Expert said in future AI could become greater threat than nuclear weapons
Artificial intelligence (AI) technologies may, in the next stages of development, eclipse nuclear weapons in terms of danger due to the risk of a person losing the ability to disable such systems in the field of critical infrastructure – as noted by the Director of the National Centre for the Development of Artificial Intelligence under the Government of the Russian Federation, Sergei Nakvasin, TASS reports
In early November, representatives of 28 countries participating in the first global summit on the risks associated with artificial intelligence held in the UK, including the USA, China and India, agreed on the Bletchley Declaration, devoted to the development of a neural network. Previously, The Daily Telegraph wrote that the British Cabinet plans to create in London the headquarters of the first international AI regulator, which will be modeled on the International Atomic Energy Agency (IAEA).
“International regulation mechanisms will actively develop in the coming years; there is a need for this. Comparing nuclear energy and AI, this metaphor itself seems to me not without meaning. I think that AI at a certain stage of development can be worse than nuclear weapons or nuclear energy. The difference is that nuclear energy is presented in the form of separate infrastructure facilities with corresponding safety contours, and AI claims to be the subject acting at this stage as a source of recommendations,” Nakvasin explained.
He also said that the ethical standards of the neural network envisage the ‘red button’ principle, according to which the system must stop working on first demand.
If in a similar system that manages critical infrastructure or a nuclear power plant, e.g., this mechanism does not work, then AI can turn from a decision support system into a decision making system. This will not be a monkey who was given a grenade, but a monkey who was enraged and ran first to the grenades and then to the guns,” the expert believes. However, he is confident that this risk may become relevant in future stages of the development of artificial intelligence in about a hundred years.
“Now AI is a safe kitten,” he assured.
