There is no opting out: so-called artificial intelligence is broadly defined, in all media. It scares people, fascinates them, and blows them away, but is going into forced circuits in companies. According to Gartner, it will be embedded in 50% of analytical applications within 3-5 years. R & D, IT services and customer support concentrate...
There is no opting out: so-called artificial intelligence is broadly defined, in all media. It scares people, fascinates them, and blows them away, but is going into forced circuits in companies. According to Gartner, it will be embedded in 50% of analytical applications within 3-5 years.
R & D, IT services and customer support concentrate the bulk of AI business projects. Faced with US giants (GAFA in the lead) and Chinese (Baidu, Alibaba and Tencent), Europe is trying to resist thanks to its excellent scientists and engineers. But when the giants do not ask philosophical questions about AI, France, proclaims at the turn of the tests that “it does not exist” (Alain Julia) or that “it is largely overestimated (Pierre Blanc, Athling). One and the other are probably right. But these algorithms that imitate the natural man, whether we like it or not, are about to call into question a millennia of practices.
The intelligence of metadata
Technology has given voice in recent years in smartphones. Since then, AI has been disrupting business methods and business services. For professionals, this new algorithm plays the grails of data analysis. Every six months that passes sees another new study appear, smashing on the AI. One of the latest, PWC’s “Sizing the prize”, even says that global GDP could grow by 14% by 2030 thanks to AI. Productivity gains from artificial intelligence technologies are expected to account for half of the expected economic benefits. It is time to create new jobs or to reconsider human employment on a new foundation.
But what is AI? Reference Standard ISO 2 382-28 defines it as the “ability of a functional unit to perform functions generally associated with human intelligence, such as reasoning and learning.” From this generic technological envelope, branches of the AI tree grow in all directions. This is for example machine learning (ML) – the development of learning process by which the machine evolves, the application, to build a predictive model. It’s about designing algorithms that can quickly analyze huge volumes. ML establishes correlations between two events rather than a causal link. But within ML, it is Deep Learning that packs the machine: “deep learning” is based on an artificial “neuronal” system. Hence the ability for the Frenchman Mediawen to offer the first of its kind in the world, the simultaneous translation is straight out of Science Fiction …
What AI already knows how to do alone
AI is for everything. Since the detection of melanomas from images of moles to the improvement of the predictive epidemiological watch. But all sectors and businesses will one day be driven by “cybernetic” technologies, an old term that includes the mathematical theory of information. In a report entitled Notes from the AI frontier, insights from hundreds of cases – published in April, McKinsey Global Institute is ready for AI capability, among other things, to optimize customer services, purchase recommendations and dynamic pricing. Not to mention the improvement of logistics and predictive maintenance.
Specifically, AI, via machine learning, can identify and quickly process all the data relating to consumer behavior on a website. In order to personalize the customer journey and improve the experience on e-commerce sites.
As a result, marketing departments have issued more and more announcements of AI coupled with big data: to be convinced of this one needs to examine the state of the art of global marketing established at the end of 2018 by Salesforce. In France, a third of marketing managers use it, up 47% compared to last year, mainly by automating, by AI technique, Customer Relationship Management (CRM) databases. Hence the emergence of chatbots – deemed increasingly effective in decrypting the demands of consumers or interlocutors. If the human interlocutor ignores it, pretending to be an advisor for everything human, at least for the most advanced applications.
Until now, these interfaces were “analyzing” from the detection of keywords. With AI, they begin to “understand” the natural language. There is no doubt that the biggest dream of Steve Jobs, the creator of Apple, embodied in the first big clip of the future of the year 1987 (Knowledge Navigator, visionary precursor of Siri), will soon be a reality: a teacher was talking with his camera in a way that could not be more natural, his virtual “assistant”, in the form of a young man with a bow-tie, answered him on the spot, provided him with dynamic data that the man worked in real time. warned of incoming calls or explained to the interlocutor that his “boss” would come back in an hour …
These computer programs could disrupt many industries in the longer term. Being able to take on different roles such as: salesman, doctor, consultant, stylist, lawyer, tour guide, culinary critic … In Germany, the Lufthansa group uses a chatbot to guide travelers in their search for the best price. In France, several companies are conducting tests or have deployed first versions of their bot: SNCF, Direct Energy, Accor, PMU …
Lack of relevant data
However beware of excess optimism because many AI models are not generalizable. Artificial intelligence still has some progress to make. Just as businesses have. The previously cited McKinsey study shows that it is still very difficult to explain complex models to decision makers in a simple way. Another obstacle is the lack of relevant data. This study highlights the difficulty of obtaining sufficiently large and complete data volumes to be exploited by certain trades or sectors. The resulting dataset will have to go through a formatting or transformation phase before the end user “uses” it, starting with the general instructions. Because a given is not a fraction of the truth. It is always subjective. It is therefore necessary to sensitize professions that are not supposed to master the intricacies of complex statistical methods.
But it is especially the lack of qualified profiles that represents the main obstacle. Many organizations lack skills in machine learning and data science. They are also looking for profiles that can identify and value cases of professional use of AI.
Globally, there are only 300,000 AI researchers and practitioners, while the demand is in millions, according to a survey by Tencent Research Institute in December 2017. So, HEC, Essec, Polytechnique or Telecom ParisTech adapted their courses and created specialized masters in AI.
To loosen the stranglehold, the mathematician Cédric Villani had proposed in his report to triple the number of people trained in AI in three years, including expanding the field of talents to baccalaureate +2 and +3. And these talents will not be disappointed by their wages. According to a 2019 Data Recruitment study, a data scientist starts at 40/44 K euros to reach 80/100 K euros and even more for people who have more than ten years of experience.
For France and Europe, the stakes are high because it is about resisting AI leaders in the United States and China. France does not count as a heavyweight in this area. On the other hand, France abound for almost nearly 270 start-ups. In 2017, $ 141million was invested in these seedlings. Some have distinguished themselves as Prophesee (ex-Chronocam). Launched in 2014, it has developed a patented innovation, unique in the world, in which machines imitate the human treatment of images, but at what rate! Actility, which specializes in the Internet of Things, Shift Technology, which tracks insurance fraud, and Dataiku, which publishes software aimed at improving business data analysis, are virtually successful.
20 billion euros of investments in Europe
In Europe, British company Darktrace has achieved distinction. It uses algorithms that mimic the human immune system to defend corporate networks against cyberattacks. Another nugget to attract investors, Acrolinx. Its marketing platform uses AI to create targeted content for brands
But will these nuggets long resist the appetite of Americans or Chinese? To stay in the race, Europe is multiplying its support. Its investment in research and innovation amounts to 1.5 billion euros for the period 2018-2020 under the 2020 Horizon program. In addition, the European Fund for Strategic Investments will be mobilized to support businesses and start-ups, thanks to additional support, to invest in AI. This fund aims to mobilize more than 500 million euros of investment in total by 2020 in a range of key sectors..
Another initiative is AI4EU (Artificial Intelligence for European Union). Launched in January 2019, led by Thales, which has a budget of 20 million euros over three years. Vocation: to bring together and animate the European AI community within a single entity.Overall public and private investment in the EU is expected to reach at least 20 billion by the end of 2020.
An essential envelope so that Europe does not miss the AI train on the move. The Washington Post, in September 2018, published the words of Kai-Fu Lee, the American venture capital investor, writer and researcher based in Beijing, China, who clearly indicated that AI in Europe has made the continent a colony of the American technological empire … Kai-fu Lee knows this area well. He created a $ 1.6 billion investment fund dedicated to AI. Many of his works are considered essential to the current development of AI. Let’s be positive: he also said that the General Data Protection Regulation (GDPR) will give European entrepreneurs a chance to create a more user-oriented experience, which is generally lacking in US companies …
Article taken from EcoRéseau written by Philippe Richard