PIE: How does AI affect businesses in their day-to-day business and activities?
Juliet Matioli: There are several strong examples of AI’s impact on organizations. We have companies that use this technology to better manage call centers, their B2C inboxes, and to automate their mailboxes with customer complaints that need to be processed. For example, one of the first AI applications in the call center is to be able to automatically sort emails to the right person depending on the problem (maintenance, invoice, etc.).
The problem today is that SMEs and small businesses, which are not necessarily technology companies, can still benefit from this technology. Especially worth mentioning Hub France IA, an association that helps small structures understand and develop AI-based applications, may affect areas we have never considered before (optimizing customer journey, marketing, fashion, etc.).
This discipline applies to everyday activities: knowledge management of a company, optimization of processes. Another example, in the case of human resources, for matching between demand and supply, for the first selection of CVs when applying for a job, is somewhat like the “metic” of a job. There are really many activities where we can use AI
PIE: Are we able to measure the economic benefits of AI in a practical way? Do you have examples?
JM: Yes completely, we can measure both social and economic benefits. In Thales, for example, we did a study for a maintenance system for which there were (regular) maintenance contracts. We’ve been able to optimize the maintenance of these fleets for AI and implement predictive maintenance that makes it possible to optimize these maintenance schedules and reduce production maintenance costs by about 50%. We may also consider using AI to optimize a process, especially through machine learning to understand the non-quality of certain parts of the production chain. Other uses make it possible to win new customers or contracts faster. So there are many uses where economic benefits are measured very quickly, especially in the case of operations or industrial production.
PIE: How would you describe the gap in the progress of AI research between countries?
JM: This is a problem for both the way and the population. The dominant countries in this sector were once England and the United States. Today, China is beginning to overtake the United States, if only in the number of publications at major scientific conferences. Population plays an important role in these results, with China necessarily having more researchers than France due to the differences between the two populations.
PIE: Will AI be able to compensate for demographic vulnerabilities in certain countries?
JM: It depends on what is called “demographic vulnerability”. AI will compensate for this demographic vulnerability in some cases but not in all. AI will not replace people but will complement them as a great help for the elderly population. We can imagine solutions based on AI such as health monitoring system, mini-companion for family, automation of certain tasks that make it possible to take care of the elderly. In addition, new occupations will appear while others will disappear and some will develop. Yet, human and social relationships cannot be replaced by one machine.
PIE: How is AI’s disaster revolutionizing information warfare and how can France protect itself from this new emerging threat when it has the weapons to respond?
JM: Information warfare is not new, the way to do it is growing through AI This is especially true for fake news (deep fake propaganda, distorted images, etc.)
To avoid this fake news, we must learn to be skeptical when using digital media where the emergence of this media allows for the viral sharing of this false information and therefore can lead to information warfare. Some big news agencies or media outlets are launching AI Cells to pick out the real ones. Three or four years ago, a TF1 reporter explained to Vinatech that while cross-checking was increasingly necessary to confirm the veracity of information, we return to the idea of trusted AI. Indeed, the accessibility of these means of creating credible false information and disseminating it in digital media is becoming more democratic.
PIE: In AI, we saw computing power through quantum computing as a catalyst for its development, what’s less well known?
JM: It is certain that Quantum more broadly represents a major breakthrough in AI and digital technology. As soon as we start being able to do quantum computing, new algorithms can be created and we will enter another dimension.
Digital technology, on the other hand, is nothing more than “green”. Their carbon footprint is not good because they consume a lot of electricity. To answer this problem, there is a movement around AI and machine learning, to think about frugal AI, moving towards a green AI in terms of energy (and data), which consumes little energy. Energy that does not pollute the planet. The study was conducted exclusively by a researcher named Julie Grolier, who designed highly energy-efficient architectures (outside of CEMOS).
PIE: What are the ethical issues of AI?
JM: There are very technical issues of ethics, responsibility and transparency. For example, when we learn, we inevitably have biases, and we must develop tools and methods to enable us to identify and control these biases. These biases are also present when modeling (cultural modeling bias). In fact, one European or one Asian will not have the same solution to the problem. The ethics are quite complex, fortunately Europe has issued recommendations, which allow you to ask the right questions.
PIE: Is the suspension of AI in Europe a break on France’s power in the field and especially in geopolitical strategies to take advantage of technology? With the risk of addiction that may include this.
JM: Legislation is being created, with AI legislation that would impose restrictions or standards on AI for critical systems or when people are in the loop. This can be a break because critical systems have been used for a long time and already have regulations or certifications (aircraft, cars, etc.). Adding additional restrictions because we’re using AI could potentially slow things down and make companies reluctant. Thus, an AI-based critical system must respect the same limitations as the “classic” critical system. For example, some systems, such as video surveillance and their extreme use, already have laws that protect us from potential abuse. So yes it could be a break if we go too far in legislation and standardization, which is why France is building itself up to be a player in standardization, including Pillar 3 of the Great National Challenge of the trusted AI.
PIE: Today, what are the next necessary steps in the development of AI?
JM: For AI to improve, it needs to increasingly reach out to systems, products, and solutions for all. For this, it is necessary to educate a population like Finland, which has launched a very interesting initiative: to create an accessible MOOC for all to raise the awareness of citizens about the use of AI with the aim of training at least 2% per year. Population. So the first step involves educating the people about artificial intelligence.
The second phase focuses on scientific and technical training. It is important to train the population very early, not to be an AI engineer and not necessarily a researcher after graduation very quickly. Promoting this culture and making it accessible to engineers will make it possible to transform the proofs of concepts to revisit most areas of activity through the prism of artificial intelligence.
Interviewed by Yassin Ivalitin For Club Information-intelligence
Part One: [CONVERSATION] Juliet Matteoli: Artificial intelligence, a new growth lever for economic intelligence? [1/2]
For the future: