Elon Musk warns AI bots will swamp social media

Mr Sunak said his concern was that elections would be held in a host of key countries next year, including India, Indonesia, the US and Britain, and the challenge of bot-driven misinformation and disinformation had to be tackled beforehand.

“This is probably the first set of elections where this is a real issue. Figuring out how we manage that is mission-critical for people who want the integrity of our democracy,” Mr Sunak said.

A ‘future of abundance’

During the conversation, Mr Musk departed from his usual dire warnings about the rise of AI, to paint a utopian alternative where humans did not have to work, incomes were high, and AI served as teachers, customers service workers and even friends.

“If you have an AI that has memory, and remembers all of your interactions, and you can give it permission to read anything you’ve ever done, it will really know you better than anyone, perhaps even yourself,” he said.

“And where you can talk to it every day and those conversations multiply each other, you really will have a great friend.”

But he worried about AI-powered humanoid robots, which – unlike a Tesla – could chase people upstairs or trees, requiring kill switches.

“If a robot can follow you anywhere, what if they just one day get a software update and they’re not so friendly any more? Then we’ve got a James Cameron movie on,” he said.

He proposed “a local sort of off-switch, where you perhaps say a key word or something and then that puts the robot into a safe state”.

For Mr Musk, AI was like a “magic genie” that could grant every wish – but stories about magic genies rarely had a happy ending.

“One of the challenges in the future will be: how do we find meaning in life, if we have a magic genie that can do anything we want?”

The Sunak-Musk conversation came at the conclusion of a two-day AI Safety Summit, attended by almost 30 governments and a host of tech executives, at the World War II code-breaking centre of Bletchley Park, north of London.

Governments step up AI activism

On the first day, the governments issued a “Bletchley Declaration”, which appeared to overcome their contrasting approaches to regulating AI. Countries including the US, China and the European Union agreed to manage the risks of cutting-edge AI in a collaborative fashion, so the tech would respect human rights but regulation would not stifle innovation.

On the second day, Thursday, they fleshed out their co-operative program by agreeing to work individually and collectively on building governmental capability to vet cutting-edge AI tech for its safety and risks.

The governments said they would look to develop shared standards, methodologies and capacities to test the safety of so-called “frontier AI models” that are proposed by tech companies.

They also agreed to produce a scientific study of “the latest, cutting-edge, research on the risks and capabilities of frontier AI models”, to inform the two AI Safety Summits that will be held next year, in Korea and France.

A team of AI academics and expert advisers will conduct the study, led by Canadian computer scientist Yoshua Bengio and supported by a secretariat of British government officials.

Industry and Science Minister Ed Husic said the summit had delivered a “seismic shift” that would help countries “get the balance right” between protecting people from harm while encouraging innovation.

“You’ve got the companies and civil society that help test and push some of the thinking, plus governments being able to set up a framework of regulations to give people assurance and comfort and, importantly, trust about how technology works,” he told reporters.