Восстание машин: системы ИИ могут устроить войну, где человечество окажется расходным материалом

Academician of the Russian Academy of Sciences Igor Kalyaev expressed the opinion that the division of artificial intelligence systems created in the world according to national, gender, religious and other characteristics could lead to a war between AIs, in which humanity will only be expendable.

Image source generated by the DALL•E 3 neural network

The scientist commented on the words of American entrepreneur Elon Musk that the amount of human data available for AI training is almost exhausted, and now specialists need to switch to training neural networks on synthetic data (created by AI itself).

Kalyaev noted that training AI on synthetic data requires preliminary training that forms a kind of “mentality”. According to him, the nature of the generated data depends on those who create it, and may reflect social or cultural characteristics.

All this may lead to the fact that in the future there will be a gradual division of AI according to national, gender, religious and other characteristics. And as the entire history of mankind shows, it is this division that is the main reason for the emergence of all kinds of conflicts, up to the bloodiest ones.

Igor Kalyaev, Doctor of Technical Sciences, Professor and Academician of the Russian Academy of Sciences

Kalyaev believes that each AI will consider itself “correct”, and other AIs — “incorrect” and subject to destruction. This, in turn, may lead to wars between AIs, where humanity will become expendable with a minimal role.

He also expressed doubt about the effectiveness of training AI on synthetic data to improve its abilities in the real world. According to him, this can be compared to trying to teach a child about life exclusively through cartoons.

The academician emphasized that, despite the use of real data, AI has still not reached the level of human intelligence. The principles of the human mind remain misunderstood, but are clearly more effective.