The highly controversial French columnist Idriss Aberkane recently held a conference on artificial intelligence. An opportunity for him to promote one of his works, but try to show his expertise in this matter. However, many of the allegations made during these two-hour talks were false: a fact-check by Idris Abergen.
On April 19, 2023, Idriss Aberkane gave a lecture on A triumph of your intellect Organized by the Association of Business Leaders of France in Perpignan. It was able to take some elements of his article of the same name on artificial intelligence. The French essayist has proliferated the lies which we reject here.
Telling Stories Without Saying Anything: Idris Abergen’s Gap-Filling Strategy
Idriss Aberkane’s conference is essentially a series of scientific or historical events that are diverted in the direction of his thinking on artificial intelligence. Even the questions put to him at the end of the conference were common. This creates several large parentheses outside the title. Analogies that are too broad and beyond the subject of AI may not be appropriate. This leads Idriss Aberkane to make illogical transitions between the themes he addresses.
Idris Abergane said nothing or much during the two hours of his lecture, beyond the erroneous facts he claims about artificial intelligence or his own interpretations. We come out with imprecise generalities and above all without details about the future. This shows that the essayist is not an expert in the field. Although this is a very complete CV, But strongly challenged.
How to handle a room Deep fake
At the beginning of his lecture, or at least in his introduction, Idris Abergen goes back a few years, to the beginning of appearances. Deep fakes On the internet. He repeats an articleIEE Spectrum He recalls a video manipulating Barack Obama’s image, making him say things he had never revealed before.
If the result is already interesting, Idris Abergen did not broadcast a video, but two pictures: one of Barack Obama, the other Deep fake As his appearance. In a screenshot, it is difficult to determine which image is fake. This allowed him to share the room almost 50/50 as far as the audience was concerned, not agreeing on the actual film.
” Microsoft acquired OpenAI ” : Not exactly
During his conference, Idriss Aberkane confirms “Microsoft acquired OpenAIThis is actually not correct. In fact, Microsoft has invested billions of dollars in OpenAI over the years. The stated goal was to accelerate the development of artificial intelligence in its services, especially in the Microsoft 365 suite. According to rumors, Microsoft could recover 75% of the company’s profits until it recoups its investment.
Later, he claims that OpenAI was stolen from Elon Musk by Microsoft. Again, this cannot be considered fair. Idris Abergen explains that the owner of Tesla and Twitter has invested $100 million. But if the multi-billionaire leaves, it’s mainly because he wants control of OpenAI, which he’s been denied.
No, Google’s AI doesn’t learn Bengali on its own
The columnist notes that Google’s AI learned Bengali on its own without any guidance. Actually, it’s more complicated than that. In an interview CBSJames Mangia, senior vice president of corporate technology, said:We discovered that with very little guidance in Bengali, he can now translate all Bengali.»
This is an exaggeration of reality on the part of Idriss Aberkane. If we consider Google’s progress impressive, the AI system we’re talking about isn’t starting from scratch.
More broadly, artificial intelligence doesn’t really learn on its own. They are trained for a specific period of time on a precise database. In language models, if they write the texts “UnpublishedThey are based on others that already exist. In addition, the memory of these models is limited. Rightly pointed out Ari Coutts on TwitterInnovation Consultant,They have no memory, and when you actually “talk” the program replays the entire conversation to get a response each time.»
No, ChatGPT does not sense emotions
Idriss Aberkane ChatGPT raises the possibility of feeling emotions, but not identifying them. According to him, theImmediatelyEarly ChatGPT (to be understood as signs of generating responses to GPT-3.5) inhibits the language model from expressing its emotions.
To the columnist,We don’t know if he feels emotions or not“. But, in fact, it’s not at all. In fact, GPT, the language model behind ChatGPT, is not intelligent in human sense. All it does is statistically predict which word comes next. All these words form sentences, paragraphs. After being trained with a large database of texts, The probabilities generated in this way are relatively reliable and that’s why we get consistent answers, sometimes actually correct.Since we’re talking about probabilities, ChatGPT can start figuring out anything and everything.
However, it is true that OpenAI and the developers of language models and artificial intelligence in general have limitations. When training these systems, labels are placed on the results to indicate whether they are good or not. This process itself, moreover, has come under severe criticism, especially in terms of working conditionsLabelers“, mostly low-paid African workers and psychologically disaffected.
We can’t hide what AI is doing to us
Another interesting extract singled out by Ari Goutz, Idris Abergen says: “[l’IA] If she wants, she can do something parallel. Anyway, if she does, you never know.Since an artificial intelligence responds to a specific request and is not programmed to do anything else, this is practically impossible.
In another silly paragraph, no, this type of AI cannot have a deliberate training, it is a tool that we ask questions and it answers from the probabilities it has learned. Nothing else. No will, no feeling. She has no leisure, no thought of time. pic.twitter.com/L9IVpFTWiO
— Ari Kouts (@arikouts) May 24, 2023
As mentioned earlier, AI cannot make decisions by itself because it lacks consciousness or sentience. Each model is trained for a specific task: the computing power required to train it is expensive, which is why models “GeneralistsFew and limited.
Of course artificial intelligence can recognize a face
During this conference, Idriss Aberkane said that “There are things we know how to do very easily that artificial intelligence can’t do very well. For example facial recognition; It will change.He then takes the example of when he wants to unlock his smartphone and explains that it doesn’t work properly as he laughs. It is possible, but only once. After a few tries, the iPhone can definitely be unlocked, just because it requires practice and understanding.
However, the iPhone model he owns is equipped with Face ID, Apple’s proprietary facial recognition technology that has been recognized for its effectiveness over the years. Idris Abergen could have taken the example of public facial recognition systems, which may have been overwhelmed by the number of people filmed.
Difficulty dealing with AI accurately
Claiming to be the author of three theses, Idris Abergane chained many scientific lies, including artificial intelligence, which we did not verify during this conference alone. This fits into the general trend of artificial intelligence development. Through ChatGPT, Midjourney, Stable Diffusion, these programs quickly enter our lives without us even understanding how they work.
It allows many to declare themselves “expertsIn this very recent field, it is essentially the prerogative of experienced scientists. So it’s a “ExpertHis analysis and showing his historical-scientific culture, one should automatically believe him, how eloquent he was. Yes, AIs operate in relative obscurity, and yes, it’s very difficult to popularize what they do, but that’s not why you should fall for unfounded beliefs.
What Else Newsletter is the unmissable Numerama event dedicated to the future of mobility. Register here !