Should we be afraid of artificial intelligence?

At times, we swear that the generalization of software and intelligent machines is the worst threat threatening the planet. It can easily be argued, however, that artificial intelligence raises much wider (and subtle) questions than whether they will sound the death knell of our world.

According to the media, which relentlessly relay the words of Elon Musk or Stephen Hawking, the rise of artificial intelligence marks almost the end of the human species, as inescapable as premature. You talk to Siri on your iPhone, or you buy flowers via a bot on Facebook? You are cooked. Your days are numbered, and it's even more serious if your name is Sarah Connor.

In fact, today's reality, as well as the short- and medium-term prospects, seem very different. If artificial intelligence technologies are experiencing real progress, and their applications are already ubiquitous around us, harvesting human beings trapped in a simulation of the 3D world is still far from being on the agenda.

Artificial Intelligence is plural

We almost forget it, there is today no "artificial intelligence", as there would be a person or a "human spirit". The expression describes all a scientific discipline, with disparate problems, much more than it evokes a hypothetical intelligent, versatile and omnipotent entity. There is no "artificial" intelligence, which remains a concept, but rather a software, machines, and robots each possessing abilities that demonstrate a form of intelligence. And scientifically, the discipline is also conjugated in the plural - machine learning, visual recognition, speech recognition or machine translation, forming many distinct and more or less advanced research areas.

That said, in our pockets, "smartphones" understand our words and can give us advice. On our wrists, "smart" bracelets analyze our vital data or our sleep and detect unusual changes or anomalies. In our homes, appliances of all kinds, alone, optimize the energy consumption and health of our plants or adapt to our mood. On the Web or in our ears, tools translate phrases in foreign languages on the fly. Our cars already know how to park without our help and will soon be able to transport us from one point to another without having to touch the steering wheel.

The good news is that the overwhelming majority of today's "artificial bits of intelligence" (and probably to a large extent tomorrow's) have intelligence only in very specific, compartmentalized and limited tasks. What reassures the worried. The worst AI can do today, even if ill-intentioned, is to miss a spot on the tile that it is supposed to clean. We are far from machine revenge, and there is little risk that your Amazon Echo will rebel against you, other than stubbornly refusing to turn off the light in the living room.

The other good news is that these software and smart devices have no other purpose than to simplify our lives. Who would sulk the pleasure of swapping keyboards and mice against "virtual" interfaces, much more natural and user-friendly?

Artificial Intelligence is stronger than you

Nevertheless, the generalization of AI can lead, even today, a form of "malaise", which is due to the fact that we "delegate" some of our typically human capabilities.

Artificial intelligence has achieved impressive results for a long time. It surpasses the man in many sectors of activities

This month, Google announced reaching a 93.9% pass rate for its automatic image caption tool. A machine can describe a scene (even complex) taking place on a photo, almost as well as a human, and much better than it did four years ago (89.6%). Last month, Stanford University and Baidu of China showed that speech recognition software is much more efficient than humans to transcribe in SMS what it hears: in English, the software is three times faster and makes 20% less error, and it's even better in Chinese, just as fast but with 63% less error. Enough to jealousy all good millennial broken texting frenzied, OMG WTF.

Jean-Gabriel Ganascia, who heads the ACASA team of the LIP6 laboratory, dedicated to cognitive agents and machine learning, recalls: "Artificial intelligence has achieved impressive results for a long time, surpassing humans in many areas of activities ".

"Today, what is surprising is the results achieved in tasks that hitherto seemed rather reserved to men, such as the recognition of speech, scenes and especially faces. deep learning techniques involving millions of examples, which leads to the design of autonomous vehicles, among other things, and many human skills can be automated by machines, "he explains.

Will we accept being "worse" than a machine?

Thanks to deep learning, the machines will only become better, more efficient, more relevant, ending up excelling in areas that we not only knew how to master, but that we were the only ones to master. Faced with a technical or moral questioning, will we be able to admit to being wrong, and agree with a computer that has become omniscient? Will we accept being "less good" than a machine, thus losing our status at the top of the pyramid of living beings? Do we live well to become so "inferior", at least on some levels? We can think that machines will one day make us swallow our intuitions and our emotions, to reveal to us the real "raw" - a cold reality, prosaic, calculated, relentless. Will we be furious and frustrated? Will we end up relying on software to establish whether a person is trustworthy, sincere or "objectively" attractive?

It can scare you, yes. For the greatest change brought about by the generalization of artificial bits of intelligence might be less what they bring us than the way they transform us - and which we accept that they do.

COMPUTING IT VERSUS

The terms IT and IT appeared at about the same time in the literature. The first created by the German engineer Karl Steinbach in 19571, the second by Harold J. Leavitt and Thomas L. Whistler in 1958 and appeared in Harvard Business Review. We already see the different origin of the two terms, scientific for the first, business for the second. The appearance of these terms has been a few years after the term computer, introduced for the first time in 1955 by IBM France, following a request from the manager of the advertising department of the company, François Girard. Previously, the term "computer" had only the meaning of "Calculator".

  

In the general population, the term computer is related to the concept "computer", in all its commercial forms: PC, tablet, mobile, etc., that is to say, the "Machine" and what it can do as a convenience (internet, games video, electronic gadgets). But this last common interpretation is obviously too reductive. In fact, IT encompasses all areas of scientific, technical and industrial aspects involving the automatic processing of information by "machines", whether at the theoretical, practical or industrial. The term computer is derived from the origin of the concatenation information and automatic words. The computer is an example of these "Machines", robots and automatons are others.

The term "information technology" (IT) refers to the application of "machines" for the storage, retrieval, transmission and manipulation of data, the more often in an industrial, business or corporate context. (The term Technology is defined as "the application of scientific knowledge to the production of goods and services "4 ). Because the notion of data transmission is inherent to IT, it's hard to talk about IT without referring to telecommunications, hence the broader term "information technology" and communications "(ICT), which is as widely used as IT and encompasses all types of "communicating machines", their use and the underlying tools (computer, telephone, mobile, TV, radio, microelectronics, photonics, etc.). In Europe, we also find the term telematic5  as synonymous with ICT but it is rarely used.

From these definitions, it can be concluded that the term computer is broader than that of IT because it is specific to the application aspects. But this interpretation is not accepted by all and certainly debatable, especially when we talk about IT and we have in mind ICT ... Certainly, reading these definitions, it is obvious that there is no clear separation between the two concepts but certainly, at least, a juxtaposition. To clarify (or mix!) further, here are some examples of areas that I believe are more IT or IT.

•    Rather "computer" type: programming languages, processing automatic visual, sound and text information, software engineering, "data analytics », knowledge representation, semantics, systems operating, software ergonomics, software development, interaction the analytic aspects of big data.

• More like "IT" (ICT): computer design, networking, maintenance systems, Internet, social networks, interoperability, architecture, hardware, management, processor, storage and databases, web technologies, cloud computing, microelectronics, the storage and accessibility aspects of the "Big data".