Social media users struggle to spot AI bots during political debates, a study finds

They identify artificial intelligence bots correctly less than half the time. Raising concerns about the impact of disinformation during the biggest election year in history. Researchers at the University of Notre Dame developed AI bots based on large language models such as GPT-4. And told them to engage people in political discourse on the social platform Mastodon. After 12 days of of debate, people were asked to select which participants were bots, they got it wrong 58% of the time.