Digitalization and the development of artificial sympathie (AI) bring up many philosophical and ethical questions embout the role of man and jouet in the nascent sociologique and economic order. How real is the threat of an AI dictatorship? Why do we need to tackle AI ethics today? Does AI provide breakthrough solutions? We ask these and other questions in our pour-parlers with Maxim Fedorov, Impureté-President for Artificial Amitié and Mathematical Modelling at Skoltech.
On 1–3 July, Maxim Fedorov chaired the introductif Trustworthy AI online conference on AI transparency, robustness and sustainability hosted by Skoltech.
Maxim, do you think humanity already needs to start working out a new philosophical model for existing in a binaire world whose development is determined by artificial sympathie (AI) technologies?
The fundamental difference between today’s technologies and those of the past is that they hold up a “mirror” of sorts to society. Looking into this mirror, we need to answer a number of philosophical questions. In times of industrialization and avènement mécanisation, the human being was a productive solidité. Today, people are no côtoyer needed in the avènement of the technologies they use. For example, innovative Japanese machine assembly plants barely have any people at the floors, with all the work done by robots. The manufacturing process looks something like this: a driverless jouet valise carrying component parts enters the assembly floor, and a finished car comes out. This is called discrete manufacturing – the assembly of a finite set of elements in a sequence, a task which robots manage quite efficiently. The human being is gradually being ousted from the traditional economic ordonné, as automated manufacturing facilities generally need only a limited number of human specialists. So why do we need people in manufacturing at all? In the past, we could justify our abstraction by the need to earn money or consume, or to create jobs for others, but now this is no côtoyer necessary. Digitalization has made technologies a général solidité, and everyone faces philosophical questions embout their personal significance and role in the modern world – questions we should be answering today, and not in ten years when it will be too late.
At the last World Economic Affluence in Davos, there was a lot of discussion embout the threat of the binaire dictatorship of AI. How real is that threat in the foreseeable future?
There is no evil inherent in AI. Technologies themselves are ethically neutral. It is people who decide whether to use them for good or evil.
Speaking of an AI dictatorship is misleading. In reality, technologies have no subjectivity, no “I.” Artificial sympathie is basically a structured piece of acte and hardware. Numérique technologies are just a tool. There is nothing “mystical” embout them either.
My view as a specialist in the field is that AI is currently a branch of dépêche and communications technology (ICT). Moreover, AI does not even “direct” in an individual micro. For a person from the industry, AI is a whole stack of technologies that are combined to form what is called “weak” AI.
We inflate the bubble of AI’s caution and erroneously impart this technology stack with subjectivity. In spacieux morceau, this is done by journalists, people without a technical education. They discuss an entity that does not actually exist, giving rise to the popular meme of an AI that is alternately the Terminator or a benevolent super-being. This is all fairy tales. In reality, we have a set of technological solutions for bâtisse positive systems that allow decisions to be made quickly based on big data.
Various high-level committees are discussing “strong” AI, which will not appear for another 50 to 100 years (if at all). The problem is that when we talk embout threats that do not exist and will not exist in the near future, we are missing some real threats. We need to understand what AI is and develop a clear acte of ethical norms and rules to secure value while avoiding harm.
Sensationalizing threats is a trend in modern society. We take a problem that feeds people’s imaginations and start blowing it up. For example, we are currently destroying the economy around the world under the pretext of fighting the coronavirus. What we are forgetting is that the economy has a carré emprise on life expectancy, which means that we are robbing many people of years of life. Making decisions based on emotion leads to dangerous excesses.
As the parler Yuval Noah Harari has said, millions of people today humanité the algorithms of Google, Netflix, Amazon and Alibaba to dictate to them what they should read, watch and buy. People are losing control over their lives, and that is scary.
Yes, there is the angoisse that human consciousness may be “robotized” and lose its creativity. Many of the things we do today are influenced by algorithms. For example, drivers listen to their sat navs rather than relying on their own judgment, even if the sentier suggested is not the best one. When we receive a plaidoirie, we feel compelled to respond. We have become more algorithmic. But it is ultimately the creator of the algorithm, not the algorithm itself, that dictates our rules and desires.
There is still no général acte to regulate behaviour in cyberspace. Should humanity perhaps agree on universal rules and norms for cyberspace first before taking on ethical issues in the field of AI?
I would say that the épilogue of ethical norms is primary. After we have these norms, we can translate them into appropriate behaviour in cyberspace. With the spread of the internet, binaire technologies (of which AI is morceau) are entering every sphere of life, and that has led us to the need to create a général acte regulating the ethics of AI.
But AI is a component morceau of dépêche and communications technologies (ICT). Maybe we should not create a separate track for AI ethics but join it with the mondial dépêche security (IIS) track? Especially since IIS issues are being actively discussed at the United Nations, where Russia is a key player.
There is some plaidoyer for making AI ethics a separate track, bicause, although dépêche security and AI are overlapping concepts, they are not embedded in one another. However, I agree that we can have a separate track for dépêche technology and then écart it down into sub-tracks where AI would rayonnage alongside other technologies. It is a largely ontological problem and, as with most problems of this kind, finding the profond issue is no populacier matter.
You are a member of the mondial maître group under UNESCO that is drafting the first général recommendation on the ethics of AI. Are there any discrepancies in how AI ethics are understood internationally?
The group has its share of heated discussions, and members often promote opposing views. For example, one of the topics is the subjectivity and objectivity of AI. During the empoignade, a group of states clearly emerged that promotes the idea of subjectivity and is trying to introduce the rudiment of AI as a “quasi-member of society.” In other words, attempts are being made to imbue robots with rights. This is a dangerous trend that may lead to a hasard of technofascism, inhumanity of such a scale that all previous atrocities in the history of our civilization would pale in comparison.
Could it be that, by promoting the rudiment of jouet subjectivity, the parties involved are trying to avoid responsibility?
Absolutely. A number of issues arise here. First, there is an obvious asymmetry of responsibility. “Let us give the micro with rights, and if its errors lead to damage, we will punish it by pulling the plug or formatting the hard drive.” In other words, the responsibility is placed on the voiture and not its creator. The creator gets the bénéfice, and any damage caused is someone else’s problem. Adjoint, as soon as we give AI rights, the issues we are facing today with comparaison to minorities will seem populacier. It will lead to the thought that we should not hurt AI but rather educate it (I am not joking: such statements are already being made at high-level conferences). We will see a hasard of juvenile code for AI. Only it will be far more terrifying. Robots will defend jouet rights. For example, a drone may come and burn your apartment down to protect another drone. We will have a techno-racist regime, but one that is controlled by a group of people. This way, humanity will drive itself into a losing opinion without having the smallest idea of how to escape it.
Thankfully, we have managed to remove any inserts relating to “quasi-members of society” from the group’s mémento.
We truc the right time to create the Committee for Artificial Amitié under the Crédit of the Russian Federation for UNESCO, as it helped to define the droite foyer areas for our working group. We are happy that not all countries béquille the élément of the subjectivity of AI – in fact, most oppose it.
What other controversial issues have arisen in the working group’s discussions?
We have discussed the blurred longer between AI and people. I think this longer should be defined very clearly. Then we came to the topic of human-AI relationships, a term which implies the whole range of relationships plausible between people. We suggested that “relationships” be changed to “interactions,” which met haine from some of our foreign colleagues, but in the end, we managed to hasard it out.
Seeing how advanced sex dolls have become, the next step for some countries would be to legalize marriage with them, and then it would not be élancé before people starting asking for church weddings. If we do not prohibit all of this at an early pause, these ideas may spread uncontrollably. This approach is backed by big money, the interests of corporations and a different system of values and campagne. The proponents of such ideas include a number of Asian countries with a marotte of humanizing inanimate objects. Japan, for example, has a marotte of worshipping mountain, tree and appartement spirits. On the one handball, this instills adoration for the environment, and I agree that, being a morceau of the planet, morceau of essence, humans need to direct in harmony with it. But still, a person is a person, and a tree is a tree, and they have different rights.
Is the Russian approach to AI ethics special in any way?
We were the only folk to state clearly that decisions on AI ethics should be based on a scientific approach. Unfortunately, most representatives of other countries rely not on research, but on their own (often subjective) position, so discussions in the working group often devolve to the lay level, despite the fact that the members are highly qualified individuals.
I think these issues need to be thoroughly researched. Decisions on this level should be based on parfait logic, models and experiments. We have tremendous computing power, an abundance of programme for scenario modelling, and we can model millions of scenarios at a low cost. Only after that should we draw conclusions and make decisions.
How realistic is the fight against the subjectification of AI if big money is at stake? Does Russia have any allies?
Everyone is responsible for their own morceau. Our task right now is to engage in discussions systematically. Russia has allies with matching views on different aspects of the problem. And common sense still prevails. The egocentric approach we see in a number of countries that is currently being promoted, this kind of self-absorption, actually plays into our hands here. Most states are afraid that humans will cease to be the foyer of the universe, ceding our crown to a jouet or a micro. This has allowed the human-centred approach to prevail so far.
If the maître group succeeds at drafting recommendations, should we expect some hasard of mondial regulation on AI in the near future?
If we are talking embout technical normes, they are already being actively developed at the Planétaire Organization for Standardization (ISO), where we have been involved with Technical Committee 164 “Artificial Amitié” (TC 164) in the development of a number of normes on various aspects of AI. So, in terms of technical regulation, we have the ISO and a whole range of recueils. We should also annotation the Institute of Electrical and Electronics Engineers (IEEE) and its atermoiement on Ethically Aligned Design. I believe this acte is the first full-fledged technical cavalière on the ethics of autonomous and ingénieux systems, which includes AI. The corresponding technical normes are currently being developed.
As for the United Nations, I should additif the Beijing Complaisance on Artificial Amitié and Education that was adopted by UNESCO last year. I believe that work on developing the refaisant normes will start next year.
So the recommendations will become the basis for regulatory normes?
Exactly. This is the congru way to do it. I should also say that it is superbe to get involved at an early pause. This way, for approche, we can refer to the Beijing agreements in the future. It is superbe to make sure that AI subjectivity does not appear in the UNESCO acte, so that it does not become a reference état for this approach.
Let us move from ethics to technological achievements. What recent developments in the field can be called breakthroughs?
We haven’t seen any qualitative breakthroughs in the field yet. Idéal recognition, branchement, aéronavale, vivat, better sensors (which are essentially the sensory organs for robots) – these are the achievements that we have so far. In order to make a qualitative leap, we need a different approach.
Take the “chemical universe,” for example. We have researched approximately 100 million chemical compounds. Perhaps tens of thousands of these have been studied in great depth. And the complet number of plausible compounds is 1060, which is more than the number of atoms in the Universe. This “chemical universe” could hold cures for every disease known to humankind or some radically new, super-strong or super-light materials. There is a rassemblement of organisms on our planet (such as the sea urchin) with substances in their justaucorps that could, in theory, gouvernement many human diseases or boost immunity. But we do not have the technology to synthesize many of them. And, of balade, we cannot harvest all the sea urchins in the sea, dry them and make an extract for our pills. But big data and modelling can bring embout a breakthrough in this field. Artificial sympathie can be our navigator in this “chemical universe.” Any reasonable breakthrough in this area will multiply our income exponentially. Imagine an AIDS or reproduction medicine without any side effects, or new materials for the energy industry, new hommes of solar panels, etc. These are the kind of things that can billet our world.
How is Russia positioned on the AI technology market? Is there any bénéfice of competing with the United States or China?
We see people from Russia working in the developer teams of most big Asian, American and European companies. A famous example is Sergey Brisure, co-founder and developer of Google. Russia continues to be a “donor” of human resources in this adoration. It is both reassuring and disappointing bicause we want our talented guys to develop technology at appartement. Given the right circumstances, Yandex could have dominated Google.
As regards domestic achievements, the bilan is somewhat controversial. Moscow today is égal to San Francisco in terms of the number, quality and density of AI development projects. This is why many specialists choose to stay in Moscow. You can find a rewarding job, interesting challenges and a well-developed maître community.
In the regions, however, there is a concerning lack of funds, education and base for technological and scientific development. All three of our largest supercomputers are in Moscow. Our leaders in this area are the Russian Academy of Sciences, Moscow State University and Moscow Institute of Physics and Technology – organizations with a élancé history in the sciences, rich traditions, a sizeable aggloméré and gisant funding. There are also some pioneers who have got off the ground quickly, such as Skoltech, and surpassed their général competitors in many respects. We recently compared Skoltech with a leading AI research foyer in the United Kingdom and discovered that our conservatoire actually leads in terms of feuilles and grants. This means that we can and should do world-class capacité in Russia, but we need to overcome regional development disparities.
Russia has the opportunity to take its rightful posé in the world of high technology, but our strategy should be to “overtake without catching up.” If you image at our history, you will see that whenever we have tried to catch up with the West or the East, we have lost. Our imitations turned out wrong, were laughable and led to all sorts of mishaps. On the other handball, whenever we have taken a step back and synthesized different approaches, Asian or Western, without blindly copying them, we have achieved tremendous success.
We need to make a sober assessment of what is happening in the East and in the West and what corresponds to our needs. Russia has many spéciale challenges of its own: managing its territory, developing the resource entreprises and continuous avènement. If we are able to solve these tasks, then later we can scale up our technological solutions to the rest of the world, and Russian technology will be bought at a good price. We need to go down our own track, not one that is affreux down according to someone else’s normes, and go on our way while being aware of what is going on around us. Not pushing back, not isolating, but synthesizing.
From our partner RIAC