Today, artificial intelligence technologies have become an integral part of our daily lives, becoming indispensable assistants in solving a multitude of tasks — from effectively supporting our efforts in performing routine tasks, processing and analysing large amounts of unstructured information, to helping us realise our creative potential and develop our creative skills. And in industries such as healthcare, autonomous transport and public safety, AI technologies help to make better and faster decisions when every second counts and those seconds can save lives! Most existing solutions are examples of highly specialised artificial intelligence technologies that require human configuration and verification. In order to solve diverse and complex problems as well as humans do, machines must learn to build causal models of their environment and navigate different contexts, rather than simply maximising success in solving a narrow task.
They must understand the physical, psychological, and other laws of our world and be able to connect new information into a big picture with what they already know. To achieve this, we need to overcome the next technological frontier: the creation of Artificial General Intelligence, or AGI. Reading various books on artificial intelligence and assessing the retrospective of scientific achievements in the field of artificial intelligence technology, the ‘explosive growth’ of technology in recent years has been particularly impressive. made possible in large part by the exponential growth of available computing power and the removal of technological limitations on the accumulation of large amounts of data for training algorithms. All this has allowed long-established architectures of multilayer neural networks to successfully solve problems in a wide variety of human activities.
At the same time, despite the special attention paid to neural networks and deep learning, this is far from the only path to General Artificial Intelligence. Everyone probably wants to understand what approaches can be used to create General Artificial Intelligence and what form it is likely to take in various areas of application. Based on the experience of developing modern machine learning technologies, we can see that the path from idea to industrial solution may be much shorter than it seems at first. Today, it is already clear that any strategy chosen by them for the development of General Artificial Intelligence and well-organised management efforts will inevitably face a key question: how to choose the correct criteria for classifying a particular study as belonging to the field of General Artificial Intelligence.
To do this, it is important for all of us to have a good understanding of the structure of the field being studied, to establish a unified and consistent theoretical basis, and then to identify the most promising approaches to the development of technologies and their subsequent industrialisation. We all believe that only by relying on effective cooperation between specialists in various fields — from deep learning and probabilistic programming to robotics and cognitive science — and by supporting interdisciplinary research, can we achieve the first tangible applied results in the development of General Artificial Intelligence. An extremely important factor in the creation and development of General Artificial Intelligence technologies is ensuring end-to-end goal setting between applied (or industrial) tasks, fundamental research, and the education system—the so-called Practice-Education-Research triad. We all also consider it important to pay special attention to the professional community on the topic of compliance with the principles of ethical application of modern technologies.
In particular, it is necessary to develop uniform and universally recognised standards at the international and inter-industry levels to ensure the safe and socially beneficial application of General Artificial Intelligence technologies. In this context, appropriate efforts should be made to ensure the stability and interpretability of the algorithms underlying General Artificial Intelligence technologies.