Spread the love

Why AIs sometimes write mind-blowing nonsense ?

© Unsplash/Adi Nugroho

You've probably already felt an uneasy feeling while interacting with ChatGPT or one of its competitors. The AI ​​is indeed capable of uttering something stupid or lying by using a peremptory tone, which can sometimes be confusing.

In an article published in Nature Human Behavior, Anthony Chemero, professor of philosophy and psychology at the University of Cincinnati, was interested in this phenomenon that the developers call hallucination.

The AI ​​is totally disembodied

The researcher begins by recalling that these AIs are trained from massive quantities of data from the Internet. Often, the latter are imbued with the prejudices and views of the people who wrote them.

In a press release published for the occasion, the scientist then goes into detail:

LLMs generate impressive texts, but they often make things up from scratch. They learn to produce grammatical sentences, but need much, much more training than humans. They don't really know what the things they say mean.

In some cases, you shouldn't even too much effort on the user to get them to say “nasty things that are racist, sexist, and otherwise biased”, affirms the author.

In reality, we must also get away from the word intelligence, because these machines are in no way intelligent. Humans are embodied beings who are surrounded by other people with material and cultural environments. “This’is why we care about our own survival and the world we live in”, notes the expert .

ChatGPT and its competitors, on the contrary, have no survival instinct or inhibition. They in no way understand what they are doing or the issues, but just try to string sentences and words together in a logical way.

With this approach, we now understand better the numerous hallucinations to which these language models can be subject. AIs improve over time, but it seems unlikely that they will get rid of this nasty flaw for good.

As a reminder, filmmaker James Cameron had felt these limits well. During an interview, he explained that he was very skeptical about the use of text to write scenarios: “I don't believe that a disembodied spirit that regurgitates what other incarnate spirits have said – about the life they had, about love, about lies, about fear, about mortality – and who puts it all together in a word salad and then regurgitates it. I don't imagine that this is something that could move the public”.

The director even joked that if in 20 years an AI wins the Oscar for best screenplay, he will have to bow.

📍 To not miss any news from Presse-citron, follow us on Google News and WhatsApp.

[ ]

Teilor Stone

By Teilor Stone

Teilor Stone has been a reporter on the news desk since 2013. Before that she wrote about young adolescence and family dynamics for Styles and was the legal affairs correspondent for the Metro desk. Before joining Thesaxon , Teilor Stone worked as a staff writer at the Village Voice and a freelancer for Newsday, The Wall Street Journal, GQ and Mirabella. To get in touch, contact me through my teilor@nizhtimes.com 1-800-268-7116