Artificial Intelligence and the Changing World of Work

Back in 2018, I had the honor of being invited to a “Marketing & Services” internal conference in TotalEnergies. The event’s topic was the changes that will shape the world of work by 2040. I was asked for my point of view on the possible impact of artificial intelligence (AI). As I had a lot of fun preparing the talk, and while the world of artificial intelligence has been evolving at breakneck speed for several years now, I decided to publish my train of thought here, adding a few updates to my state of mind at the time. I must admit my thoughts didn’t change a lot, even if the technological state of the art has evolved a lot with the raise of generative AI.

I won’t be speaking today about general artificial intelligence, which is to say a superior artificial intelligence created by man that pursues its own goals, independently of its creators. A lot has been written on the subject but, however philosophically interesting it may be, I don’t wish to dwell on it here as we still lack the scientific and technological capabilities to create such an intelligence. I’d prefer to focus on the issues that will arise in a nearer future, that is, when using narrow AIs, which are designed to achieve the goals of their creators.

I’d be telling you nothing new if I said that humankind has been trying to automate work for a long time — first by mastering nature (by using animal power, for example), then by mechanization, which experienced a tremendous expansion during the industrial revolution. So, we have a long history of robotizing physical work. AI is a new form of robotization applied to information processing. Here, too, I would like to remind you all that automation has already revolutionized how we work several times. Originally, “computer” was a human profession. It was the job of thousands of people to calculate for others — a job that was rendered obsolete by the arrival of the first computers.

It would be wrong to deprive ourselves of the potential of AI…

AI further extends the scope of automation, making it suitable for logical processing, learning and decision-making tasks in increasingly complex situations. AI excels in more and more areas, often even outperforming people. This has long been the case with arithmetic, chess and game shows like “Jeopardy!” For several years now, machines have been beating people when performing vision and driving-related tasks, or playing Go, Dota 2 or even Poker.

For this reason, the rise of AI as hyper-rational and hyper-specialized machines for processing information is unstoppable and will continue to develop fast. It would be wrong to deprive ourselves of this potential. Furthermore, I think it will be a relief for us, as we all waste so much time finding and compiling information and trying to solve low value-added problems, such as planning a meeting in Outlook or retrieving information in big corporate databases.

TotalEnergies has significantly invested in R&D and digital expertise in Artificial Intelligence by developing in-house solutions and partnerships to accelerate energy transition and those of its customers. Indeed, AI is an extremely powerful lever for transforming the way we work, whatever our business. It is in our E&P platforms to anticipate equipment breakdowns, in our refineries to control electricity and gas consumption and reduce our carbon footprint, in our Saft factories to optimize battery production flow, on construction sites to detect high-risk situations, or in M&S, through the development of fuel pricing solutions.

The question that we must ask ourselves is: “How can I rethink my work processes? In my daily tasks, what can I automate and to what point?” I invite you all to think about it because, honestly, it’s the key to the future of your work. This will clearly transform a lot of jobs, but it will also create new ones. Mine is one of them, for example, and there are others at TotalEnergies, such as data managers and data scientists, who are preparing the ground for massive AI deployment.

…but we need to be very aware of the current limitations of AI

The next question to be asked is: “As a human being, what is my added value compared to AI? What job opportunities can I expect if I don’t want to become an AI programmer?” I think we need to be very aware of the current limitations of AI. Whatever fantasies the media may have, at the current state of scientific knowledge, AI is still very limited in many areas where people excel.

First, AI is less efficient when it comes to putting itself in someone else’s shoes. It thinks from its own objective, fact-driven world perspective, and finds it difficult to take into account ideas about how other agents — human or robot — represent the world. This is known as “theory of mind”. Researchers are working on the problem (referred to as epistemic planning), but still only offer limited responses when it comes to “real life” applications. Human beings, however, excel in this area. In many of our choices, we factor in what we think that others believe. In the future, then, I would recommend that people work on their listening skills, their openness to others and their empathy so that they can be the feeling counterpart to cold and rational AI.

Second, we still aren’t very good at creating “social” AI. For some time now, multi-agent robotics has made it possible to elicit smart behavior from several synchronized agents. But individually, they remain very basic, and the social responses that we manage to simulate are closer to the shoal behavior observed in fish than to that of human society.

So, we are still far from being able to synchronize today’s hyper-specialized AIs to automate complex decision-making processes. In the future, I would advise people to learn how to look at the big picture and hone their ability to perform complex, systemic reasoning. This will ensure that they can effectively take on board input from other people and from potentially contradictory AI to make high-level decisions.

The more machines are able to solve problems, the more we will need discussion

Lastly, I would like to discuss goals. Modern AI can solve complicated scientific problems but remains “philosophically” limited. Overall, AI comes down to maximizing or minimizing a mathematical function. Personally, as things stand, I believe it’s unrealistic to think that the major issues driving humanity, or even our own small, existential questions, can be reduced to a mathematical function. Humankind must remain capable of formulating high-level, overarching goals, rather than restraining itself to the level of problem solving. AI is an excellent assistant for solving a problem. But a problem always leads to another, and it’s important to maintain an overall view, to steer a more general path, in order to give meaning to actions, be they moral, religious, philosophical, economic, hedonistic or other. Today, a machine cannot set such goals.

To conclude, my feeling is that if you really want to take advantage of AI, some organizational issues need to be considered. Highly mechanistic organizations, where each employee focuses on a very specific information processing task, possibly don’t make a lot of sense. A more appropriate target would be to create organizations that encourage discussion and exchanges to foster openness and debate and create of shared sense of purpose. I think that the more machines are able to solve unitary problems, the more we will need discussion, transparency and cohesion, to ensure we can collectively commit to a shared vision of a world that satisfies us.