Masayoshi Son is one of the most optimistic VCs in technology, and he has a similarly optimistic prediction for AI.
The SoftBank CEO, known for his bold bets and long-term vision, recently laid out a roadmap suggesting artificial intelligence could become a billion times smarter within just a few years. This comes at a time when AI is rapidly evolving, with models like GPT-4 showcasing impressive, albeit sometimes flawed, capabilities. Son’s vision, however, leaps beyond these current advancements, promising an exponential increase in AI’s cognitive power.
His prediction hinges on project Stargate through which Softbank along with a couple of other partners has pledged to plough in as much as $500 billion into building infrastructure for OpenAI. “Stargate will provide more chips, more compute. So how much more compute, how much more intelligence would there be? Stargate will provide at least 10x the number of chips per year,” he says.
He continues, building on this foundation of increased chip production: “Each chip will be about 10 times more compute-capable per chip. So 10 times 10… and then each model, infused by this opening AI, becomes exponentially more capable. So 10 times 10 times 10 – 1,000x in one cycle.”
Son envisions this cycle repeating rapidly: “A second cycle [comes] every 12 to 18 months. So, the second cycle comes, 1,000x – that’s a million x. The third cycle comes, that’s another 1,000x – that’s a billion x.” He then addresses the potential skepticism surrounding such a dramatic projection: “So if you think, ‘Oh, it still has, you know, some mistakes in understanding, blah blah, oh, I am smarter than you,’ well, that thing you are looking down on from your perceived superiority…that little thing is becoming a billion times smarter. How dare you condescendingly refer to it as a ‘little thing’?”
Son’s history is peppered with similarly ambitious, some might say outlandish, bets. His early investment in Alibaba yielded astronomical returns, solidifying his reputation as a visionary. However, not all of his gambles have paid off. The WeWork debacle, in which SoftBank invested heavily in the co-working space provider only to see its valuation plummet amidst corporate governance concerns, serves as a stark reminder of the risks associated with his aggressive investment strategy. More recently, Softbank’s Vision Fund has struggled with significant losses driven by downturns in the tech market.
This context is crucial when considering Son’s AI pronouncements. While his track record demonstrates an ability to spot transformative technologies, it also highlights a tendency towards exuberance. Whether his “Stargate” project and the billion-fold increase in AI intelligence it promises will materialize remains to be seen. The rapid pace of development in the field certainly lends some credence to the idea of exponential growth, but the scale and speed Son predicts are breathtaking. Only time will tell if this latest prediction will be remembered as another stroke of genius or another example of overreaching ambition.