AI's superintelligence dilemma grows

Celebrities and scientists urge a halt as the world grows uneasy with its own creation

SAN FRANCISCO/ PARIS:

More than 700 scientists, political figures and celebrities including Prince Harry, Richard Branson and Steve Bannon on Wednesday called for an end to the development of artificial intelligence capable of outsmarting humans.

The appeal, published by the US-based Future of Life Institute, came as new European research found AI systems unreliable for news - even as OpenAI unveiled a new AI-powered browser directly challenging Google's dominance.

The appeal urged a global prohibition on developing superintelligent AI until the technology is proven safe, controllable, and supported by the public. Among the signatories were Geoffrey Hinton, the 'Godfather of AI' and 2024 Nobel laureate in Physics; University of California Professor Stuart Russell; and Yoshua Bengio of the University of Montreal.

Other supporters included Virgin Group founder Richard Branson, Apple co-founder Steve Wozniak, former White House adviser Steve Bannon, and Susan Rice, who served as national security adviser under Barack Obama. The Vatican's AI expert Paolo Benanti, Prince Harry and Meghan, and US musician will.i.am also endorsed the call.

The statement coincided with OpenAI's launch of ChatGPT Atlas, an artificial intelligence-driven web browser that allows users to summarise pages, automate tasks, and even interact with websites on their behalf - marking a direct challenge to Google Chrome.

Built around ChatGPT's conversational engine, Atlas enables users to perform complex online tasks such as comparing products, analysing data, or completing purchases through an "agent mode" that executes commands end-to-end.

In a live demo, OpenAI developers showed the browser locating a recipe and automatically purchasing the required ingredients through Instacart. The Atlas browser, currently available on Apple's macOS, will later be released for Windows, iOS, and Android.

The launch follows OpenAI's rapid expansion into consumer tools since the debut of ChatGPT in 2022, as it competes with Google and Anthropic for dominance in the AI sector. Analysts said the move could reshape competition for digital advertising.

"Integrating chat into a browser is a precursor for OpenAI starting to sell ads," said Gil Luria, analyst at DA Davidson. "Once it does, that could take away a significant part of search advertising share from Google, which currently holds around 90% of that spend category."

Despite mounting competition, Chrome remains the world's leading browser with a 71.9% market share as of September, according to StatCounter. A recent court ruling in Google's favour also allowed it to continue promoting its search engine through partner payments, a practice the judge said was less problematic now that AI-powered search is transforming the landscape. Meanwhile, a major study by the European Broadcasting Union (EBU) underscored the continuing unreliability of AI systems as sources of news. The report found that ChatGPT, Microsoft's Copilot, Google's Gemini, and Perplexity collectively made factual errors in nearly half their responses to current affairs questions.

Of the four assistants, Gemini performed worst, with serious issues in 76% of its answers, including wrongly naming Pope Francis as still alive when he had been succeeded by Leo XIV and mistaking a satirical story about Elon Musk for a factual report.

"AI assistants are still not a reliable way to access and consume news," said Jean Philip De Tender, deputy director general of the EBU, and Pete Archer, head of AI at the BBC. Despite such shortcomings, a Reuters Institute study in June found that 15% of people under 25 already use AI tools weekly to get news summaries.

This increasing dependence on technology that remains error-prone has intensified global concern - now echoed by scientists and public figures - that the race toward ever-smarter AI systems risks outpacing both safety and public trust.