What if a robot was able to perform the tasks you would like to exempt yourself from? All these repetitive, boring actions without added value, those that you do out of obligation... And if he did them for you, with rigor, accurately, quickly, everywhere, all the time, in total autonomy and that he learned from his own mistakes and adapted to your preferences and habits? Would you agree to share your daily life with him? For many, rightly or wrongly, this is artificial intelligence (AI). But do we really know this technology which, little by little, is making a place in our lives?

Where are we really in AI?

AI is a catch-all term that feeds fantasies but doesn't stick to reality, because what convinces scientists so far is only machine learning, a small part of what will make up the AI of tomorrow. Today's learning systems are becoming remarkable and allow people to perform feats that we thought were impossible not so long ago, such as translating languages ​​correctly, recognizing speech, or reading medical images.

In some areas, machine learning has overtaken humans. Take the game of go as an example. While it takes a lot of intelligence for a human to play it, the AlphaGo Zero software, produced by DeepMind, a Google subsidiary specializing in AI, has shown the general public that we can increase our level not by boosting our intelligence. but learning from millions of games against oneself. Memory, deduction, anticipation, self-learning, acting out, analyzing situations considered and experienced, and correcting mistakes made to improve performance is not a matter of intelligence but of performance.

In this context, fundamental questions arise: does a machine need to be intelligent to be useful? Can it perform tasks that humans would not be able to perform and vice versa? If so why ? How to establish a stable complementarity between these two parts? What AI can do, it can also undo, how to ensure good use of this technology? How can humans keep control over their smart innovations?

AI, threat or opportunity for our societies?

AI projects are multiplying in France and the Villani report could accelerate the pace. Beyond the buzzword of the moment, underlying technologies such as machine learning, more commonly known as Machine Learning, and its various variations represent a real technological breakthrough. Concrete applications will change our lives in the years to come, for example in the automobile industry with autonomous cars, in health with early detection, or in finance for the fight against fraud. But today the frenzy of building systems incorporating AI is very often done without sufficiently taking into account cyber risks.

Let us never forget: if a machine can detect the origin of a threat, the extent it could take, the faults of a complex system, identify the weak points and bad reflexes of users, it is also capable of creating an attack and bypassing the defense systems put in place to counter it.

On the one hand, AI can make cyber attacks (distributed denial of service, botnets1, ransomware, etc.) more powerful and efficient. The first consequences of the automation of cyber attacks are already being felt, whether they attack critical infrastructures or seek to manipulate public opinion. In October 2016, a distributed denial-of-service attack by the Mirai malware turned internet-connected objects into botnets, remotely controlled botnets, besieging US DNS provider Dyn and French host OVH and causing hardship. access to a large number of websites. Politically-driven social bots can also mimic human behavior on social media to influence users at key political times.

On the other hand, AI technologies can make it easier and better to detect and mitigate cyber threats. More and more applications are integrating AI techniques and in particular Machine Learning techniques in order to better identify threats and anomalies. Machine Learning allows machines to reproduce and recognize given behaviors from examples. Machines are thus able to process large amounts of data in a short time and recognize the smallest changes in their environment, differentiating in real-time normal behavior from those that represent a threat.

Today, to guard against cyberattacks, organizations are turning to Machine Learning. This intelligent technology makes it possible, among other things, to strengthen human analytical skills and capacities, thus providing a more powerful deterrent. But is it enough? The answer is clear: no!

Why is AI so worrying?

In a report recently published by the organization Ifop, 64% of French people say they are “worried” about the development of what they consider to be AI. If their main concern is the impact that this technology will have on their professional life, the consequences of the digitalization of their environment in the personal sphere also worry them. The logic is relentless: the more the population depends on connected devices, the more they could be misused by hackers, the more users will be vulnerable and exposed to cybersecurity problems.

According to PwC's The Global State of Information Security study, the vast majority of French people fear piracy of the technologies that surround them: 36% fear an interruption in the functioning of machines, misuse of their sensitive data, an infringement of property physical, 32% a threat to product quality and 25% risk to their lives.

Philippe Trouchaud, Cyber ​​Intelligence partner at PwC, notes “a real professionalization of attackers. Their motivations are now much more of economic interest (embezzlement, theft of assets, etc.) than mere maliciousness. The attacks, more targeted, better prepared but also better distributed over time, cost more. In fact, both in France and in the rest of the world, cyberattacks are increasing. And this worrying phenomenon is not about to stop because very many public and private actors are investing massively in AI to develop their economy through the modernization of the companies and societies in which we live.

Why is AI so popular with businesses?

In a rapidly changing global economy, PwC's international Sizing the Prize study reveals that global GDP could grow by 14% by 2030 thanks to the rise of smart innovations. This market is expected to weigh around 13.5 trillion euros over the next twelve years, more than the current combined GDP of China and India. The productivity gains generated by AI should represent half of the expected economic benefits. Gaëtan Bodmer, Partner at PwC, comments: “Our analysis highlights the considerable value creation potential linked to AI for companies, arguably much greater than that induced by automation. In the short term, AI will have a strong impact on the offer by allowing companies to automate certain routine tasks, strengthen the skills of their employees, and free up time to devote themselves to tasks with higher added value. In the longer term, it is the growing demand for smart products and services that will drive economic momentum.'

The two big winners in the rise of AI: China and North America, which should generate significant economic benefits, with a respective GDP gain of 26.1% and 14.5% by 2030, that is to say, 9 150 billion euros in total and 70% of the economic benefits at the scale of the planet.

Initially, productivity gains should be faster in North America, due to the region's greater maturity with regard to AI technologies such as Machine Learning, digital assistants, or robots. conversational - more commonly known as chatbots - and the many jobs likely to be replaced by more productive technologies. China could, on the other hand, start to widen the gap with the United States in ten years, once the country will have caught up with its delay in terms of technology deployment and expertise.

In Europe and the developed countries of Asia, the economic benefits of AI should also be significant even if they will never equal those of the United States and China: 9.9% of the GDP of Northern Europe, 11.5% for that of Southern Europe, and 10.4% of the GDP of developed Asia in 2030. In developing countries (in Latin America and Africa among others), the impact will be more limited (less than 6% of GDP) as adoption rates of AI technologies are expected to be much lower.

The industries that stand to benefit the most from AI globally include healthcare, automotive, and finance. Intelligent technology will, in all three areas, increase productivity, product value, and consumption. In health, for example, diagnoses will be much more precise and delivered more quickly, the prescribed treatment will be more personalized and will pave the way for major transformations such as, among others, smart implants. For François Royer, director specialist in data analytics at PwC: “all sectors and all companies will be in one way or another impacted by AI. On the sole criterion of productivity, the consequences could transform the competitive environment. Also, companies that fail to adopt AI quickly could be weakened in terms of lead times, costs, and customer experience. As a result, they could lose significant market share. The major challenge for companies is to recruit the right people, to equip themselves with the right technology and access to data in order to fully exploit the opportunities offered by AI.'