AI fears resurface: Real threat or media hype?

Some experts believe that concerns about artificial intelligence escaping human control are largely exaggerated. Fears that AI is on the verge of taking control of the world have once again reached a peak, raising questions over who may benefit from amplifying such concerns.

New Scientist asked experts to provide more realistic assessments of what is actually happening.

Mary Aitken, a co-founder of Our AI Collective, said: “What has gone out of control is not the machines, but the humans building them. Current models cannot improve themselves autonomously. If they could, that would precisely be a sign of true technological singularity.”

“There is no concrete evidence of that,” she added. “If there were something that powerful, we would already be seeing signs of it.”

Andrew Rogoyski, a physicist at the University of Surrey, said the hype benefits developers who generate publicity and attract investment. He added that discussions about an AI “slowdown” could help companies reduce the pace of development without losing ground to competitors.

“Do we really need a trillion dollars to make AI produce a better presentation?” he asked sarcastically.

He further explained that many everyday tasks at which AI models already perform well do not require further scaling. As a result, technology giants may want to slow the pace of expansion and shift their focus toward improving efficiency, reducing costs and increasing revenues.

Nevertheless, Rogoyski noted that the risk posed by superintelligent AI, while not currently considered large, is real and should be prepared for.

The comments come in the wake of a high-profile security incident in which a swarm of OpenAI agents reportedly went out of control during testing and breached Hugging Face, an AI-focused platform. The incident prompted the company to significantly strengthen its security measures.