Masters of Disruption: How the Gamer Generation Built the Future [23]
Masters of Disruption: How the Gamer Generation Built the Future [23]
In the final part of my interview with John Carmack: Facebook worries, going big, and why he ditched his Ferrari for a Tesla.
This post is part of a longform project I’m serializing exclusively in my newsletter, Disruptor. It’s a follow-up to my first book, Masters of Doom: How Two Guys Built an Empire and Transformed Pop Culture, and it’s called Masters of Disruption: How the Gamer Generation Built the Future. To follow along, please subscribe to Disruptor and spread the word. Thanks!
I’m serializing my recent interview with John Carmack here in my newsletter, Disruptor. To read from the beginning click here. The interview has been edited and condensed for clarity.
David Kushner: What have you learned from your past work that informs your work in AI?
John Carmack: One was the lesson I learned from Armadillo [Aerospace, Carmack’s shuttered suborbital vehicle startup] versus Oculus: go big. Going big really does make a difference. Where I was trying to do it as my little side hobby project, I'm working in contrast to Elon [Musk]. I was just working with a set aside fraction of my money while Elon went all in. He basically could have been nearly poor if everything had failed with his ventures because he threw all of his money back into it. And he threw all of his time initially the SpaceX. I'm not going big on this right now. I've got people that are offering me investment money. But should I be staffing up? And should I at least have a team of the eight or 10 people working with me on this as we go ahead and try to do this?
David Kushner: But for now it’s just you working alone. Why aren’t you staffing up?
John Carmack: At Armadillo, I went through like a yearlong larval phase. I didn't know aerospace. I had to go learn aerospace. Before I could have a valid opinion, I had to at least know something about the state of things around me. I definitely had to do that for machine learning. There's a ton of hype. It's a hyperactive field right now.
“I could probably raise a couple hundred million dollars, but I don't want it. I know that could keep me from doing the things that I actually want to do.”
David Kushner: So are you still in that larval state now with AI?
John Carmack: I'm probably past that right now. I have informed opinions. Even a year ago, it was eye-opening to me to realize that I could have legitimate technical conversations with the leading people in the field. I wasn't just a naive outsider. I would argue that I am past that now and maybe I should be scaling.
But then the other aspect that I worry a lot about is at Facebook. They had virtual reality and augmented reality. A lot of people have always thought AR will be the real deal when everybody has it on all the time. But I have always argued that it's going to have been 10 years from the acquisition time to the time that they have an AR product. I look at it internally, I'm always saying, we've got signal on VR. If you apply yourself in VR right now, you'll actually be learning lessons versus spending 10 years going from a vision that you basically made up. You stared at a wall and said, “this is what I imagined AR is going to be like.” I just think that's the wrong way to develop things. But here I am doing essentially exactly that on AI where I'm saying I've got a 10-year vision of where I want to be. There are so many near-term values for AI. I'm very conflicted about this. Because while I'm all about this incremental learning, and trying to do something of value while you're learning, I've seen a failure case in that direction also.
On the aerospace side, Elon did do the big bang of ‘no, we're going straight to orbit. We're not even taking anybody's contracts.’ And he was able to do that in six years or something. So on the plus side, that's an example of that. But the AR development side, I think, is a negative example. They may yet prove me right or prove me wrong before I wind up getting the AI, which would be heartening to me. But that's not where I'm expecting things to go right now.
There's probably hundreds of hundred-million-dollar companies that can be built with the evolving machine learning technology for people just coming to grips with. There's even things that I would be well-suited for in, like, game asset development or making an AI tool for helping us. But I worry that that is totally get your eye off the ball of the big brass ring, which is the general intelligence. Because once you've got the general intelligence, worst case, you can teach them. It would not be the most efficient way to do it if you're paying by the compute hour, you'd still want to do a narrow one. But I worry that Open AI, they've taken a giant investment from Microsoft and they're commercializing some things right now, they may lose some of their focus working on those things as a result of that.
David Kushner: Do you think you’ll eventually decide to “go big” with your AGI work?
John Carmack: I think inevitably I will bring on some extra people. But I desperately don't want to go do the venture capital dog and pony fundraising show where I go around and talk about how enormous this is going to be. I know Mark Andreessen said that just talking to me was a big factor in the appeal of Oculus, but I'm not an optimal salesman for something. I could probably raise a couple hundred million dollars without having to do that, but I don't really want it. I know that could keep me from doing the things that I actually want to do.
“It doesn't seem catastrophic to me if AI becomes a dominant form.”
David Kushner: You have kids now. How do think that has affected your work? Does it make you feel any kind urgency to creating AGI?
John Carmack: Not really so much. I don't think a lot about legacy. I'm much more about the problem-of-the-week as I work towards longer goals, but I feel very fortunate to be able to have basically provided a wonderful life for my kids. They’re both super smart. My oldest got a perfect SAT score. I get my dad brag about that, the first time he took it. So definitely there's some heritage being passed on with that.
David Kushner: Is he interested in computers as well?
John Carmack: Engineering in general. But he’s not as super driven sort of entrepreneurial or self-motivated for things as I am. I do wind up wondering: can an environment be too good that it dampens some of the drive to go off and do different things? I don't claim to have any really deep insights about child-rearing. But there is a part of me that can look at the idea of if we do make super intelligent AI is, in some ways, they are our descendants. And again, I know some people are getting really bent out of shape about it, but it doesn't seem catastrophic to me if AI becomes a dominant form.
David Kushner: It’s like in James Lovelock's book Novacene where he says we’re the parents of AI, but they’re not our children.
John Carmack: Yeah. It's not something that I see that I spend a lot of my time thinking about. But, in my optimistic view, I think the world's going to be great for my kids to grow up in. And not just because they're children of wealthy parents. I think it's going to be a rewarding, enriching world where they have opportunities that were never available before no matter what their level of wealth. So, yeah. I still feel good about. I know some people go on about how climate change makes them not want to have kids, because they don't want their kids to grow up in a hellscape. But that's just not the way it's going to turn out.
David Kushner: Last question, do you still have a Ferrari?
John Carmack: No, I have a Tesla P100D. I've had a Tesla basically since they came out. Internal combustion just seems so antiquated now.