The year 2017 was marked by numerous issues in the cyber domain. We can mention the successive ransomware waves but also the imminent arrival of the GDPR (General Data Protection Regulation) which is the subject of many conferences, posts, forum topics, etc. Another great subject was the “second birth” of artificial intelligence and machine learning. Concerning the artificial intelligence (AI), we are still far from Skynet from Terminator, I can assure you. However, it has to be said that a lot of progress has been made recently and that it seems to be the beginning of a great and beautiful adventure.
On LinkedIn, every day there is a post mentioning the GDPR or the artificial intelligence. My purpose today is to clearly redefine what machine learning is in relation to artificial intelligence.
Artificial intelligence (or AI) is a subset of computer science (the one that designs intelligent machines) while machine learning is a subset of AI. These two expressions have been frequently used lately in the scientific literature or on the Internet. However, it is important to keep in mind that “machine learning” and “artificial intelligence”, although closely related, both present subtle differences.
The artificial intelligence and by transitivity the machine learning fields appeared in the second half of the 20th century. Machine learning is therefore nothing new, it is the effective implementation of methods and algorithms which support the artificial intelligence.
The purpose of machine learning is to allow a computer to provide answers and solutions to problems it may never encounter. It begins with a learning phase, during which the computer (the program) is fed with examples from a learning database in order to allow it to modify certain parameters that will provide answers and solutions to these examples. Once the learning phase is completed, a second phase begins during which (by way of illustration because it is not the only possible application in this field) a predictive system will allow the computer (the program) to provide answers to problems it may have never encountered.
On September 8, 2017, Prime Minister Edouard Philippe entrusted the mathematician and Deputy Cédric Villani with a mission on artificial intelligence. On that occasion, the Prime Minister presented the four main challenges of artificial intelligence.
- To begin with, there are economic challenges. Indeed, all forecasters and specialists agree that AI presents a strong and steady rise over time. Double-digit growth rates have been announced for many years to come. However, it should be noted that machine learning has lately expanded with the arrival of the Big Data and especially the availability of enough computing power.
- Secondly, we can see that there are also social issues which can have an impact on the labor market, on how we work in the medium term. Machine learning can, for example, allow us to achieve significant gains in productivity and ergonomics.
However, I would like to point out that even if the digital revolution may eventually destroy jobs, I believe above all that it is transforming the professions. In fact, from my point of view, we will always need human expertise to complement the work of a machine.
- The third type of challenge concerns the security and sovereignty issues. Indeed, it is important that we maintain our independence in France, that we be a proactive force and a leader in the field.
- Finally, there are also ethical issues that raise a number of reflections. The best known being the determination of liability when an autonomous vehicle is involved in a road accident or “who deserves to die” (the old one or the young one), for example. Questions that I will not answer today…
Against these four challenges, France has major assets including specialized companies in the domain. These companies offer reflection, innovation and intelligence. Systancia is fully engaged in this domain with its history, its research and innovation center “La Forge” and above all, its solutions.