[AI Journey track] Engineering AI. Scientific foundations for industrial applications

In the not too distant past symbolic AI was the dominant paradigm of AI research. However, with the advent of statistical AI methods, which have now resulted in the discovery of deep neural networks approaches (Neural AI), the scientific community has largely forgotten about the concept of symbolic AI. Nowadays, modern LLMs is a new "engine" capable of "breathing life" into symbolic AI. It is becoming obvious that merging of these two concepts is a necessary step for the further development. In my talk I will present a new concept of Engineering AI. It is not yet AGI, but it is no longer a solution to highly specialised tasks by a standard Artificial Narrow Itelligence, which is mainly based on Neural AI nowadays. In a nutshell Engineering AI is a multi-agent system, in which the symbolic AI is used to orchestrate the process of solving engineering problems, extracting relevant domain-specific knowledge for this, and the Neural AI is a set of tools to generate the corresponding computational workflow. I will describe the concept of Engineering AI in detail, provide some recent scientific results on its foundations, and present several relevant applied use cases, one of which have been already successfully realised in production.