Published
Update
Reading time: 4 min
:quality(50)/2023/11/10/068-aa-07112023-1410540-654e6dafded18838321239.jpg)
The boss of A strategy which supports its “capacity to increase the visibility of relatively extremist dialogues”, according to Maxime Moffront, senior consultant at Rhapsodies Conseil and specialist in AI.
Does the Tesla boss have a problem with Google? In a first tweet on February 22, Elon Musk describes Google’s generative artificial intelligence, called Gemini, as “woke racist”. Cause: problems regarding the Gemini feature, criticized for historical inconsistencies in the images it creates. On social networks, users pointed out certain inaccurate creations, particularly in terms of gender and origin, appearing to underrepresent white people. A query on a German soldier from 1943, for example, resulted in images of Asian or black-skinned soldiers.
The next day, Elon Musk ensures on having spoken on the phone with a Google executive, claiming that the underrepresentation of white people and men is a “racial and sexist prejudices” of the tool.
If Google CEO Sundar Pichai denounced the errors “totally unacceptable” of its artificial intelligence application Gemini in a letter to its teams, and confirmed that “the efforts [de Gemini] to eliminate prejudices” are not to the point, Elon Musk’s comments give the impression that he rules all over the world on subjects related to artificial intelligence (AI). “Elon Musk is characterized by a strong involvement in these subjects but with very few concrete results”says Maxime Moffront, senior consultant at Rhapsodies Conseil and AI specialist.
“His statements are not really taken seriously enough to require a response” from the targeted companies, he adds. Moreover, Google’s attempt with Gemini did not overly worry Silicon Valley according to the AI specialist: the firm simply paused the generation of human images.
The controversy was mainly taken up by Internet users marked to the extreme right, in the United States as in France. “But this shows Elon Musk’s ability to increase the visibility of relatively extremist dialogues and with it, the significant increase in disinformation about X in particular,” he said.
Elon Musk, ambiguous on AI
Thus, according to the expert, the boss of SpaceX and Tesla is banking more and more on “a techno-solutionist discourse which relies on a relatively ‘right-wing’ and ‘extreme’ population of the media spectrum”. A strategy exacerbated by the new functioning of X since its acquisition by Elon Musk, “particularly regarding user remuneration on the Premium version”, recalls Maxime Moffront. Indeed, the more content is visible and generates interactions, the more profitable it is. However, it is violent, hateful or conspiratorial content that is shared the most and generates disinformation for which the X network is often singled out.
Furthermore, Elon Musk does not seem to be a serious player in generative artificial intelligence. “The Grok experiment was of little use”, recalls the Rhapsodies consultant. Indeed, the project of this “anti-chatGPT” whose trademark was responses tinged with sarcasm is in fact “a selling point for X Premium”, according to Maxime Moffront without being a real competitor to Meta, OpenIA and Mistral, the real behemoths of the market.
OpenIA was however founded by… Elon Musk himself, “as a research organization on artificial intelligence but it withdrew from the system in 2018, therefore before the major advances that we have the opportunity to see at the moment with GPT-4” he explains. The whimsical businessman had also called for a pause in the development of new artificial intelligence systems a few months before Grok’s release, illustrating his ambiguity on the subject.
How representative are generative AIs?
It not remain that “the ethical issues in AI are very important today”, explains Maxime Moffront. And as for Google, it seems to have to prove itself after previous scandals. Already in 2015, “Google was singled out when it released the Aria facial recognition and automatic photo tagging tool on certain phones.” An American developer denounced the racist bias present in the tool which identified black people as gorillas. A problem still not resolved almost ten years later according to Numerama.
“There was also the dismissal of a researcher, Timnit Gebru, in 2020 who was the ethics specialist for Google’s AI and who was fired to avoid the publication of an article on the flaws and risks of large language models (LLM) used by AI”, relates Maxime Moffront. “Several elements tended to show that Google had little interest in ethics in artificial intelligence. Google’s efforts today with Gemini therefore give a positive signal about Google’s efforts in terms of ethics”, he analyzes.
In addition, Google needs to regain a place of choice against Microsoft: “Today, despite Google being a huge player in artificial intelligence, it is no longer an opinion leader on the subject compared to Open AI and, to a lesser extent, Mistral, two entities supported by Microsoft”, explains Maxime Moffront.