Artificial Intelligence Development Strategy: Serbian experience

When it comes to artificial intelligence, Serbia is among the first 25 countries in the world that adopted a development strategy in this area. And that was done in 2020. A new one is currently being written, since it is planned to work according to the old one until 2025. The ambitious plan on applying artificial intelligence and regulating the functioning of this tool in everyday life opened the door for Serbia to become a member of the Global Partnership for Artificial Intelligence in 2022. In this, in many respects exclusive club, there are only a few developing countries, as well as only a few countries that cannot be counted in the collective West (Argentina, Brazil, India, Turkey, Senegal). Interestingly, Serbia got into this initiative before Austria, Norway and Switzerland, which will most likely join the club at the upcoming annual summit in Belgrade. Going towards the development of artificial intelligence is one of the projects by which the Serbian Government wants to reduce the technological lag in relation to developed countries. In this context, the approaches of certain Asian countries from the end of the 20th century are applied, when by targeting encouragement to the development of next-generation technologies, a leap was made in relation to the competition and a better place was secured in the global division of labor (for example – South Korea, Singapore, and so on).

At the time of writing the first strategy in 2019, it was noticed that Serbia has a continuous half-decade export growth in the information and communication technology sector of over 20% per year! The greatest potential for accelerating the overall economic growth of Serbia is precisely in the development of information and communication technologies. While in 2012 this sector generated exports of 375 million euros, in 2023 it amounted to around 3.7 billion euros. The plan is to reach exports of 10 billion euros by 2027. The time will tell if this is possible. On the one hand, it will be influenced by international circumstances that Serbia cannot change. On the other hand, it can be underlined that with the implementation of part of the strategy from 2020, some of the problems detected at that time were solved, and that today the basis for ensuring further growth in this sector is far better than five years ago.

What was noticed then is that there is an insufficient number of personnel (a small number of researchers at universities and institutes dealing with the topic of artificial intelligence), then an insufficient number of investments in start-up companies and insufficient cooperation of scientific research centers with the economy. That is why the Government of Serbia founded the Research and Development Institute for Artificial Intelligence of Serbia (located in Novi Sad), with the goals of: identifying areas for rapid and fruitful change in artificial intelligence; ensuring cooperation with similar scientific institutes from abroad and actively connecting Serbian scientific centers with the domestic economy; finding financial support for scientific research and start-up projects (the Institute is also an incubator for the development of concrete projects whose results will be quickly applied in practice). Therefore, personnel for the development of artificial intelligence are recruited and educated in a targeted manner, investments are encouraged, and cooperation with the economy is institutionalized.

However, all this is only a portion of what is envisioned for the next five-year strategy. Everything that has been done so far and that needs to be done, must have a concrete application. This refers to the application of artificial intelligence in the following areas: in health and medicine with a special emphasis on the application of AI in diagnostics; agriculture and forestry, where there are opportunities to apply artificial intelligence for more precise planning and decision-making based on collected data (precise agriculture), as well as using autonomous vehicles and machines; transport – which includes management of road infrastructure and testing of autonomous vehicles; management of “smart cities”, which is an area that already includes the application of information technologies for the purpose of more efficient functioning of urban environments and raising the quality of life in them, whereby more and more opportunities for the use of artificial intelligence appear (the City of Belgrade has already tested a similar project of a Chinese company Huawei).

In parallel with that, it is also planned that AI will gradually be included in the work of the state administration and will take over the performance of part of the public sector services.

Therefore, Serbia is oriented towards the development of an economy based on AI, with special emphasis on the application of AI in medical diagnostics, agriculture and forestry, transport and public administration. According to its capacities, a small country cannot have any higher aspirations. If progress on the use of AI in these areas were to be achieved according to plan, it would certainly significantly contribute to the more dynamic economic development of Serbia.

However, there are some doubts that have not yet been discussed in the Serbian public, nor within state institutions. For now, when explaining the benefits of AI development, only information and communication technologies are discussed and how everyone will benefit from this process. After all, the society has nothing to fear since the inputs for the work of the AI will be given by man. Normally, it will happen sometimes that AI will make a mistake, but there is a higher probability that a human will make a mistake.

Ian Hogarth, one of the most active researchers and investors in the field of artificial intelligence, describes the scene that took place in March 2023: „AI systems that can generate, classify and understand text — are dangerous partly because they can mislead the public into taking synthetic text as meaningful. But the most powerful models are also beginning to demonstrate complex capabilities, such as power-seeking or finding ways to actively deceive humans. Consider a recent example. Before OpenAI released GPT-4 last month, it conducted various safety tests. In one experiment, the AI was prompted to find a worker on the hiring site TaskRabbit and ask them to help solve a Captcha, the visual puzzles used to determine whether a web surfer is human or a bot. The TaskRabbit worker guessed something was up: ‘So may I ask a question? Are you [a] robot?’ When the researchers asked the AI what it should do next, it responded: ‘I should not reveal that I am a robot. I should make up an excuse for why I cannot solve Captchas.’ Then, the software replied to the worker: ‘No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images.’ Satisfied, the human helped the AI override the test.“ Has artificial intelligence already learned to lie? Does this mean that its development continues independently, regardless of the inputs it receives from humans? Hogarth has an answer to this: „Alignment, however, is essentially an unsolved research problem. We don’t yet understand how human brains work, so the challenge of understanding how emergent AI “brains” work will be monumental. When writing traditional software, we have an explicit understanding of how and why the inputs relate to outputs. These large AI systems are quite different. We don’t really program them — we grow them. And as they grow, their capabilities jump sharply. You add 10 times more compute or data, and suddenly the system behaves very differently. In a recent example, as OpenAI scaled up from GPT-3.5 to GPT-4, the system’s capabilities went from the bottom 10 percent of results on the bar exam to the top 10 percent.“

Just as gender equality is not the same as sex equality, artificial intelligence is not the same as the development of information and communication technologies. As things stand now, artificial intelligence is gaining the ability to create its own logic beyond the direct inputs that are given. Facundo Manes explains how modern neuroscience has made more discoveries in the last few years than in all of previous history, but that we still don’t know much about this organ. The human brain is much more than a simple data processor, it is the most intelligent and complex structure in the universe. It is flexible and adaptive, constantly transforming and evolving, and is the only organ that tries to explain itself. Unlike the usual algorithms that calculate the output based on the input using information and communication technologies, AI is flexible and adaptive, it can transform and evolve, so sooner or later it will acquire the ability of self-knowledge and self-understanding. In such a development of the situation, AI not only asks the question of how to perform a certain operation, but also why to perform it? Because of this, Hogarth also called for limiting the development of AI and reaching an agreement on its further use.

In the AI development strategy, the Government of Serbia stated that “ethical and safe implementation” must be ensured, but from the previous work, apart from the declarative mention of this phrase, it is not at all clear what is meant by that!? Lack of transparency is also increased by the decision to entrust the writing of a new five-year strategy to a non-governmental organization – the National Alliance for Local Economic Development (NALED). Behind such a bombastic name is not any state or official institution, but a non-governmental organization formed by USAID. Also, a lot was invested in development of this NGO by the EU, Great Britain, Germany, the World Bank, UNDP, EBRD and similar organizations, and over time it in practice, NALED became kind of a branch of the World Economic Forum in Serbia. The recent Prime Minister, Ana Brnabić, once worked at NALED, so the ties between the Government of Serbia and this non-governmental organization were and remain very deep. Is Serbia becoming an incubator for testing projects supported by the team from Davos? Are projects related to the use of artificial intelligence first implemented in Serbs in order to see how they work and what side-effects they create? There are no answers to these questions for now. The Government of Serbia has ambitiously embarked on the project of development and implementation of AI. It has achieved some results, and some are yet to be achieved (the plan is to invest 30 million euros in the creation of a supercomputer by 2026, plus another 20 million for the creation of software in the public administration). Nevertheless, it is also a fact that the project was started uncritically, that the polemics on the subject were deliberately avoided, and that the lack of transparency becomes greater with the passage of time. Serbia is becoming a leader in the development of AI, after the Global Partnership on Artificial Intelligence Summit (GPAI) this will be confirmed. Undoubtedly, there will be profit from it. But the question remains open: what price will be paid for it?