Elon Musk’s AI Warning Sparks Debate

After the UK artificial intelligence (AI) summit on Nov. 2, Inflection AI CEO and DeepMind co-founder Mustafa Suleyman criticized Elon Musk in a BBC interview.

Musk used his trademark sensationalism in an interview with U.K. Prime Minister Rishi Sunak after the two-day event. Musk called AI “a magic genie,” but warned, “Usually those stories don’t end well.”

The richest person in the world warns that AI will someday do almost every profession, which will make it hard for humans to find meaning.

Musk also emphasized AI’s existential threats, including the need for a “physical off-switch” so humans can govern the machines.

Sunak concurred with Musk that Hollywood AI scenarios like The Terminator seemed to shape both men’s views on the technology.

“All these movies with the same plot fundamentally all end with the person turning it off,” joked Sunak. What technology the two men meant is unknown.

Distributed and cloud computing and server technologies make most AI systems produced in the previous decade resistant to “turning it off” with a single physical switch.

Suleyman, who attended the U.K. AI Summit, called Musk’s comments banal in a BBC Question Time interview.

“This is why we need an impartial, independent assessment of the trajectory of this technology. [Elon Musk is] not an AI scientist. He owns a small AI company. He has many other companies. His expertise is more in space and cars.”

In an interview, Suleyman stated that he is not the first AI specialist or CEO to challenge Musk’s scientific understanding of AI.

Gary Marcus, a New York University computer science professor and best-selling author, and Vivek Wadha, a Carnegie Mellon and Harvard distinguished fellow, disputed Musk’s 2029 AGI prediction in 2022.

The two experts placed a bet of $500,000 on Musk completing AGI before 2029.

AGI is undefined and has no benchmarks or measurement standards. The assumption is that AI technology will eventually be able to do any intelligent work owing to unforeseeable technical impacts.

Some believe AGI, or sentient AI, may exist, while others say present systems are too dependent on training, programming, processes, and guardrails to be as clever or capable as humans or other animals.