Head of AI Projects at LCL

It has been more than a year since ChatGPT invaded the world, and AI projects in companies have sprung up like mushrooms. Among these charming representatives of the plant world, some are delicious, some barely edible, some will give you a good fever, and a few others are simply lethal. Would it be the same with AIs? This parallel even holds regarding the glorious EU AI Act, which is the envy of the world and whose risk classification bears a striking resemblance to the general typology of evil in use (or in ingestion) that makes this gourmet text resemble a health manual.

And it's true that AI raises questions at a societal level, but the important issues it raises are likely to be ignored in favour of a race to enhance more business. So many billions have been invested, especially across the Atlantic, that it's out of the question not to get a return on it for minor ethical considerations! Not to mention ecological issues—the poor relation of technological upheaval—leading us collectively to a worrying form of schizophrenia (it's up to your neighbour to be frugal; luckily, he thinks the same of you...!)

Does AI have something in the pipeline to lift our spirits? The answer is undoubtedly "yes". AI is a technology and discourse, but first and foremost, it is a tool and even a whole toolkit. With the growth of digital flows, companies are required to generate and then process ever-growing masses of data, the full potential of which (or a substantial part of the potential, let's say...!) has been understood for several years now. Obviously, there is a long way to go from this awareness (and it's not for lack of repeating it to anyone who will listen) to the operational—and operative—activation of the levers that AI enables.

In this respect, the wave of generative AI is good news: no one can ignore the existence of artificial intelligence, or its power to act, any longer (although we must calm down the magical representations that most people still have of it; some fantasies are hard-boiled...). Some want generative AI everywhere when it's not necessarily the best solution. We must be careful not to give in to the siren song of technological solutionism, especially as these technologies are recent. In the corporate world, it's best to do what we've always done—and what a surprising thing it is to have to be reminded of this!—we use a tool when we have a specific need. It would never occur to anyone (our managers included, wouldn't them?) to acquire a screwdriver when there are no screws to screw in. Business needs should be at the forefront of AI projects, especially since the arrival of these chatty conversational agents, and it's the desire to use them that drives the project. For what kind of results?

Rest assured; generative AIs will penetrate the intimacy of companies on their own as the major players in the market make software upgrades. And after all, who's going to complain about being able to speak to his computer in sentences rather than having to code in abstruse languages? We're still not there, but we are progressing towards this aim...! For the moment, wizards are offering you the opportunity to write meeting minutes. That's a good start, even if the real report of a meeting is more the action plan that follows than the automatic—and rather flat—summary of the exchanges...Other use cases will come, perhaps making us regret the AI that’s wrong in a binary way compared to these AIs that claim to bring answers when they only bring text. The enthusiasm is there, but it never ceases to surprise us: would our fellow creatures be content with the low-end productions of generative AI? (Indeed, how much garbage is there for a few exceptional successes in image or video synthesis?)

But let’s not deny ourselves the pleasure! Generative AI comes with remarkable technical underpinnings, the LLMs (“Large Language Models”), which make it possible to finally tackle the dreaded question of chatbots and search engines, with some hope of cracking them. For the rest, these projects are bringing other more familiar, less fashionable, but surely more useful projects into their wake: stories of the automation of numerous, off-putting documentary tasks, which are often the remnants, left to employees, of current processes’ blatant inadequacy with respect to an information system in which there has been insufficient investment. Computer vision still has a long way to go until the digitization of the world takes over (but we've been hearing that for ten years now...!). AI is enjoying a new spring unless we've already reached summer, which is, alas, now tending towards global overheating. Let's hope for the best!