A study shows that AI is not so smart after all: this is where it always fails

How intelligent is AI? This is a question that many people probably ask themselves today. Not in vain, this technology has more and more weight within society, and we can’t stop talking about it. There are promises around it that range from a future without the need to work to immortality, among other incredible phenomena.

There are also those who think the opposite, of course. Some time ago an expert argued that artificial intelligence is no smarter than a kitten. Well, a new study is precisely more along these lines. It doesn’t go into too much about how far AI will go or not, but it delves into its own nature. No, she wouldn’t be that intelligent, and the idea that many have of her is wrong.

Don’t fool yourself, artificial intelligence doesn’t think

It has been an investigation shown on the technological portal TechSpot which has cast a bit of skepticism around artificial intelligence. Generally, many associate this technology with science fiction, Terminator either Matrixand amazing machines that reason as well as or better than humans. But it would be something very far from reality, according to specialists.

And why? Well, for one fundamental reason: that no one be fooled by its name, artificial intelligence does not think. Not long ago, Stephen King said he was calm about the possibility of losing his job. That they didn’t see real ingenuity or creativity in generic AI texts. And perhaps the well-known writer found precisely the key to the matter.

As discussed in the research carried out, language models such as the most popular of all, ChatGPT They work through textual patterns, associating words, and not really understanding how the world worksfrom a conventional point of view. That is, they do not reason. Hence their “daydreams” or the blunders they usually make.

Is this really the case with all models of this technology that currently exist? In the opinion of these researchers, yes. Therefore, although many jobs seem to be at risk because of AI, perhaps we should be cautious about this. Especially if what you are looking for are genuine results and, so to speak, a true understanding of a problem.

Problems understanding the world

To reach these conclusions, MIT scientists did different tests, for which they were based on different variable situations (and that is the key to the matter, situations dynamic). A good example to understand it: traffic in New York. During the tests, The AI ​​indicated the city routes perfectly, but when something changed, it started to get confused.

Whether it was because it was simulated that there was an accident, restrictions or any phenomenon that introduced the need for a change: that is where artificial intelligence really began to have problems. And what does this prove? Well, AI, although it seems intelligent, is not. Something that many people, and especially companies, should know before betting all their cards on it.

Get to know how we work in ComputerToday.

Tags: Artificial intelligence