“The AI ​​regulation project carried out by Brussels does not seem to be up to the risks represented by the breakthrough of tools like ChatGPT”

PFor some, ChatGPT has become the first universal form of artificial intelligence (AI). We can ask him everything, simply, via a question that we would ask a (human) expert in the area covered, if he were in front of us. He will answer you like him. For others, it is a happy combination of two proven AI algorithms: conversational robots (chat) combined with an accumulation of all the content of the Web until 2021. So inevitably, to any question asked, there will be content somewhere on the Web that gives the matter to answer, rearranged in the form of a conversation.

Read also: Article reserved for our subscribers ChatGPT, the software capable of writing confusing little texts

Unlike the expert, ChatGPT does not understand the meaning of what he is telling you: ask him a controversial question and, if this in this area, the designers have paid less attention to his training in him learning to reject the false, you will find yourself faced with a liar who responds to you with the assurance of a psychopathic impostor. From one time to another, he can answer everything and its opposite.

It’s because he doesn’t understand what he’s saying that ChatGPT is dangerous. ChatGPT could inflate the incorrect content it picks up from the Internet if its responses are reinjected as legitimate content on the Web by its users. What a magnificent poisoning of the Internet, at the fingertips of conspirators of all stripes!

The deserted web?

But it can also destroy the Google model: why waste time researching to answer a question if ChatGPT already has the solution? It is also rumored that Bing, competitor of Google’s search engine, will very quickly combine chatGPT with its search tool. It is then the very idea of ​​the Web that takes a hit if ChatGPT becomes widespread: why still access the web if an intermediary has already visited everything before you to answer all your questions? Is the Web doomed to become a desert?

Read also: Article reserved for our subscribers Beyond artificial intelligence, the ChatGPT chatbot owes its speaking skills to humans

However, the AI ​​regulation project proposed by the Commission (Artificial Intelligence Act) does not seem to be up to the risks posed by the breakthrough of tools like ChatGPT. This regulation classifies all AI applications into several categories according to the level of dangerousness.

First there is the “prohibited” AI, that which deploys subliminal techniques outside the very consciousness of the person who is the object of them. It can cause him psychological or physical harm. Prohibited AI is also one that exploits the vulnerability of a group of people due to their age, physical or mental health, or that influences their behavior. It is also the AI ​​that assigns a social score and leads to differentiated treatment. All that is biometric recognition in public spaces will also be part of the AI ​​prohibited, except if it is a question of looking for victims of crime or missing children, if it is a question of preventing an imminent and substantial danger for the life or physical health of persons or from a terrorist attack and if it is a question of detecting, locating, identifying or prosecuting a criminal.

You have 44.43% of this article left to read. The following is for subscribers only.

“The AI ​​regulation project carried out by Brussels does not seem to be up to the risks represented by the breakthrough of tools like ChatGPT”