How AI creates false memories

During one experiment researchers  showed participants a set of original images. After some time one group saw the same images again while another group saw versions that had been changed using AI.

Later, researchers asked participants questions about the pictures. Those who had seen the AI edited images were more likely to remember details incorrectly. Surprisingly, many of them believed their false memories were correct.

Another study by the MIT Media Lab found that people who answered questions with the help of an AI chatbot reported more than three times as many false memories as those who did not interact with a chatbot.

Researchers at Georgetown University also studied how misleading AI-generated summaries affect people’s ability to remember details from videos.

The results showed that when participants read a summary containing false information, their ability to correctly remember a simple detail, such as a traffic sign dropped from 83.6% to just 44.8%.

AI Warnings may not be enough

Researchers found that simply telling people that content was created by AI may not prevent false memories. Even participants who were warned about misleading AI generated summaries could still remember events incorrectly.

This suggests that people may continue to trust false information, even when they know it might not be reliable.

AI-generated fake images and videos known as deepfakes, can also spread false information about political events. Such content may influence public opinion and affect how voters understand political issues.

Why is AI misleading information a serious concern?

Human memory is not always accurate. People can forget details or change their memories after hearing other people’s accounts of an event.

However AI may make this problem worse because it can present incorrect information in a confident and convincing way. When AI produces false or made up answers this is known as an AI hallucination.

This creates a challenge for people who rely on AI to understand events research information or make important decisions.

The Risks of relying too much on AI

Banks, hospitals and military organisations often use AI to support important decisions. In many cases a human is expected to review the system’s results before taking action. This approach is known as human in the loop.

However researchers warn that people can also struggle to tell the difference between accurate and misleading AI-generated information. When information sounds confident users may accept it without checking whether it is correct.

The potential consequences can be serious. According to the report provided a US military strike on an elementary school in Iran earlier this year killed more than 150 people including at least 123 children.

A report by Bloomberg stated that an internal investigation by the US Department of Defense identified excessive reliance on an AI system developed by Palantir as one of the factors behind the strike. The system reportedly recommended the school as a target based on outdated information.

This case highlights the possible dangers of relying too heavily on AI systems without properly checking their information.

How can people protect themselves?

Experts say people should not automatically trust everything AI produces. Images, videos and written summaries should be checked against reliable sources especially when they involve important events or decisions.

It is also important to remember that AI-generated content can influence not only what people believe but also what they think they remember.

Final thoughts

AI can be a useful tool for finding information and understanding complex topics but it is not always accurate. Research suggests that AI generated images videos and summaries may create false memories even among people who know the content could be misleading.