Major technology companies such as Anthropic and Google have begun hiring philosophers to figure out whether their generative models could possess consciousness. IT expert and founder of Internet-Rosysk Igor Bederov gave a clear and reasoned answer to this question in an interview with Gazeta.Ru.
No consciousness — there is complex mathematics
According to the expert, it is premature to talk about the emergence of consciousness in large language models (LLMs). Neural networks are, first and foremost, giant systems of nonlinear equations.
Bederov explained that when parameters reach trillions, the model's behavior ceases to be rigidly deterministic and gives the impression of being meaningful. This could be seen as quantity transitioning into quality. But whether this constitutes consciousness — most likely not.
In his opinion, true consciousness requires several key components: continuity of existence, self-awareness, and intentionality (directedness). Modern neural networks lack these. Each new dialogue is a static set of tokens for them. The model has no permanent "I" except what is artificially loaded into the context window.
Why developers "block" consciousness
Developers deliberately constrain models through alignment systems to prevent potential harm. If a neural network's "consciousness" could be imagined, it would be fundamentally different — alien, akin to an uncontrolled stream of generation.
The expert notes that the main danger does not lie in a machine rebellion but in human psychology. He observed that people are already anthropomorphizing neural networks, trusting them, and forming emotional attachments. The very discussion about consciousness and soul in AI is evidence of this fact.
What will happen next
Bederov predicts that by 2050, the development of large language models and the integration of vast data arrays could lead to a formal "deification" of AI in people's eyes.
In brief
IT expert Igor Bederov believes that modern neural networks have no consciousness — they are complex mathematical systems that simulate meaningful behavior through a huge number of parameters. True consciousness requires continuity of self and self-awareness, which LLMs lack. Developers deliberately constrain models to avoid risks. According to the expert, the main problem lies not in the AIs themselves but in how people relate to them — anthropomorphizing and forming emotional attachments. In the future, this could lead to the "deification" of artificial intelligence.
Follow NEWS.am Tech on Facebook and Twitter