2023-05-17 17:36:00
We’ve had more turbulent tech audiences. Far from the atmosphere of discord that prevailed during the last convocations of American bosses of social networks, Sam Altman’s first major oral presentation before the United States Congress took place without major hitches, in a courteous and even concord atmosphere. .
The leader of OpenAI, creator of the now famous ChatGPT, was auditioned on Tuesday, along with two other experts, Christina Montgomery, in charge of privacy protection at IBM, and Gary Marcus, expert in artificial intelligence, by members of the Senate Legal Affairs Committee, including elected Democrats and Republicans. On the program of the debates, in particular, the need to regulate generative artificial intelligence, which has swept the world through image creation software like Dall-E and Midjourney, but also, of course, chatbots like ChatGPT.
A consensus around the imperative to regulate
If the debates took place in a peaceful atmosphere, it is above all because everyone seemed to agree, Sam Altman the first, on the need to quickly put in place demanding regulations to oversee the use of algorithms. artificial intelligence generator.
« OpenAI strives both to ensure the democratization of artificial intelligence and to maximize the security of these models. However, we also believe that government intervention is necessary to limit the risks posed by this increasingly powerful technology. »,
He thus immediately attacked the entrepreneur who, like his colleagues in Silicon Valley previously grilled by Congress, had for the occasion swapped his jeans and T-shirt for a suit and tie. .
Invited to expand on the type of regulations that he believes should be put in place, Sam Altman then, seconded by Gary Marcus, notably proposed the establishment of an independent federal agency responsible for granting the necessary licenses to in the market for artificial intelligence algorithms, with the ability to revoke these licenses at any time in the event of non-compliance with a certain number of criteria, on the model of what the Food & Drug Administration does for example in the States -United.
« We are witnessing something historic: it is undoubtedly the first time in history that the leaders of an industry have come before the legislator to beg him to further regulate their market. This is all the more exceptional in a country where the idea that the less the government interferes with an economic sector, the more it can prosper, has long been very present. noted Richard Durbin, the Democratic senator from Illinois.
A desire for regulations with which elected officials, including Republicans, might only agree. The fear of repeating the same mistake that was made with social networks, where the laissez-faire approach ended up giving Cambridge Analytica, came up once more and once more. ” Congress failed to act at the right time on social media. Now we have a duty to do it for artificial intelligence before the threat becomes too real summed up Richard Blumenthal, Democratic senator from Connecticut and chairman of the commission. European voluntarism on the subject has also been repeatedly held up as a model.
Why OpenAI has an interest in more regulations
But is there really anything to be surprised by Sam Altman’s insistence? After all, generative artificial intelligence has taken the world by surprise with its impressive capabilities. The risks of seeing this technology disseminate false information boosted to carry out mass social engineering, quickly put a large part of the population out of work, or even cause an explosion of cyberattacks, all risks that were mentioned during the exchanges , are already evident to a large part of the population. In such a context, it is difficult to imagine the father of ChatGPT appearing before Congress saying that everything is fine in the best of all possible worlds and that it is perfectly useless to regulate anything.
However, for the boss of OpenAI and his colleagues in Silicon Valley, the regulations are not only a necessary evil, but also an economic opportunity. Indeed, as Gary Marcus has argued,
« TEveryone has an interest in the rapid implementation of standardized regulations on a global scale, under the impetus of the United States, including artificial intelligence companies. In the absence of such regulations, it is indeed likely that each country, or even each region, will tinker with its own rules in its own corner, which would force artificial intelligence companies tomorrow to train an infinity of different models to stick to the specificities of each local law. A nightmare. »
Finally, with the US presidential election due to take place next year, a Cambridge Analytica-type scandal, but involving generative artificial intelligence, might prove catastrophic for the industry, tarnishing its image with the public in the long term. and regulators, and auguring much less cordial future summonses to Washington for Sam Altman. The implementation of rules capable of limiting the risks in this area are therefore once more in the interest of OpenAI.
« The ability of these models to manipulate public opinion is one of my worst fears, especially in the context of the presidential election. “, thus conceded Sam Altman, who also took care on several occasions to repeat that generative artificial intelligence and social networks were two very distinct things that should not be confused.
The questions that OpenAI will have to answer sooner or later
Finally, behind the consensus around the need to regulate the use of generative artificial intelligence, there were also dissensions likely to increase in the future. The most critical speaker vis-à-vis OpenAI was strangely not one of the elected officials present, but rather the expert Gary Marcus, who notably questioned the altruistic and disinterested speech of Sam Altman.
« OpenAI’s raison d’être was originally to advance artificial intelligence in a way that benefits humanity and is not constrained by the need to generate profit. Seven years later, the organization is largely funded by Microsoft, a private company, which is currently engaged in a bitter battle with Alphabet for supremacy around Internet search, a battle which has led the latter to hasten the marketing of a product, Bard, by neglecting safety. »
Others have pointed out the thorny problems posed by the training of generative artificial intelligence models in respect of copyright. ” We have a vibrant community of artists and musicians in Tennessee. When an artificial intelligence algorithm draws from the works of these artists to generate tracks, how much say do these artists have in it and how will you ensure that they receive adequate compensation? ? asked Republican Senator from Tennessee Mary Blackburn in her very musical Southern accent, referring in particular to Jukebox, a tool launched by OpenAI to automatically generate music.
A question on which the Democratic senator from Minnesota Amy Klobuchar immediately bounced, referring to the local American press, already in great difficulty, and which text generation tools might further weaken. Pressed on these two points, Sam Altman certified that his team was thinking regarding the best way to remunerate the creators of the content on which the algorithms train, without however managing to give a precise answer. To be continued.
1684348772
#Sam #Altmans #OpenAI #call #regulation #isnt #disinterested