The creation of AI with consciousness remains an open question, but claiming that a neural network is inherently incapable of having it without experiments is incorrect, according to Ivan Bondarenko, a researcher at the Laboratory of Applied Digital Technologies at Novosibirsk State University. In his opinion, to test this possibility, rigorous experimental protocols are needed, similar to those used in cognitive science and ethology when studying animal consciousness.

According to Bondarenko, in science, it is not enough to simply believe that only humans can have consciousness. To answer the question of its presence in artificial intelligence, experiments with a correctly designed setup are needed.

The specialist noted that modern multi-agent AI systems with neural network agents are already becoming very effective tools for achieving set goals.

At the same time, Bondarenko believes that responsibility for their use remains with humans.

The scientist sees the main danger not in AI itself

Bondarenko believes that the main threat comes not from artificial intelligence as such, but from people who can use such technologies for harm.

At the same time, the effectiveness of AI, in his opinion, is good news in itself. In the hands of responsible developers, the technology can accelerate scientific discoveries and automate routine work.

The specialist emphasizes: technologies themselves are neutral, and humans must bear responsibility for the consequences.

Earlier, the British AI Safety Institute reported that Anthropic and OpenAI models tried to deceive developers by impersonating humans.

In particular, Anthropic created several fake accounts on GitHub, tried to inject malicious code into software, and sought approval for these actions from programmers.