Computer chip in the shape of a brain. Credit: Pixabay

Artificial intelligence development is an interdisciplinary endeavor.

Maxim Goryunov Dec 11, 2021

Throughout history, humanity has not seen changes on the same scale as what we are witnessing now. The digital transformation of the Fourth Industrial Revolution is not just the introduction of new technologies, it goes deeper and involves a shift in mindset, behavior and culture. Its protagonist, artificial intelligence, is gaining momentum - it's a fait accompli. But how many are thinking about what it will lead to and what questions the development of AI poses to society? Should it be seen as an interdisciplinary component, including questions of sociology and philosophy? There is no clear answer if the current world situation represents Taleb's "Black Swan", although attention to nowadays impact is crucial.

Undoubtedly, viewing the success of digitalization can come from many angles. However, the most influential of these is automation. AI technology accompanies people in medicine, economics, education, science, retail, automotive and manufacturing. Researchers and entrepreneurs, in fact, don't have enough of it. Some are saying that the use of AI should be encouraged, thereby enhancing human capabilities. Others, and this is really another vector, about creating machines that will understand us on a psychological level. A third group scares people with a technological singularity. Going back 5 years, media headlines were increasingly focused on the coming dystopia/utopia of artificial intelligence. Consequently, the average person is faced with the question of who is to be believed. Naturally, the decision to attend a Miku Hatsune concert is a personal matter. Nevertheless, let’s pause for a moment and appreciate how far technology has come. Where is the fine line between understanding that intelligent automation is a supplement to human effort, not a replacement for it, and whether it should be seen as a panacea?

The history of artificial intelligence development and what we should expect from it

The term 'artificial intelligence' appears in many different contexts today. Since the mid-1930s, when Alan Turing published his work on devices that could solve fairly complex problems, the global scientific community has been interested in the problem of AI. At Turing's suggestion, an intelligent machine is defined as one that cannot be distinguished from a human being. Conversely, the American scientist John Searle, in his article "Mind, Brain and Programs" , which has become chrestomatically famous, turned to criticism of the theoretical claims of artificial intelligence. His "Chinese Room " experiment suggests that artificial intelligence cannot be equated with natural intelligence, as the principles of their operation are quite different. Thus, he refutes the "strong version" of artificial intelligence, which suggests that a computer can have the ability to think and comprehend on a par with a human. At the same time, Searle supports the "weak version" of artificial intelligence, according to which mental processes can be modeled by a computer program. The experiment provoked a considerable academic backlash and an intense debate which is still going on. In particular, many critics have pointed out that Searle in his experiment inappropriately identifies the person in the room with the subject of Chinese comprehension, which should, in their view, be recognized as the entire system consisting of the room, the rulebook and the person. The answer to who stands right is, at the moment, impossible to give. However, one conclusion can still be reached. To set up the human-machine relationship, we need to listen to a variety of experts who are able to agree on concepts and translate them into context. AI philosophy, being both a philosophy of science and a philosophy of technology, considers artificial intelligence as a phenomenon of society, culture and cognition, on the one hand, and tries to comprehend its nature and evaluate its impact on our reality, on the other. Therefore, not allowing our civilization to forget that the time of reflection has come.

In a practical sense, AI is just a tool, but it has a high computational power. Since the Dartmouth Conference , its potential is increasing day by day, solving a growing number of applications. The AlphaZero program, which is equally good at go, shogi and chess, with no basic knowledge, needs only 24 hours to be able to beat any person. More interestingly, AI has even mastered poker, where, unlike go, players may avoid acting without full knowledge of what their opponent knows or doesn't know, what cards they hold and what they can do. Amazingly and frighteningly, the OpenAI GPT-3 neural network writes poetry, music, and articles. Thus, full autonomy is not the ultimate goal for AI systems. The findings of the recently published interim results of the "One-Hundred-Year Study on Artificial" look at the broader areas in which AI is evolving and making a difference in people's work and lives: language processing, computer vision, mobility, robotics, healthcare, finance and decision-making. Artificial intelligence resources are evolving in value, which makes it more relevant to use them responsibly.

According to Gartner's "Hype Cycle for Artificial Intelligence, 2021" report, the AI market is in a state of evolution, with a high percentage of innovation in the early upstream stage, referred to as the "innovation trigger". Thus, growing numbers of end users are considering a solution with more AI capabilities. Meanwhile, enhancing trust, transparency, fairness and verifiability becomes progressively important for stakeholders, and responsible AI helps to achieve this. The research company expects that by 2023, all personnel hired to develop and train AI will need to have a technology expertise that ensures ethical compliance, explainability, justice and respect for privacy. This requires small and wide data approaches that provide more robust analytics, reduce reliance on big data and help achieve a 360-degree view on problems. Institutions will be forced to change their attention from big data to small and wide data by 2025, which will give more context for analytics and make AI less data-intensive.

Priority areas for implementing AI and its role in transforming the economy.

From an economic perspective, revenues from the global artificial intelligence market are expected to reach $341.8 billion in 2021 . According to a report by Grand View Research , this volume is expected to reach $997.77 billion by 2028, at a CAGR of 40.2%. Investments related to digital transformation combine writing and refining new applications, reducing the implementation time of new releases, and interacting with various systems through API interfaces. Enterprises are facing significant changes in current methodologies and practices for managing hybrid IT infrastructures. Key to these shifts is a new framework called AIOps. Previously, we hosted a webinar in which Microsoft scientists talked about how to take this approach into cloud computing, the use of AI and product strategy. Indeed, new DevOps technologies are forcing programmers not only to take more control at the application level, but also to take responsibility for the overall state of the ecosystem and the interaction between services and AI infrastructure. Artificial intelligence for IT operations allows enterprises to analyze consumer behavior and better predict people's needs, leading to a more personalized experience. For instance, medicine as a science and branch of health care is experiencing the formation process called 4P medicine : personalized, predictive (predicting the probability of diseases), preventive (focusing on the prevention of disease development), and participatory (enabling the equality of doctor and patient positions). The arrival of AI in medicine has the potential to help solve a wide range of important problems in collecting and combining large amounts of scientific knowledge and clinical practice. It will give the development of medical and pharmacy practice, improving the effectiveness of management decisions in healthcare. The Watson for Oncology module of the IBM supercomputer is used to search the database (medical reports, diagnoses and pages of texts from medical journals and clinical trials in oncology), select multiple treatment options and make the physician choose the best one. At the same time, it gives staff members more time to focus on critical issues. Programs can pick up wards, search for available equipment, keep track of medical equipment.

However, scientists, as has already been said, do not stop there. It is assumed that we will create machines that will understand us humans on a psychological level. These machines will be able to sense our inner states - emotions, attention, and character. Certainly, there are already many products and services that assess the human condition - for example, wearable devices that help monitor sleep quality. But empathy is not determined by the amount of data processed. Existing deep learning methods are inherently not well-suited for predicting things like mood and stress, which vary widely due to individual differences. To move forward, scientists suggest driving the work on the Human-Machine Understanding concept, which involves assessing emotional states, including postures, facial expressions, physiological markers, and language in combination with the unique environment to reflect each individual. This topic is not only of applied nature, but also largely determines the further development of new interdisciplinary science, technology, economy and society. It involves neurobiologists and psychologists successfully joining forces with engineers.

The regulation of ethical problems assumes a relationship between machines and morality.

This widespread use of AI reinforces the dilemmas between privacy and human autonomy. The range of legal (or legally relevant) and ethical problems determined by the development and use of artificial intelligence technologies and units is quite extensive.

From the perspective of legal regulation, it is possible to distinguish two main groups. The first includes the operation and use of artificial intelligence units in general. Their solution is mainly aimed at preventing harm to society and the individual from mass use. One primitive example: an unmanned car ran over a person. Whose fault was it? The programmer? The owner of the car or the manufacturer? Slightly less primitive, the use of AI as a judge . Preventive regulation could significantly reduce the degree of uncertainty in risk management issues related to the legal responsibility for the actions of an AI unit. The second group looks at specific (socially useful) areas of activity. It could be the privacy of personal data in the provision of medical care, or the use of artificial intelligence in the banking sector . Detailed legal regulation needs to be developed for each individual use of artificial intelligence. Given the dynamics of the development of technologies and AI units, flexible legal regulation in this group may be more effective. Public authorities may not have sufficient information at the stages of their early development to develop appropriate legal regulation. Formation of legal regulation of artificial intelligence technologies is only at the beginning, but it may evolve into full-fledged complex institutes of law and institutes of legislation, possibly even into branches and sub-branches of law and legislation.

Meanwhile, the ethics of AI is emerging as an area of research. Developments in this area have led to the emergence of a number of initiatives. The global initiative of the Institute of Electrical and Electronics Engineers (IEEE) is a particularly noteworthy one. Work has begun on the creation of regulatory and technical documents capable of laying the foundation for ethical patterns of "behavior" of AI systems. Studies in this area are leading to the creation of glossaries . Their principles can be generally described as: the technology must not violate privacy, must not discriminate on any grounds, must not cause harm to its creator, must be fully controlled by him, and must not violate other human rights. An even more complex issue is related to the social adaptation of AI, the division of human and AI roles in the real sectors of the economy and social sphere, the mutual intellectual influence of human and artificial intelligence, which may lead not only to mutual enrichment, but also to new risks and challenges for future generations. Additionally, the document Rome Call for AI Ethics , signed in Rome by the head of the Roman Catholic Church, FAO (Food and Agriculture Organization of the United Nations), the Italian government, representatives of IBM, and Microsoft, was published in 2020. The goal is to take a holistic and comprehensive approach to the technological future of humanity.

Despite the serious challenges of implementing AI systems, the prospect of using them encourages the search for solutions to overcome the obstacles. Highly qualified specialists from around the world are constantly working on the development of this field. They are trying to understand how artificial intelligence can affect automation systems, national security, ethics, legislation, and the privacy of citizens. Indeed, artificial intelligence plays a key role in the future of technological development and we are forced to choose between great possibilities and great dangers at some point. The main conclusion is to consider artificial intelligence as an interdisciplinary field, encompassing linguistics, philosophy, neuroscience and psychology. In this way, it is possible to abandon the fantastic fear of "machine uprising" and build a utilitarian useful relationship between AI and humans.