On Thursday on BBC Today, Suleyman called the approach of the American AI company Anthropic, which, according to him, treats AI as if it were human, “mistaken”. Earlier this week, he published an essay in which he nevertheless called Anthropic head Dario Amodei and his team “intellectually honest people”.
In the same essay, he wrote that AI does not have consciousness, does not feel, does not experience, and does not suffer, and also does not have innate preferences or its own motives. According to him, such systems are “sequence completion systems”, internally empty and created to follow instructions and carry out goals set by people.
Suleyman warned that companies should not create AI systems capable of independently setting goals, making money, and owning assets, since this could mean “in effect creating a new silicon species” that would compete with humans for resources. He said that people have every right to be concerned about AI, but there are quite practical ways to keep it under control.
The head of Microsoft AI also called for greater transparency in how AI systems are trained and evaluated, as well as independent verification of their behavior and stronger control tools. Wendy Hall, a professor of computer science at the University of Southampton, UK, called this exactly the conversation needed at the international level and contrasted it with the “hysteria” of some AI companies.
In a conversation with the BBC about the upcoming AI summit, Suleyman said that future technologies must be “subordinate” to humanity. He added that the question of how much autonomy can be given to AI systems is increasingly sharply dividing the industry.
Sources
