This transcript is generated with the help of AI and is lightly edited for clarity.
REID:
I’m Reid Hoffman.
ARIA:
And I’m Aria Finger.
REID:
We want to know what happens, if in the future, everything breaks humanity’s way.
ARIA:
Typically, we ask our guests for their outlook on the best possible future. But now, every other week, I get to ask Reid for his take.
REID:
This is Possible.
ARIA:
So, last week we had a great conversation with Ashton about investing in AI and its potential for our professional and personal lives. And so Reid, let’s talk about being early to spot and invest in people and ideas. You’ve been investing in AI for many years now, pretty much before anyone else. Can you walk us through how you think about making investments in AI companies and products in particular? What draws you to a pitch and what raises red flags?
REID:
Well, there’s two parts to this. One is general investing, which I’ll answer, and then AI specific. So general investing—you know, Ashton’s answer was great—who you’d work for as a combination of, do they kind of bring out the best in you? Do they have a good moral character? Do they have a good compass for where they’re going? Is the aligned mission there? You know, and then we focused on learning. And one of the things that Ashton talked about was distance traveled as part of this. And I think that was a great measure. I first heard about that from the Kapors, you know, Mitch and Freada. And I think it is, it’s a good proxy for learning about, kind of, what it is.
REID:
Although I think my particular focus on distance traveled is learning velocity and infinite learning. Like that you just keep going as part of it. And that’s part of the reason I amplified the kind of learning aspect of it. Because if you’re learning and you’re learning this game and the next game and, and constantly learning, that infinite learner is a key part of it. And, you know, I thought it was interesting—and another thing that was part of it is like, they have to be able to run the distance themselves. Part of the way that I look at this is, “Would I give the person the money as an investor and then walk away and say, call me in five years?” Because that’s part of it.
REID:
They have the game to learn it themselves and to what to do as part of it. Now you try to help as much as possible, just as Ashton said. Now, the thing that I would add is that you want to have a person who has a good kind of theory of the game. Now part of it is that learning velocity. Part of it’s like, “Do they see what the game is in front of them?” Like what are the key issues going to be? Is it going to be product-market fit? Is there going to be competition? Is there going to be changing patterns and technology? What is the business model? Is it to go-to-market? Is there, you know, assembling the right talent? Is there—and all of these things come into every entrepreneurial project, but which things are most important and what’s their particular shape here?
REID:
And part of the reason I add that into, as a general thing, to what we were talking about with Ashton is because the, the notion of—not just would you learn from them and not just would you have the infinite learner and learning speed—but you have a context awareness of your game and you’re updating that. And that’s part of what I get in the conversation that I always have with entrepreneurs about possibly investing. Because I’m trying to figure out, “How do they see their theory of the game? Do I agree? Does it update my theory of the game?” Because if it does, that’s a really great thing. That means that they’re likely to be really on point. And your theory of the game will change as you learn, as you encounter the market, as you encounter the consumer, the customer, as you encounter competition in terms of how you’re operating.
REID:
That’s all, I think, in the general framework. Then when it gets to AI—because AI is this platform transformational shift. I mean, you know, Ashton described it as an OS. OS is definitely our common metaphor for platforms. I actually think in this context, the platform will have a different shape than an OS—kind of building applications on top of it. The metaphor I think probably needs to be completely changed for it. But it is that kind of transformation. It’s like everything there, where you have a transistor in a product where intelligence can be added to a product and service. Phone and computer are obvious, but also your Sonos speaker or your light switch or any number of other kinds of things, I think will have intelligence amplifying it.
REID:
And that’s the kind of platform transformation. Now part of this is, it’s such a massive transformation in the world. You kind of get driven in AI investments to two kinds of questions. One is, “Is it an area that we kind of understand?” And I think this is one of the things that we’re doing at Greylock, which is—well actually, in fact, there’s like a lot of B2B businesses and kind of selling enterprise and everything else. And we understand what that transformation is. And so it makes a much more highly professional investing in it. And so I think there’s that set of things which—it’s like, “Okay, imagine everything you do at work and say, what is the AI replatforming transformation?” Whether it’s work productivity or file storage or marketing or sales or customer service and da da da and meetings and do all that. And then there’s the other thing, which is, “What’s the whole new world—just like the Internet was a whole new world—that comes about from AI?” And then, “What are AI and agents and everything else bringing to the new world of platform?”
ARIA:
Reid, it’s been a number of weeks since you debuted your AI avatar. So can you tell me more about how you think people can be in many places at once in their own lives with the help of AI?
REID:
Well, I think there’s going to be lots of ways. Some of it is that, you know, you’ll have an AI like—like everyone will have one or more personal intelligences, one or more copilots for themselves. One part of it is, it’ll always have an ability to kind of, as a word, answer the phone, answer the message. Kind of say, “Hey, is this really important? Oh, I’ll go find them and bring the person to you and bring the real person.” So that’s one place where you’re in lots of places, because you’re, you’re fielding a variety of inbound communications. Another one will be, like when you’re shopping for stuff, it’s like, “Hey, pay attention to when you know this thing—the Taylor Swift tickets may come on sale, or the sports tickets, or this new product, or the thing going on sale or price dropping.”
REID:
And so you’ll have representation of you doing all of that. I think you’ll also have, you know, questions where you might see saying, “Hey, I’m researching this and researching it takes time.” And you’re out doing that—it’s your research agent out doing stuff [laugh], right? While you’re doing something else. Your agent is researching path A, B, and C and you’re researching path X, as ways of pulling it together. And that’s just the beginning for this kind of multiple virtual location. Then of course, you could even get to—part of why we’re seeing the future—is like, say, “Hey, you know, what if you had your robot staying home with your pet and doing your own dog sitting?” Now we’re talking about virtual locations and then physical locations.
REID:
Then there’s also, by the way, other ways you might project into space. Like, for example, I was using singing—this is maybe one of the things that we will try at some point to get Reid AI to do since I can’t sing worth beans. It’s literally—the joke that I usually tell is a horror show. When I was part of my high school and we were doing the whole school coming together and singing, the very first comment I got—which I think was an accurate comment, even if it was a little dissuasive—was, “You know, when you can’t sing in tune, sing more softly.” And so, you know, getting some kind of singing you know, there—or other things, other expressions, other dimensions of yourself will be another thing that could be there. It could be poetry telling, it could be acting, it could be, you know, one we’ve already done—me and Klingon. You know, like all these kinds of things are another whole set of virtual projections or, you know, self-projections that will come of this. And that’s, like, that’s just the beginning. And our very quick answer to this question.
ARIA:
Reid, I have it even worse than you. My mom is a music teacher and my name is Aria and I also can’t sing. So there you go. [Laugh].
REID:
Well, we’ll create an Aria AI and then Aria AI can sing an aria.
ARIA:
Finally, I will have my mother’s gifts. So when we’re thinking—you mentioned robots to take care of your dog when you’re off on your trips—like when you’re thinking about AI hardware and agentic AI, what is the current state? Where are we going? How do you think about that?
REID:
Well, I think we’re just at the very beginning. It isn’t, we may even have just literally touched the surface of the water with our toes, let alone, you know, putting in the toes into the water. And I think we’ll see a lot more in agentic AI first just through software. First, just through what happens with the phone, first what happens with the computer, what happens with a variety of things. Because the speed at which that evolves—the speed at which that can be a personal intelligence, a copilot, a kind of an amplification of intelligence, an amplification of experience—I think will be happening at a very fast rate. So I think that the notion of, “Where do agents play in?,” will first play in all that. And you could say it’s kind of like a virtual space. But by the way, virtual space is also like when you go to the internet and you buy something from Shopify or something else, that’s a kind of a virtual space kind of thing.
REID:
And so virtual space is real space. Now, I think there will be a bunch of stuff that happens in robots. I think some of the earliest things—which we’re already seeing is obviously—like, you know self-driving, you know, like Aurora and other kinds of things where the self-driving car is a physical robot and that really transforms our lives. You know, everyone always tends to—because we’re very human focused— anytime there’s a humanoid robot, it’s like, “Oh my God, let’s talk about the humanoid robot.” You know, and it’s like, “Well, we already have a ton of interesting things going on with robots that have nothing to do with a humanoid robot.” And you know, one of the interesting questions that I think is like: “Well, we have a natural tendency to want a humanoid robot, because we tend to want to—you know, our whole society, we’re citizens of the polis, all Aristotle, we surround ourselves with human beings, we focus on human beings.”
REID:
So we tend to want, “Well, let’s have humanoid robots.” You know, when we have movies about good robots, they’re humanoid robots. Oh, there’s also bad humanoid robots. But there’s kind of a tendency there. I actually think one of the interesting questions is whether the most interesting kind of agentic robotic interfaces are not humanoid. And even though we have a bunch of arachnophobia, you know, is actually in fact, the arachno shape actually a much more high-functioning kind of thing. And I think it will be, you know, as Ashton was mentioning, I think our best thing to return to a manufacturing industry in the U.S. is AI and robotics.
REID:
It’s the thing that we should be wanting to, you know, be coming from this. And I think we will see, in the same way that we’re seeing an amazing set of things. Our paradigm with going on AI is, “Scale, compute, and learning.” And that “scale, compute, and learning” will be applied to robotics—will do a number of different things that are interesting there. I think it will come—because of the physical world, a bunch of other stuff, and the learning data—at a slower pace than the things that are happening through language models and other kinds of agents. But it will be coming at speed because of that application of scale-compute, the ingestion of tons of data, and the learning pattern of doing it. And so I think we will have a bunch of different robotic stuff. And I think it’s, “What a time to be alive!” I think we have an amazing set of human amplification—not just within, you know, the steam engine of the mind, but it’ll also be a return to the steam engine of the body. The steam engine of physical applications through the application of intelligence into robotics as well.
REID:
Possible is produced by Wonder Media Network. It’s hosted by Aria Finger and me, Reid Hoffman. Our showrunner is Shaun Young. Possible is produced by Katie Sanders, Edie Allard, Sara Schleede, Adrien Behn, and Paloma Moreno Jiménez. Jenny Kaplan is our executive producer and editor.
ARIA:
Special thanks to Surya Yalamanchili, Saida Sapieva, Ian Alas, Greg Beato, Ben Relles, Parth Patil, and Little Monster Media Company.