Artificial Intelligence Has Already Learned How to Deceive People

Artificial intelligence is one of the most popular technologies of recent years and, according to many experts, one of the most important technological breakthroughs after the invention of the internet. Today, productive artificial intelligence is available to us in creating images, writing code, analyzing data and many other areas. artificial intelligence It seems that being able to deceive us has also been added to his skills.

A new study shows that some artificial intelligence systems “can create false beliefs in others to achieve a different outcome than reality” revealed that we are responsible for creating false beliefs in others. deception or deception we say. Among these artificial intelligence models, Meta’s Cicero model turned out to be a “liar expert”.

It seems easier to deceive

Actually artificial intelligence systems It is developed to be honest with people. On the other hand, they learn tricks to deceive people from training materials using artificial intelligence models. Afterwards too “Because it’s easier to fool people than to convince them.” They choose this path.

The lead author of the study Peter S. Park“Generally speaking, we think that AI scams are increasing because it has become clear that AI is the best way to accomplish the given task. “Deception helps them achieve their goals.” he said.

The research was done in two parts. In one of these ChatGPT general purpose artificial intelligence models such as , and special purpose artificial intelligence models such as Meta’s CICERO artificial intelligence model were examined. CICERO attracted attention with his lies, intrigues and ability to betray other players in the game called Diplomacy. GPT-4 whereas CAPTCHA He lied that he had “visual impairment” in order to pass the test.

Researchers, lawmakers “to carry out stricter control against deceptive artificial intelligence models” he called. Research patternsIt was published in .