SoftBank CEO and former OpenAI chief scientist share the same exciting but scary dream

Getty Images

Synopsis

Humans haven’t yet seen the full potential of Gen AI. But big guns with money and know-how are betting big on creating something far more powerful – hyper intelligent AI.

Artificial super intelligence (ASI) is said to be the next – and final – frontier of AI technology, going even beyond artificial general intelligence or AGI which AI major OpenAI aims to create. But what is ASI, which the likes of Softbank CEO Masayoshi Son and OpenAI co-founder and former chief scientist Ilya Sutskever (who left the AI pioneer last month) have been talking about, and what would it mean for humanity? ET Prime explains. What is

Artificial super intelligence (ASI) is said to be the next – and final – frontier of AI technology, going even beyond artificial general intelligence or AGI which AI major OpenAI aims to create. But what is ASI, which the likes of Softbank CEO Masayoshi Son and OpenAI co-founder and former chief scientist Ilya Sutskever (who left the AI pioneer last month) have been talking about, and what would it mean for humanity? ET Prime explains. What is ASI? ASI is an AI system with cognitive performance exceeding that of the human mind across all domains of interest, whether scientific creativity, general wisdom, or social skills.“In essence, an ASI would be an inexhaustible, hyper-intelligent super-being. A nearly perfect supercomputer available 24/7, with the ability to process and analyse any amount of data with speed and precision that we can’t yet comprehend,” an IBM blog summarises. According to Softbank CEO Masayoshi Son, ASI would be smarter than any human genius by a factor of 10,000, while AGI – AI which can match or surpass human intelligence in various cognitive aspects and tasks – is smarter by only one to 10 times. He said that the future could see ASI models interacting with each other like neurons in a human brain. Does ASI exist currently? Thus far, ASI remains hypothetical as the technology has not progressed that far. Creating AGI would be a step towards achieving ASI. While OpenAI’s charter says it seeks to build AGI, Son said last week that Softbank Group’s mission was to help humanity’s progress by realising ASI. “SoftBank Group has done many things until now that have all been a warmup for my great dream to realise artificial super intelligence,” Son told shareholders at the group’s annual general meeting. OpenAI co-founder Ilya Sutskever made a high-profile exit from the company last month to begin a new venture, Safe Superintelligence Inc, whose focus is to create ASI.How far in the future can we see ASI? “Superintelligence is within reach,” Sutskever-led Safe Superintelligence Inc’s website says. It adds that building safe superintelligence is the most important technical problem of our time.Softbank’s Son said last week that we could see ASI in 10 years. Its subsidiary Arm Holdings, the British chip designer, plans to launch AI chips by 2025. Others say the huge compute and energy requirements for ASI would limit how soon we see it. It may be significant to keep in mind that developments in quantum computing are believed to be around the corner, and that should help speed the research up.How significant would ASI be? OpenAI in a blog says that superintelligence could arrive in a decade, adding, “Superintelligence will be the most impactful technology humanity has ever invented.” Nick Bostrom, an expert on superintelligence, has said that ASI is the last invention that humans will ever need to make, because then future inventions can be more efficiently done by them. He compares the coming of ASI not to tech breakthroughs like the advent of mobile internet or blockchain but rather the emergence of homo sapiens and life on earth itself. Experts caution however that superintelligence can pose existential risks to humanity and may go out of human control.It could, for instance, manipulate systems or even gain control of advanced weapons. How humanity keeps control of ASI will depend on how it gets to grip with AI in the first place.