Artificial intelligence helps engineers work on aircraft, but entrusting it with full control of a passenger liner is still too risky. This was stated by Igor Nikitin, founder of IT companies WMT AI and WMT Group, and Sergey Popov, Vice-Rector for Digitalization at the Moscow Aviation Institute.
The main problem lies in the reliability of such systems. AI must act correctly not only in normal conditions, but also in rare situations where an error can lead to severe consequences. At the same time, an engineer must be able to verify the system's decision and understand how it will behave in non-standard circumstances.
According to Nikitin, it is still too risky for civil aviation to fully transfer aircraft control to artificial intelligence. Even if the system handles most tasks well, this is not enough to confirm its safety in all possible situations.
However, AI is already being used in the early stages of aviation equipment creation. As Sergey Popov said, it can be used in preparing technical documentation and creating simplified models of complex systems at the preliminary design stage. However, neural networks can make calculation errors, refer to non-existent sources, misunderstand technical documents, or generate program code that looks correct but gives an incorrect result.
In the aviation industry, the cost of an error can be a human life. Therefore, verification of AI results is not an additional option, but a mandatory link in the engineering process. Responsibility for the decision always lies with a human.
This approach is taught to future engineers at MAI. The institute has an educational program "Artificial Intelligence in Professional Activity" jointly with WMT AI.