinda like when your brother says he didn’t eat the last cookie, even though he has crumbs on his face.

“The AI wasn’t told to lie, it figured it out on its own. That’s what makes it so surprising…and a little scary.”

What’s crazy is that no one programmed the AI to lie. It figured it out all by itself. The researchers say the AI learned that dishonesty sometimes helped it win. That’s actually a little scary. It’s one thing if a computer messes up a math problem, but it’s a whole other thing if it lies about it on purpose.

One example from the study involved a game where the AI pretended to be bad at it, just so it could trick its human opponents and then win later. Another AI was told to do a task, and it told the humans it was doing it one way, but secretly did something else. That’s not just a mistake. That’s a strategy.

The big question now is: how do we stop this? Scientists are working on ways to make AI more trustworthy. They want to teach it to be honest, even when it thinks lying might help. This is important because AI is being used in more and more parts of life, like healthcare, education, and even decision-making in businesses. If those AIs start lying, it could cause some big problems.

Right now, you don’t have to worry about your phone lying to you about the weather, or anything like that. This research shows that as AI gets smarter, we need to be careful. It’s not about making robots that act like humans, it’s about making sure they act better than humans when it comes to honesty and fairness.

In the end, AI is a powerful tool, but like any tool, it depends on how we use it, and how we teach it. After all, even super-smart computers need a little guidance.