As we’ve previously reported, artificial intelligences tend to make up many more things than people believe. Although the most famous cases are those of Microsoft‘s new Bing and Google‘s Bard AI, in which they could shamelessly lie to users and even gaslight them, the popular ChatGPT is no exception.
And the worst part is that, apparently, it’s a problem that will never end. Associated Press reports in a recent article the opinions of a group of experts in artificial intelligence who claim that the “hallucinations” of language models would not be an easy problem to solve.
“This is irreparable,” says Emily Bender, a linguistics professor at the Computational Linguistics Laboratory at the University of Washington. “It is inherent in the mismatch between the technology and the proposed use cases.”
Indeed, not everyone necessarily sees it as something negative. “In fact, hallucinations are an added advantage,” says Shane Orlick, president of Jasper AI. “We have clients all the time who tell us how ideas came to them: how Jasper created storylines or approaches that they would never have thought of themselves.”
The same can be seen in image-generating AIs, capable of creating images with very “peculiar” aspects, especially when generating the complete image of a human being. However, while it is true that, creatively speaking, AI can be a plus, their “inventiveness” can fuel the flame of misinformation.
“The large language models are designed to come up with things. It’s the only thing they do,” said Bender. “Even if they can be adjusted to get things right most of the time, they will still have flaws, and those flaws are likely to occur in the cases where it’s harder for a person reading the text to realize, because they are more obscure.”
Some of the links added in the article are part of affiliate campaigns and may represent benefits for Softonic.
