Vinod Khosla is hotter than ever. The Sun Microsystems co-founder turned high-profile investor, first at Kleiner Perkins and for the past two decades at his own venture firm, Khosla Ventures, has always been sought out by founders for his forthright advice and the firm's track record, which includes investments in Stripe, Square, Affirm, and DoorDash. But his $50 million bet on OpenAI in 2019 thrust Khosla Ventures and Khosla himself into the spotlight, even though it was far from clear that the outfit would be successful at its current scale.
He's having a lot of fun. When I sat down with Khosla at the Collision conference in Toronto last week, before taking the stage, he told me that he's been appearing in public multiple times a week these days: onstage, on podcasts, in TV interviews. I asked him if the schedule—he'd flown to Toronto just hours before the sit-down, for example—was exhausting, and he brushed it off.
Certainly, there are things he likes to talk about, but the art of the deal isn't one of them. “Frankly, the investor side isn't that interesting to me,” he said, when I asked him about something he'd heard recently: that Khosla Ventures now has $18 billion in assets under management, yet he hasn't taken a single dollar in management fees since he started the firm. (He acknowledged this, but said it only applied to himself, not the firm's overall policy.)
He's very enthusiastic about the startup opportunities he sees in a landscape that is changing every day due to advances in AI, so we spoke about some of this void. We also talked about what he's most concerned about regarding the ripple effects of AI, FTC Chair Lina Khan, and why, in his view, “Europe is regulating itself too much to take the lead in any technology sector.”
We started off by talking about Apple's splashy new deal with OpenAI, which allows the company to integrate ChatGPT into Siri and its generative AI tools. Apple likely has similar deals with other AI models, including Meta, but not surprisingly, Khosla, an OpenAI investor, is bullish on this one, which is the only one Apple has publicly announced to date.
Khosla called this a “validation of OpenAI technology.” [Apple] Apple said it “could have signed anyone” as its first partner. By announcing the deal with OpenAI at its high-profile developers conference, Apple is ” [OpenAI CEO] Sam [Altman] Lead [developments in AI] “It's not going to change in the next five to 10 years,” Khosla said. “When a company like Apple makes a bet on technology, it's usually not going to change in the next year.”
As I've said before on TechCrunch, many startups are likely to be disrupted and killed by Apple's latest features. I asked Khosla if this was the case with its portfolio companies. I wondered in part about Rabbit, an AI-powered hardware device backed by Khosla Ventures that promises to be a kind of AI assistant for users.
Asked whether Apple could make the device obsolete, Khosla suggested the device is more flexible than people might imagine and could be used in hospitals and other businesses, including in emergency room environments. He characterized the device as one in a growing number of devices that “monitor what you do, see what you do and automatically respond.”
Indeed, Khosla suggested that his team has been actively avoiding what could become “roadkill” as large-scale language models like those from OpenAI get more advanced, and he singles out at least one company that’s not in the portfolio: Grammarly, the writing assistant startup that was most recently valued by backers at $13 billion.
“If you're doing, say, Grammarly, it's a really light wrapper in today's model and Grammarly can't keep up. Grammarly should never have been an app. It demonstrates the need for that functionality, but it's going to be part of Word or Google Docs. That's pretty clear. When we talk to YC companies or other companies, I can usually say, 'Half of these companies will be obsolete before the YC batch is over,'” Khosla continued.
Khosla sees a lot of opportunity in areas where expertise is nearly free, but I don't see how these companies will continue to make a profit (even after asking him.) Think private tutoring or oncology.
“OpenAI and Google are not going to train chip designers,” Khosla said. [to have on your smartphone]”OpenAI and Google aren't going to train structural engineers. They're not going to train family doctors or mental health therapists,” he said. “So [founders to mine]But you have to think about where the model is going next year and five years from now and say, “I want to take advantage of that capability.”
We also spoke about regulation: Khosla has previously said that closed, large-scale language models like OpenAI's should be protected even with a regulatory framework, which left me wondering if this meant he was abandoning other “open source” AI forever.
Not at all, he said, noting that he's a “huge fan” of open source. Sun was one of the first companies to “jump on open source” by open-sourcing its file system, he said. He also noted that Khosla Ventures was one of the earliest investors in GitLab, software that invites people to collaborate on code.
But he suggested that open source in the context of large-scale language models is a whole other story. “The biggest risk we face in AI is China and 'powerful Chinese AI' competing with America's 'liberal values,'” he said, adding that “we need to make sure China gets behind us.” Otherwise, he warned, China will end up providing the world with “free doctors and free oncologists,” and incidentally “exporting both the economic power and the political philosophy that goes with AI.”
Onstage, I told Khosla about my recent meeting with FTC Chairman Lina Khan, who doesn't believe the national champion model is a justification for coddling organizations like Google and OpenAI in their AI development.
Khan regularly hears from business executives and investors that government intervention would put the U.S. on a dangerous path. But when I met with her, she argued that America has chosen “the path of competition” time and again, which “has stimulated and driven many of the breakthrough innovations and remarkable growth that our country has enjoyed and that has enabled us to continue to lead the world.”
“If you look at other countries who have chosen the National Champion model instead, they are the ones who have been left behind,” Khan added at the time.
But when I barely mentioned Khan, Khosla turned dismissive, calling her “not a rational person” and accusing her of not understanding business.
“She shouldn't be in that position,” Khosla said. “Antitrust is a good thing in any country, in any economy. But antitrust is a good thing. [that’s] “Over-enforcement and over-execution is bad economic policy. One of the advantages the U.S. has over its European rivals is a much more streamlined business environment. That's why Europe has been regulating itself from taking the lead in any technology field. It's been regulating itself from AI, all social media, all internet startups, basically.”
Of course, some antitrust enforcement is good, but too much is not good, and the question is where to draw the line. On this point, before we parted ways, I mentioned Altman's reference to the “abundance” that will be created by AI. At TechCrunch's StrictlyVC event last year, Altman said that the “good case” for AI is “so incredibly good that when you start talking about AI, you sound like a really crazy person.”
Khosla said he agreed, but I've long wondered how society could reap all of this benefit without regulators getting more involved in what these companies do. After all, I told Khosla onstage, we're already seeing huge aggregates of wealth and power being attached to smaller and smaller companies and individuals. When will we say enough is enough?
Here, Khosla said this issue bothers him deeply: “I think that in 25 years' time, when I hope I'll still be working, there will be very little need to work.” Still, he said, while AI should generate “greater prosperity, greater GDP growth, greater productivity, all the things that economists measure,” he worries about widening income inequality “more than anything.” [ensure the] Is it possible to distribute the benefits of AI fairly?
He realizes where the tipping point is. [U.S] If GDP growth could rise from the current 2% (currently less than 1% in Europe) to 4%, 5% or 6%, there would be enough wealth to share the wealth and the benefits.”
Of course, whether and how that will happen is the bigger question, and for all his brilliance, Khosla, a self-described tech optimist, didn't have an answer. Instead, he thanked the audience for their time, walked offstage, and toward the dozen or so founders who had gathered, all waiting in the wings to listen to him for as long as they could.