Venture capital pioneer Vinod Khosla says AI will deliver broad deflation
Summary
But as the price of expertise falls toward zero, a new age of abundance can begin, Khosla says—if people let it.MENLO PARK , CALIFORNIA : At a time of widespread concerns about the safety, efficacy and economics of AI, venture capital pioneer Vinod Khosla is still all in.
Yes, artificial intelligence will become capable of replacing a good deal of human labor, Khosla said in an interview, but that will push down the price of healthcare, education and other services. Many current forms of work will even be eliminated, but society will be able to create a more robust safety net than is possible today.
“Most expertise in the world, whether you’re talking about structural engineers, oncologists, mental health therapists or primary care doctors, or journalists and teachers, that expertise will be near-free for all of us to access," Khosla said at Khosla Ventures’ offices on Sand Hill Road in Silicon Valley.
Khosla, 69, is worth listening to at a freighted moment for the development of AI, partly because of his record as a driving force in technology since he co-founded Sun Microsystems in 1982. He joined venture-capital firm Kleiner Perkins in 1986, where his investments returned over $10 billion to the firm. His $3 million investment in Juniper Networks earned $7 billion for Kleiner.
Khosla also invested in OpenAI in 2018, well before the launch of its groundbreaking ChatGPT app in November 2022. He expects the pace of innovation—there and at other AI companies—to continue and even accelerate, eventually dispelling much of the doubt and fear that surrounds the technology today.
About 80% of the work involved in 80% of jobs across the economy can be automated over time, according to Khosla. “So 64% of all jobs can be done by an AI," he said.
For him, the bigger question is whether humans get in the way.
AI will be very deflationary, for example, if we choose to let it happen, Khosla said. “Technology makes a set of things possible. Policy often allows it to happen or not happen."
Economywide deflation, or falling prices, typically reflects stalled or deeply depressed demand. Such deflation has been linked to upheavals including the Great Depression in the U.S. and the economic collapse of early 1930s Germany.
But AI in the coming years has the potential to spur not only deflation but abundance, according to Khosla, by simultaneously fueling the creation of many more goods and services.
He views the next 10 years as a transition period in which the world’s political and social structures won’t seem all that different. AI will be seen as a boost for efficiency and productivity. After that, as the mid stages of AI-driven automation take a toll on more than 25% of today’s jobs, governments will need to provide much broader and deeper social services. But there will be enough economic abundance to support it.
Over a 25- or a 50-year period, GDP growth has the potential to climb to 5%. During the second quarter of this year, by comparison, the economy grew at an annualized rate of 2.8%. “We will have in that excess GDP growth lots of room to take care of the people who will be adversely affected economically," he said.
“If medical services are a lot cheaper, if education services are near free, if eldercare becomes substantially cheaper—and today eldercare is a liability that’s looming for so many nations around the world at a level that people haven’t accounted for—the social safety net is much easier to construct," said Khosla, a Democratic donor.
In his view of the future, people’s main preoccupation may shift from earning a living to finding meaning. AI and robots would take on menial labor, leaving humans freer to explore with the support of a comprehensive social safety net.
Anticipating automation
Khosla isn’t alone in his outlook, of course—at least elements of it.
Existing generative AI and other technologies have the potential to automate work that absorbs 60% to 70% of employees’ time today, consulting giant McKinsey said in a June 2023 report, raising its estimate from roughly half of employees’ time in a pre-generative AI study.
That potential won’t necessarily be fully realized anytime soon, though, given the effort it takes to integrate new technology and the need for that work to make economic sense and pass regulatory scrutiny. Perhaps 30% of today’s work hours might be automated by 2030, according to McKinsey Global Institute Partner Michael Chui, an author of the report.
And that’s only one possible scenario, Chui said. There are historical examples of technological innovation driving the creation of new kinds of jobs and work alongside automation of existing occupations, but that outcome can’t be taken for granted. “It requires a lot of work and a lot of reskilling and upskilling and all kinds of labor mobility," he said.
Darker fears about AI also have some validity, Khosla allows.
Possible rogue AIs could cause harm, he said, but there will be other AIs to fight them if so. “We’ll have to manage the risks," he said.
Khosla has made a life out of incubating ideas that seem improbable, at least at first, from online grocery delivery and mobile payments to solid-state lithium batteries, 3-D-printed buildings, low-cost solar energy, liquid biopsies and nuclear fusion.
“I like to say most people invest to increase the probability of success, but at the cost of minor innovation and reducing the consequences of success. We don’t mind the 90% chance of a project failing if there’s a 10% chance of changing the world," said Khosla. “I don’t need to increase the probability of success if the consequences of success are really, really consequential and I have a lot of shots on goal."
AI will certainly meet the “consequential" test if he’s right.
“The only question is will we let it happen," Khosla said. “You know, maybe unions will come back and will slow it down. That’s entirely possible, but I don’t think so, because the benefits to be gained are very large."
Write to Steven Rosenbush at steven.rosenbush@wsj.com