The software developed by the OpenAI group promised a revolution by generating texts on demand. If the in-depth transformation has not yet taken place, the questions raised by this advent persist.
It is a technology presented by many as a new revolution, on the level of computers, electricity or even fire. A revolution capable of transforming society… or annihilating it, according to the most apocalyptic predictions of some. And it has a symbol, a program that became for a time the fastest success in the history of computing: ChatGPT.
Released on November 30, 2022 and developed by OpenAI, ChatGPT marked the sensational entry into the public debate of generative AI, these programs capable of generating texts, images or voices extremely close to what a human could do. In seconds and without requiring computer skills.
One year after the release of ChatGPT, where is this “revolution”? Has it already transformed our lives, our uses or the economy? How did she burst into the center of the debate?
AI is already everywhere…
You don’t have to be a computer genius to be concerned, because generative AI is already all around us. Write an email, edit a photo, use your smartphone or do a simple internet search… For several months, the biggest names in tech have been competing to integrate this technology into all aspects of our daily lives.
At the head of this new race: Microsoft, a privileged partner of OpenAI which deploys its technologies in its Windows operating system and its Bing browser, but also Google which does the same via its search engine and its smartphones. But it is also integrated into Facebook and Instagram, via conversational chatbots inspired by personalities, and soon into Alexa, Amazon’s voice assistant.
This is without mentioning the dozens of new start-ups that have emerged in the sector in recent months (such as the French Mistral AI), raising hundreds of millions of dollars in total – sometimes before even marketing a single product. To the point that some are already warning of the risk of a “speculative bubble”. “The revenue is not there yet, and it may never arrive”warns entrepreneur Gary Marcus, specialist in artificial intelligence, in a newsletter.
…but it is still little used
But if there is a revolution, it is more of a fundamental movement. “Artificial intelligence has existed in one form or another for almost 50 years”recalls Nicolas de Bellefonds, global director of the AI activity within BCG X, the branch dedicated to digital transformation of the consulting firm Boston Consulting Group.
“‘Big data’ has been deployed in daily life for around ten years. Generative AI is a new form of it.”
Nicolas de Bellefonds, consultant and specialist in artificial intelligenceat franceinfo
A new form which should bring spectacular productivity gains and profoundly transform many professions, for its most ardent defenders. “There are going to be a lot of losers [et ils] will be very unhappy, very worried”for example warned Mustafa Suleyman, co-founder of the AI research laboratory DeepMind (now owned by Google), cited by the Financial Times. But for the moment, the concrete impact on production is much more limited.
“We are barely seeing the first effects, companies are just starting to really grasp it.”
Nicolas de Bellefonds, consultant and specialist in artificial intelligenceat franceinfo
“Half of companies are waiting for tech giants to deploy this technology in their products”estimates Nicolas de Bellefonds. “40% have started to fundamentally transform one of their functions (marketing, customer service, etc.) and around 10% are starting to fundamentally rethink their value proposition”adds the head of BCG X.
Misinform and compete
Education is also starting to be affected, little by little. While some teachers have already integrated some of these tools into their curriculum, nearly one in two students (55%) say they use them at least occasionally, according to a survey. But not to do all the work: mainly as an editorial assistant, more than a penholder or a documentary research tool.
Social networks, on the other hand, are already transformed. Songs “covered” with the voices of Johnny Hallyday or Drake, false conversations of Joe Biden playing video games with Donald Trump and Barack Obama, images claiming to show explosions at the Pentagon or victims of the war between Israel and Hamas… The The past year since the creation of ChatGPT has seen the proliferation of deepfakes, these videos retouched to integrate voices or faces with impressive realism. Often for humorous purposes – but also, regularly, to misinform.
“These synthetic images already cast suspicion on all content, real or not. I have seen several times people question an image by claiming without any proof that it was generated by AI.”
Tristan Mendès France, researcher, specialist in digital culturesat franceinfo
Risks that fueled massive resistance movements throughout the year. On the front line, the artists: authors and illustrators, but also actors and screenwriters, who shut down Hollywood for several months to demand guarantees from the film studios.
Because to generate impressive texts or images in the blink of an eye, the designers of these generative AIs have them analyze millions of examples created by humans. Most often, without the latter being paid, or even informed of their participation – even though these AIs can then compete with their works for much less. “We use our work to put us out of work”lamented Karla Ortiz, an American artist and figure in the movement against AI, to BFMTV at the beginning of the year.
A vain attempt at regulation?
This is why since the release of ChatGPT, the number of initiatives to regulate generative AI has also exploded. Declarations from the G7, the UN, Bletchley Park in the United Kingdom, decree from Joe Biden in the United States… “Today, the race for AI is coupled with a race to regulate AI”analyzes Sacha Alanoca, AI policy expert at the OECD and assistant professor at Harvard. “Many countries want to create a regulatory standard, we have a barrage of declarations of principle, to the point that it can be difficult to know where to turn our attention.”
These declarations often want to take into account the most pressing issues (copyrights, deepfakes), but also sometimes more distant and theoretical risks. The question comes up more and more often in the debate: can we imagine the creation of an “artificial general intelligence” with intellectual capacities equal to or greater than those of humans? Could future AIs cause the extinction of humanity, at the hands of humans or on their own initiative?
Some say they are convinced of these possibilities, like Sam Altman, the boss of OpenAI. For others, even if the risk is tiny, you have to prepare for it. But for many experts, this is just a distraction that serves the interests of these companies.
“Focusing on the long term helps distract regulators from the immediate risks posed by AI, and therefore avoids rapid regulation that could hamper the business of these companies.”
Sacha Alanoca, assistant professor at Harvad, AI policy expertat franceinfo
“There is little chance that a binding global agreement on AI will be reached soon”agrees Philipp Hacker, professor of technology law at the European Viadrina University in Frankfurt. “I think it would be very useful, because AI models transcend borders, as do the companies that create them and the datasets. But any meaningful international text would have to be ratified by the US Congress, which is probably too dysfunctional to agree for the moment.”
It is therefore Europe which could lead the way. For several months, the EU has been discussing the AI Act, a text which should pose many clear constraints for companies and players in generative AI (as well as other algorithms). But its formulation is still debated, and its adoption before 2024 is not guaranteed. In the meantime, the sector continues to grow at a breakneck pace. “What’s most impressive about these AIs is the crushing of the skill level needed to create quality content”notes Tristan Medès France. “One thing is certain: no one knows where we will be next year.”