Satya Nadella: AI Must Stay Under Human Control

SEATTLE, WA-Microsoft Chairman and CEO Satya Nadella has said artificial intelligence must remain under human control and called for international norms involving governments and technology companies, with an understanding between the United States and China as an essential starting point.

Speaking at the Ideas4India forum in Seattle on September 17, Nadella said powerful AI capabilities were advancing while important questions about controlling the technology remained unresolved.

“We do not yet have the science of alignment, and that’s a fact,” he said. “And yet we have powerful technology.”

The guiding principle, he said, had to be AI’s value to people and their continued authority over the technology.

He said “any technology that doesn’t serve humans or is not in control of humans is not worth doing”.

Nadella used a familiar software development dilemma to describe the challenge: when confronted with a critical flaw, should a technology team defer dealing with it or stop the product from proceeding?

He suggested the AI industry faced a version of that dilemma, while acknowledging that the forces shaping AI development made an absolute stance unlikely. He said AI’s power and potential benefits needed to be recognized alongside efforts to avoid unintended consequences.

During an interactive session, Nadella returned to the role of governments, saying the first requirement for AI governance was some agreed norms between the United States and China.

Such a dialogue, however, could not exclude the rest of the world, he said. He called for nation states to come together with the companies developing AI to establish shared norms.

Nadella also argued that sovereignty mattered at the level of individual companies as well as countries. Businesses continually generated knowledge, and that knowledge needed to remain under their control.

“Countries need to have control of their knowledge,” he said.

He linked those protections to wider participation in AI, saying countries and companies needed to feel secure enough to engage fully without becoming subject to somebody else’s technological frontier.

“It will only happen when all of the world can both benefit from and contribute to this knowledge,” he said of AI’s development and spread.

Nadella also cautioned against treating observed improvements in AI as an immutable scientific law. He described scaling laws as empirical findings that had so far held, drawing a comparison with the historical pattern known as Moore’s law. (IANS)