If we had to pick just one piece of tech news in 2023, it’s a safe bet that most of us would cite ChatGPT. How can we analyze this irruption of generative AI into our daily lives, starting with our work? Is it a flash in the pan or the beginnings of a sustainable transformation? We asked Luc Julia, one of the designers of the Siri voice assistant.
This article is excerpted from the first issue of the journal Mermoz, “Work: Reshuffling the Cards”.
Is AI a positive revolution for work?
Luc Julia : AI is not a revolution: the only real revolution is that of 1789. In addition, there are AIs, there is not just one AI. On the other hand, we can talk about the evolution of AI, because it has been evolving for 67 years. The evolution of generative AI is very interesting because it carries even more data. When it comes to these AIs, the real revolution is in the ability to use the tool. And again! The revolution is not so much in the “GPT” (Generative Pre-trained Transformer) part as in the “chat” part, i.e. we can now easily access AI, through natural language. Being able to easily maintain a dialogue with AI: this is the revolution, which makes AI accessible to everyone, without the need to know how to code. It can also be a source of potential danger with good or bad uses such as, for example, the fact of generating false images ( deep fakes). In the world of work, again, this is not a revolution but an evolution because once again AIs are easy to access. This tool can be used in a very simple way to iterate and reiterate. This can be true for designers like HR, procurement, or any of the support functions of every company.
We hear that AI can be both a source of significant productivity gains in the workplace and a source of risks. How can we measure these productivity gains and risks in the short and long term?
L.J. It’s not easy to answer the question because we don’t have much hindsight. However, it is clear that all the previous tools have been a source of productivity gains because we are going faster in what we can do than before. This is true if we compare, for example, Photoshop with what you could do with paper and pencil; this is also true for generative AIs that can generate drawings. We can see that there are not only productivity gains, there are also quality gains. When it comes to risks, there are of course because these powerful tools can be used for anything. In addition, generative AIs don’t know everything, they can generate “hallucinations”, errors. For example, if you ask an AI to generate the biography of a well-known person, you may find a lot of errors. These hallucinations, or the act of inventing an answer, can happen frequently because these AIs are meant to give an answer no matter what. When it comes to work, there may be a risk that the work rendered will be very poor. For example, a lawyer asked ChatGPT to generate a plea that was totally false because the cited case law did not exist. Apart from ChatGPT, other generative AIs include Midjourney and DALL-E, which do not generate text but images. They are multimodal and all the more numerous as developer communities develop them in open source. Because the other characteristic of these AIs is that they can do fine tuning. By ingesting our data, AIs will become specialists with much finer levels of relevance.
What sectoral impacts can we expect if the uses of AI were to expand significantly in the next 10 years?
L.J. It is not a question of improvising as Madame Irma, but of plunging our gaze into history and entrusting ourselves to Schumpeter. We have to be very careful when we talk about job cuts, because they are not necessarily linked to the emergence of a new technology. In fact, it can be the opposite. For example, our industry disappeared precisely because we didn’t want to robotize. This makes it difficult to make a prognosis. For the sectors that would be most affected by generative AI, we can certainly mention support functions, i.e. jobs where documents must be produced. It is clear that we are no longer talking about workers or robotics!
On the contrary, it is less manual, more intellectual jobs that can be affected. But it can also go in the right direction, and this is particularly the case with cobotics, through which AI becomes the assistant of humans, who work in cooperation with robots.
By intruding into the lives of animal laborans, does AI offer a greater horizon of happiness to humans?
L.J. Yes, because technology is there to help us. It is important to be aware of both the benefits of technology and its limitations. It’s the man “who’s evil,” not the technology. It is humans who target the applications of technologies: they are good or bad depending on what you want to do with them. In terms of linking AI and reducing working hours, you can’t really do that. Let’s take the example of journalists, who are far from being replaced: they have even more opportunity to show the quintessence of what they know, i.e. to verify and recheck information. Investigation is emphasized even more because it becomes almost more difficult to find the truth. So, on net, there will be a lot more work for journalists!