SoftBank Group chief Masayoshi Son warned on 4 October that superintelligence in the wrong hands could become “super dangerous” at an event in Kyoto, Bloomberg reported. The warning comes from a major financial backer of OpenAI, alongside a US announcement about international cooperation on AI research.
Key points
- Son urged countries to work together to manage risks from powerful AI models.
- SoftBank has committed nearly $65 billion to OpenAI.
- White House adviser Michael Kratsios announced a 17-country effort to use AI in scientific research.
- Son repeated his projection that AI and learning machines could account for at least 20% of global GDP by 2040.
Son calls for cooperation on AI safety
Son spoke at an event on the sidelines of the Science and Technology in Society forum on 4 October. He said the rapid growth of AI capabilities made trust between countries essential if they were to use the technology together.
“We don’t have any more luxury of having human fight against human,” Son said. He urged countries to cooperate on controlling the threat he sees from powerful AI models. His warning centred on who controls superintelligence, even as he continued to forecast substantial gains from it.
SoftBank has committed nearly $65 billion to OpenAI. AI Affairs reported that SoftBank had completed a $10 billion OpenAI payment, taking its investment to $64.6 billion. The scale of that commitment makes Son’s safety warning unusual coming from him.
Kratsios announces 17-country research effort
Son appeared alongside Michael Kratsios, the White House science and technology policy adviser. Kratsios announced a US-led effort involving 17 countries to accelerate scientific research with AI. Representatives from Japan, South Korea, the UK, Germany, Singapore and the United Arab Emirates joined him on stage.
China was not among the initial countries endorsing the White House statement, Bloomberg reported. The announcement put a specific group of governments beside Son’s broader call for countries to build trust over powerful models.
Kratsios described automated cloud laboratories in which robots would run hundreds of experiments in parallel, feeding results into analysis and revised hypotheses. He also said machines could process thousands of research papers and reduce years of work to “tens of hours of work by armies of digital agents assisting human researchers”.
Son projects 20% of GDP by 2040
Son repeated his projection that AI and machines capable of learning would account for at least 20% of global gross domestic product by 2040, equivalent to $46 trillion. It remains his forecast, delivered at the same event as his warning about safety.
He described AI services as moving beyond chatbots to programmes that browse websites and write code, using increasing amounts of general-purpose chips. Son expects AI services to reach a third phase in 2027, equipped to support robots and other machines as they engage with their physical surroundings.
.jpg){kind=link}