To better understand the challenges of machine learning, particularly in the automotive industry, we interviewed Charles-David Wajnberg, who heads the department dedicated to artificial intelligence and big data within ESGI.
Machine learning offers many opportunities for businesses. © NicoElNino – stock.adobe.com
Branch of artificial intelligence that allows computers to learn to solve tasks without having been programmed specifically for this purpose, the machine learning offers the possibility of developing algorithms to analyze a very large amount of data more quickly and more precisely. ” the machine learning, or artificial learning in French, is a research and application discipline that aims at implicit programming, as opposed to traditional programming which requires the design of ad hoc computer code. Much of the machine learning concerns classes of algorithms capable of inducing representations, models and programs from computer data », Explains Charles-David Wajnberg, educational director of the Artificial Intelligence and Big Data (IABD) branch of ESGI.
The multiple challenges of machine learning, omnipresent in companies
While artificial intelligence was born in the 1940s and 1950s with the beginnings of computing, the artificial learning models designed today represent statistical techniques that apply to very large amounts of data. ” The algorithms of machine learning are essentially based on induction mechanisms, that is to say, identification of representations and rules which generalize disparate information. For example, an algorithm, which learns to distinguish on medical images a benign tumor from a malignant tumor, does not start from any medical theory, but from a simple preliminary labeling of the images. There is thus a whole stake in articulating human theories, their practices, with digital representations and their impacts. », Continues Charles-David Wajnberg.
Yes the machine learning makes it possible to exploit the full potential of big data, it is necessary to return to the context in which this massive data appeared, distributed on a very large scale. ” The idea is indeed not new because, at the end of the Second World War, scientists understand that nascent computing is inseparable from the data flows it will process. Eighty years later, here we are: data and calculations are distributed over vast infrastructure networks. Hardware, software, protocols and languages are evolving to process ever more data, in ever shorter times. The omnipresence of machine learning in our computerized societies raises the question of the possibilities and limits of these techniques, including their validity in many situations. »
With data making it possible to more or less directly link individuals, companies, whatever their size, and even States, the machine learning constitutes ” a technical fabric interwoven with all the professional, economic and political aspects of human activities. Suffice to say that the stakes are not lacking! »Thus, companies or institutions have the ability to analyze and understand what values they can derive from the data they have, how to automate the integration or trade.
The benefits of machine learning in the automotive industry
the machine learning offers many advantages to companies. In the automotive sector, it provides greater autonomy to driving. ” The algorithms embedded in an autonomous vehicle produce behavior similar to that of a human driver. Clearly, cameras, lidars (laser imaging detection and ranging, or laser remote sensing) and other sensors, allow the vehicle to make a representation of its environment and to control its mechanical systems accordingly. »It is thus possible to recognize a specific object, follow its movements and anticipate its trajectory. ” This information is communicated to a control system which has been taught the correct reactions, in a simulator or in a test environment. »
If GPS or automatic parking are the first examples of use cases, which now punctuate the daily lives of motorists, the complete autonomy of vehicles is based on a technique based on operations. ” Experiments have been launched in a large number of countries and there are now legislative and ethical questions, in particular, that arise. One example among many: who owns the driving data of the vehicle? Is it the driver or the manufacturer of the autonomous car? Who is responsible in the event of an accident? These questions were, for example, on the French legislative agenda in 2021. A decree, published on July 1, 2021, has in fact formalized the new regulatory framework for the circulation of autonomous vehicles in France, which will be authorized to drive from September 2022 on predefined routes or zones.
How to train to master the techniques of machine learning?
A complex discipline to tackle but just as fascinating, the machine learning requires mastering essential technical skills, such as mathematics or technological environments (codes and platforms). ” It is also necessary to be able to understand the context of the professions in which the algorithms will evolve, to make the link between the logic of the machines and the necessities of the associated human activities. », Adds Charles-David Wajnberg. Professional opportunities are numerous and concern all types of activity sectors: energy and the environment, transport, commerce, health, IT security or even cybersecurity.
ESGI, which trains learners in IT professions to meet the needs of companies, offers a course dedicated to Artificial Intelligence and Big Data. Two cycles are on the program, the Bachelor (bac + 3) and the Master (bac + 5), which can be followed alternately at the rate of 1 week in progress and 3 weeks in a company. ” The teaching of this sector brings students to a very high level of competence in machine learning. Research publications are studied and implemented, as is the application by the technologies in use. The ESGI ensures that students do not just have a use of algorithms, but a deep understanding of the mechanisms implemented and the associated issues, industrial or legal, which makes them responsible experts. »
Receive all digital news by email