MIT Study Shows AI Reliance Degrades Human Memory and Critical Thinking

People who use generative AI tools change the way they remember things in ways that are difficult to reverse, a group of scientists from MIT Media Lab found. This is important because it disrupts the narrative that AI-assistance is all good things and implies that constant use of AI tools may impair our ability to remember things correctly. In other words, the convenience of AI is coming with an invisible cognitive cost that impacts the way we interact with and remember reality.

Research reveals AI tools implant false memories and weaken neural connectivity

So in one study, subjects were presented with images that were digitally manipulated with AI (e.g. add some soldiers to a scene or change a face expression). If the subjects were shown these manipulated images, they were much more likely to confidently recall non-events. This shows that the AI-generated stuff can implant fake memories. This overwrites our human visual memory, despite the fact that they're specifically told that these are AI-generated.

Secondly, the researchers found that using AI tends to reduce the connectivity of your neurons, compared with the when you do things manually like write things or search the web. Thirdly, in one clinical observation, the doctors using AI to diagnose have their ability to detect cancer independently reduced by 20%, if it's a common practice. This is due to something called "cognitive offloading", where the human critical thinking and ability to remember stuff gets degraded because we offload our mind to some algorithms. The team warns us that the most likely danger is not AI taking over the world, but the degradation of human abilities in general.

'I think this is all very dangerous,' Pattie Maes, one of the researchers, said. The study is in early days. It doesn't conclusively show that the loss of memory is permanent for regular users. But it's a trend.