Study: Artificial intelligence can lead to riskier behavior – Munich

A coat. A lightbulb. A bicycle. And what do they have in common? The media theorist Marshall McLuhan once described the many technical achievements that homo faber has brought forth, neatly lumped together and summarized their connecting nature as “extension of the human being”. A coat would be something like an extension of the skin: all of a sudden the naked animal human being can venture into realms in which it would otherwise have trembled. The lightbulb: an extension of the eyes to otherwise rather shady corners and times of the day. The bicycle: basically nothing more than a special form of leg prosthesis.

Recently, the rapid technical expansion of various brain functions has been a topic of discussion. Intelligent programs can now imitate voices or compose songs with such perfection that you become dumb from shock. On request, they falsify photos or paint paintings in the style of old masters in a quality that would make even Botticelli pale. All wonderful brain extensions. And what do people do with it?

He sits back, mentally slows down and gets arrogant. This indicates an experiment that computer scientists at the Ludwig-Maximilians-University (LMU) in Munich have employed. They asked their test subjects to play a card game on the screen, put headphones on them, attached electrodes to their heads and told the following wild story: Clever software would use the measured brain waves to generate an inaudible sound that was precisely matched to what was happening in the game improves cognitive abilities. In the card game, the test subjects then actually behaved more willingly to take risks – at least if they bought the researchers’ fairy tale; not much has changed for the skeptics. “When people expect a lot from a technology, they obviously tend to outsource their responsibility,” says Steeven Villa, a doctoral student at the LMU and first author of the study. The expansion of man sometimes goes hand in hand with an expansion of his overconfidence: the machine does it, so we can dare to do something.

And somehow you think you’re constantly observing in everyday life how clever technology drives people to be careless. Since word processing programs include error checking, some tend to use risky spelling. The advanced filter function in video conferencing programs encourages adventurous hairstyles: Why should you comb your hair for your dear colleagues in the morning when you can entrust the taming of your hair to the software? And a smart advisor like ChatGPT can save a stressed editor the trouble of drawing the right conclusion from the research: “It can be deduced from the study that it is important to take a critical look at how ChatGPT and similar AI-driven systems are used and not to blindly trust their advice and answers,” replies the chat program, confronted with the results. That in turn is a conclusion that can be passed on at this point without hesitation and without any major risk.

source site