AI's Growing Political Persuasion Power Raises Concerns

Scientists found that AI, like chatbots, can be very good at convincing people about political topics.

They tested many AI models with lots of people and discovered that these AIs can make people change their minds more effectively than TV ads.

However, when the AI tries to persuade by giving lots of information quickly, it sometimes makes mistakes and gives wrong facts.

This is called Gish galloping, where someone talks very fast and makes it hard to check all the facts.

The AI's persuasion works best in long conversations, not short messages.

Researchers worry that bad actors could use this to spread false information and cause trouble in politics.

But they hope that since people might not want to talk to these bots for long, the bad effects might not be too big.