No “artificial intelligence”: think tank describes “machine learning” etc.

No "artificial intelligence": think tank describes "machine learning" etc.

In an attempt to “uncover and mitigate the damage caused by digital technologies”, a well-known US think tank wants to refrain from using the English terms for “artificial intelligence”, “AI” and “machine learning”. The director of the “Centers on Privacy and Technology” at Georgetown University cites the important computer scientist Alan Turing and his concept of the Turing test as justification. Contrary to what is usually suggested, Turing was not at all concerned with the question of whether machines would one day be able to think. But that’s exactly what the terms suggested thanks to successful PR work. Emily Tucker doesn’t want the turning away from them to be understood as a new taboo, but as “creative practice for more intellectual discipline”.

Tucker recalls that in the text “Computing Machinery and Intelligence” Turing presented the “Imitation Game” as a mere theoretical sketch. In this game – the Turing test – a human should ask a computer and another human questions without knowing which was which. Based on the answers alone, he should then find out who was human and who was computer. In 1950, Turing predicted that thanks to the increase in computing power around the turn of the millennium, there would at most be a 70 percent chance of getting it right. However, he was not at all concerned with the question of whether the machine could then think in this case, but whether one could confuse a computer with a human being. The question, “Can machines think?” I didn’t attach any importance to Turing at all, nor was it synonymous.

While 2022 still has no indication that there might soon be a computer that would reliably be mistaken for a human, another of Turing’s prophecies has unfortunately been fulfilled, says Tucker. It is now common to ascribe to machines activities that are reserved for humans. This is how computers would “think”, “judge”, “predict”, “interpret”, “decide”, “recognize” and of course “learn”. The fact that our language thus represents machine intelligence as a given has no equivalent in technology. On the contrary, the development of both areas was deliberately decoupled from each other.

The PR departments of large tech companies are responsible for these linguistic changes. They would sell products whose novelty was not based on scientific breakthroughs, but on immense computing power with which gigantic amounts of data are processed, which can be collected thanks to the lack of regulation.

Instead of searching for the limits of computers’ potential in simulating humanity, “AI” vendors would pursue the limits of human potential for predictability. The term “artificial intelligence” now helps above all to “confuse, alienate and glorify”. That’s why the majority of people no longer know what “AI” is supposed to be; many would even think that “AI” is smarter than them.

It is no accident that we are so ignorant of technology and its growing importance in our lives. In science fiction, the “AI” is often a superintelligence that robs people of their ability to act. The danger is real, not because the technology is so good, but because “corporate greed and the perfection of political control require that people give up the pursuit of knowledge about the possibilities of their own minds”.

In any case, Turing did not dream of a society in which computers, which could not be friends with anyone, would be treated as authorities. The question of what it means for people to be able to think is a liberating one. The conscious use of language is an important part of it. Tucker’s think tank therefore wants to write instead of “face recognition based on artificial intelligence” from now on, “Tech companies use massive data sets to teach algorithms to assign pictures of faces to each other”.

In addition, possible difficulties in using the technology and the providers should be mentioned. Finally, it should be made clear that people acted. “Machine training” could then be used instead of “machine learning”. This is not a creed or a new taboo. The distinction between the real and the imitation should be the focus.


To home page

#artificial #intelligence #tank #describes #machine #learning

Leave a Comment

Your email address will not be published.