Latest / Newcomer Pod / Two Titans on the Future of AI (with Reid Hoffman & Vinod Khosla)
Transcript
- 0:00Hey, it's Eric Newcomer. This is the Newcomer Podcast.
- 0:03We've got a live episode coming to you from the Schruebel Valley
- 0:06AI Summit on November 15th. We've got two conversations
- 0:10between me and prominent venture capitalists.
- 0:13This is two of my favorite conversations back-to-back.
- 0:16First, Reed Hoffman, the Co founder of LinkedIn Co, founder
- 0:20of the Foundation Model Company, Inflection Greylock Partner.
- 0:24And then after Reed, we've got Vinod Khosla, founder of Khosla
- 0:29Ventures, founder of Sun Microsystems and always
- 0:31opinionated spicy venture capitalist.
- 0:35Reed and Vinod took different stakes on some issues, so I we
- 0:38wanted to put them in conversation with each other.
- 0:41We did record the conversation November 15th, so before the
- 0:45blow up at opening Eye. So listen with that in mind
- 0:49before we get to those conversations.
- 0:51A word from our sponsors, Oracle and NVIDIA.
- 0:54Thousands of enterprises around the world rely on Oracle Cloud
- 0:58Infrastructure OCI to power applications that drive their
- 1:01businesses. OCI customers include leaders
- 1:03across industries such as healthcare, scientific research,
- 1:07financial services, telecommunications and more.
- 1:10OCI also works with NVIDIA to provide an AI training as a
- 1:14service platform for customers to train complex AI models.
- 1:18Talk with Oracle about accelerating your GPU workloads
- 1:21at the link in the description. And now my conversations, first
- 1:26with Reed Hoffman and then Vinod Khosla.
- 1:29I wanted to start off, you know, we've been so AI focused that I
- 1:33think for the first half we forgot, oh, you know, the
- 1:36president is meeting with Xi Jinping today in San Francisco.
- 1:39Besides, this is being taken over by politics and
- 1:42international relations, I I guess.
- 1:45Have you talked to any of the people in town as someone who
- 1:47gets to connect with that sort of crowd?
- 1:50Well, so yesterday I actually met with President Biden, who
- 1:54obviously was not flying out here to meet with me.
- 1:55He was flying out here to meet with President Xi.
- 1:58But of the many things that I don't think Biden gets enough
- 2:00credit for is he's actually fairly intellectually curious.
- 2:04And so they did a whole bunch of very good work on the executive
- 2:07order. And he was like, OK, was it
- 2:10good? What do we need to do next?
- 2:12You know, one of the things that I think they did that was very
- 2:14smart is they put the primary locus within commerce.
- 2:17Because part of what, you know, Biden is thinking about is.
- 2:21How do we make sure that this is really good for American
- 2:23industry jobs, your American worker and so forth?
- 2:26And that was the kind of conversation he wanted to have.
- 2:28So he was flying out here to meet with President Xi, but he
- 2:30was like, oh, you live out there?
- 2:32I can talk to you about AI come by.
- 2:34And so, of course, getting through the security cordon at
- 2:37the Fairmont was a bit of work. What, what did you say about the
- 2:42executive order? Are you unabashedly supporter
- 2:45middle ground? What's your mood on the
- 2:47executive order? I would call myself a 9598%
- 2:53supporter. I think a lot of it's very good.
- 2:56I mean the kind of notion of reporting and monitoring, being
- 3:00an agent, dialogue, revealing what's kind of what you're
- 3:02what's going on, that kind of thing is all, I think part of of
- 3:05good, smart, intelligent governance.
- 3:08I think red teaming is very good.
- 3:09I think getting. Red teaming by third parties is
- 3:13good. Starting this with voluntary
- 3:14commitments as a way of doing it and then kind of gearing it up
- 3:17and using the Defense Production Act I think is all very smart as
- 3:20ways of being kind of effective governance.
- 3:23The only part that I probably disagree with in part because,
- 3:26you know, I think a little bit of what Lina Conn's doing with
- 3:30the FTC is to say big technology is bad.
- 3:33I actually don't think that's the case, not necessarily just
- 3:37because it's big. Part of having big industries
- 3:40and businesses as they're, they set global platforms, they can
- 3:44do things from everything from American industry to American
- 3:47soft power and influence in the world.
- 3:49And the line in there that's like oh the FTC you should be
- 3:53making sure that there is, you know, nothing here anti
- 3:56monopoly. If you actually look at a lot of
- 3:57the interesting work, which is, you know, not just happening
- 3:59with inflection, but also Open AI and Entropic and a whole
- 4:04bunch of other things, you know, start-ups are not being impeded
- 4:06right now. Even though obviously there's
- 4:08great work for Microsoft, Google, you know, etcetera on
- 4:11this as well. So that was probably the the
- 4:13principal piece of the EO that I disagreed with.
- 4:16But the stack of how do we be navigating intelligently and be
- 4:21getting good data was I think smart.
- 4:24How much did China play the backdrop?
- 4:26I mean there there's this great foil for Silicon Valley which is
- 4:29if you over regulate us, you know China is going to out
- 4:33compete us in AI and that's going to be a threat to national
- 4:36security, our economy and everything else.
- 4:38Like how how much was China coming up in the conversation?
- 4:40Well, in in my conversation yesterday, not very much.
- 4:45I did ask him some questions about, you know, kind of what
- 4:48would be the set of things around, you know, his discussion
- 4:51President Xi. And it was kind of questions of,
- 4:54OK, let's let's try to make sure that we have good relationships,
- 4:58that we're competing on a fair playing ground on a number of
- 5:01things. Let's be still focused on
- 5:03climate, a set of different things.
- 5:05But. Relative to China and AI, this
- 5:09is, I think, as momentous a moment in economic elevation as
- 5:14the steam engine, right. So, you know, I kind of, I call
- 5:17this the steam engine of the mind.
- 5:19And so as such, it's intensely part of how we elevate our
- 5:23industries. You know, one could choose not
- 5:25to use the loom. That would be a mistake.
- 5:27One could choose not to embrace the steam engine.
- 5:29That would be a mistake. Doesn't mean that there aren't a
- 5:32lot of challenges as you. Kind of.
- 5:35Retool your industries and economy and what does that mean
- 5:37for industry, jobs, and unless you have to navigate that, not
- 5:41to oversimplify where you stand, but do you think TikTok should
- 5:44operate in the United States? Yes.
- 5:47Although, so here's the general thing is we as the US say that
- 5:53we think that it's an important thing to have Internet platforms
- 5:57broadly available because by the way, historically most of those
- 5:59Internet platforms have been US ones.
- 6:02If you say, well we're now uncomfortable with another
- 6:05Internet platform operating here that validates all other
- 6:09countries going well, we're Brazil and you know your
- 6:14Internet platforms are operating here.
- 6:16I think the question with TikTok is not, I mean we and we could
- 6:20say here's what we're going to do in terms of general content
- 6:24governance across country borders and then we should do a
- 6:27multilateral version of that. But I think that the thing that.
- 6:31That we need to do with Tiktok and say, look, you slant your
- 6:35local Chinese tech market away from US and other companies.
- 6:40Let Facebook manipulate you if we're if Tiktok's going to
- 6:43manipulate. Us.
- 6:44Yeah. Yes.
- 6:45Or let's have a fair playing field between the two of them,
- 6:48right. In terms of being a Democratic
- 6:50donor, maybe beyond even just the AI question.
- 6:53I mean you're sort of, I don't know, Bizarro world counterpart
- 6:56Peter Thiel on the right is saying I'm going to swear off
- 6:59like donating this time. I don't know if it's oppo
- 7:02researchers coming at him or just like he he he said in a
- 7:06recent interview that Trump was worse than he thought.
- 7:08Yeah, I'm curious, how gung ho are you for Biden and the
- 7:11Democrats in in this re election?
- 7:13And then maybe what do you make of Peter Thiel's comments?
- 7:17So you know, one of the things that I think.
- 7:20Biden, to his credit, had been so focused on doing the job, he
- 7:23hasn't been spending the time Tooting his own horn.
- 7:26He's navigated a very difficult situation around Gaza, you know,
- 7:30with, you know, grace and aplomb trying to, you know, the
- 7:33multiple tragedies going on there.
- 7:34He brought all of Europe and a lot of the world into the
- 7:39Ukraine and dealing with that situation, the Inflation
- 7:42Reduction Act with climate is very good thing post COVID
- 7:45pandemic. Economic recovery, you look at
- 7:48the job numbers, it's very good. It's got to be better bragging.
- 7:50And more bipartisan legislation than like in than in decades is
- 7:56ways to do it. Like all it sucks is very, very
- 7:58good. So like I think Biden, one can
- 8:01make a very strong case, has done a very good job over the
- 8:05last, you know three years and it's partially because of of
- 8:08experience. Now, that being said, I think
- 8:12you know we have a. Prospective presidential
- 8:15candidate who is, you know, who has former people he's
- 8:21describing as his lawyers, you know, kind of turning and saying
- 8:25and turning evidence against him.
- 8:27This is very reminiscent of mob boss, You know, kind of
- 8:31behavior. And you know, describing his
- 8:35political opponents as vermin, which is a very fascist like way
- 8:39of doing this. Yes.
- 8:41And so it just it's it's like I look, I think it's super
- 8:43important. I think, I think what we care
- 8:45about as American values in Rule of law and democracy and a bunch
- 8:50of other things is on the table and I think we should be out
- 8:53there. Doing something about it.
- 8:55So that's what I'm doing. Have you committed a dollar
- 8:57figure yet? Or?
- 8:58Well, I never pre commit dollar figures, but I've already
- 9:01started investing the oh in the Peter Thiel piece.
- 9:04You didn't touch on that? Yeah, yeah, happy to do it.
- 9:05Sorry I was just answering the 1st question, I know I asked too
- 9:07long too long of a question. So, well, look, I had a number
- 9:11of arguments with Peter about Trump.
- 9:13Matter of fact, there's probably the most ferocious set of
- 9:15arguments we've had, because I just couldn't understand why
- 9:22Peter couldn't see Trump as a. You know, it's kind of a a, a
- 9:25Chernobyl. The way that I see him And, you
- 9:28know, part of Peters and My's friendship all the way back to
- 9:32the Stanford days was based on the value of public intellectual
- 9:35discourse, of making arguments. And I've learned a ton from
- 9:39Peter about, you know, kind of how to think about, you know,
- 9:42some of the interesting perspectives and in good in
- 9:45depth perspectives on conservative arguments from him.
- 9:47And then that part of the thing I was like, Trump strikes me as,
- 9:50you know, how can you tell is he's lying, His lips are moving.
- 9:53Right, so it so it it's like, what?
- 9:56Right. And so I was not surprised that
- 9:58he then went, oh, shit like this was not as worse than I.
- 10:02Expected. But I think we still have to
- 10:04deal with the ramifications of we have a a a former liar in
- 10:07chief the you know we talked about active regulation you know
- 10:11now there's this effort responsible innovation hey
- 10:15Montenegro General Catalyst is sort of very involved.
- 10:17Do you have a stance on that like what do you think of sort
- 10:20of industry self regulation where that?
- 10:22Is positioned well. So broadly I think industry self
- 10:26regulations are good things to generally do.
- 10:29Industry usually will calibrate well to what the cost and so
- 10:32forth of doing this stuff is. I think that doing the voluntary
- 10:36commitments from the White House was good.
- 10:38I think that saying hey look investors should be responsible
- 10:41here too is a good motion. I think I kind of looked at it
- 10:46and kind of bogged down on not coming around to signing it
- 10:49because one of the dynamics that.
- 10:52You know I believe and I know you know my partners at Greylock
- 10:54believe as well is you don't impose thing on companies you
- 10:58you you only invest in good entrepreneurs who who have good
- 11:01ethical compasses and good projects in the world.
- 11:04But you don't show up saying well I have a 10 item thing that
- 11:07you have to do. It's like no do the right
- 11:09ethical thing of the thing you're doing and we work and
- 11:11collaborate with you. So it just seemed a little
- 11:13dictatorial between you know kind of investors and founders.
- 11:17So we don't almost ever do that sort of thing.
- 11:20It's like like we don't show up saying.
- 11:22Well, you're going to have a company culture with no sexual
- 11:24harassment. Well, if you're going to have a
- 11:26culture with sexual, we want nothing to do with you, right?
- 11:29We don't need that. We sort of need to screen it in
- 11:31the founder selection sort of ethics.
- 11:33So that's that's how we operate on these things and we don't, we
- 11:36we show up as the invited partners with our entrepreneurs
- 11:41versus the, Oh well, we have a set of things that you have to
- 11:43do what we tell you to do, what do you make on the other sort of
- 11:47extreme of GC and the Co signatories, Andreessen Horowitz
- 11:52is, you know, really using this as a way to be like we're wild,
- 11:54like anything goes. Do you have a reaction to that?
- 11:57Do you think it's dangerous or? Look, so I am also a techno
- 12:02optimist. I I I found it entertaining that
- 12:05Mark like Mark Andreessen quoted kind of liberally from some of
- 12:09the things that I write and speak about like homotechnae and
- 12:11other things without attributing.
- 12:14That's fine, no problem, right? You know, and I've been kind of
- 12:19a techno optimist on AI from very early.
- 12:22It's part of like publishing, impromptu and all the rest
- 12:24involved in open AI super early. Now, that being said, it's dumb
- 12:29to think when you have major technologies that there can't be
- 12:33negative side effects that you need to navigate around and be
- 12:36thoughtful about them. And so for example, every
- 12:41project that I'm part of Open AI inflection, etcetera, does have
- 12:45a safety team. But it's focused on important
- 12:48topics. Like important topics like,
- 12:50don't make it easier to make bombs, right?
- 12:54Like, like there's an easy one that everyone agrees with, you
- 12:57know, and there's a stack of these things they say, well, but
- 13:00then it's, you know, it'll talk smack about Trump but not about
- 13:04Biden. You know, like, OK, I'm sure you
- 13:07can ultimately get them to crack jokes about both of them.
- 13:10I'm sure I could go to get light bulb jokes about both of them,
- 13:14you know, as kind of ways of doing it.
- 13:16But like, for example, having an A informed point of view of the
- 13:192020 election was not actually stolen, right?
- 13:24Or if it's stolen, then every American election is stolen.
- 13:27So it's a meaningless statement and you're sort of saying it can
- 13:29look biased if the facts on the. Ground are biased.
- 13:32Exactly. So it's like, OK, so like like,
- 13:35be more thoughtful about navigating inside.
- 13:37It's not like whatever you can build with technology is grand.
- 13:40It's technology can be amazing. Let's be intentional about
- 13:45building that intentional, that amazing technology.
- 13:47What is your view right now on existential risk, X risk?
- 13:51I feel like there's almost like a backlash to it at this point
- 13:53where there are people who think it's a promotional technique for
- 13:57AI. Like where are you on the
- 13:58possibility that X risk is real? So I think the people who
- 14:02articulate XX risk existential risk.
- 14:06I presume everybody here knows that.
- 14:08Are serious and earnest, and so I value their their, your
- 14:13Inflection Co founder Mustafa, who we're talking to at the end
- 14:16of the day, seems more worried about.
- 14:18Yes, although you what you will find in talking to him but he's
- 14:20more he finds that the X risk people are occluding the real
- 14:24risks. Which is what does AI do in the
- 14:26hands of humans. What does it mean for jobs?
- 14:29What does it mean for bad human actors doing stuff?
- 14:31That's the importance of things which he and I totally agree
- 14:34with, as those are the risks to focus on versus S Cris.
- 14:37Now the thing with X risk is you know like that 22 word
- 14:40statement, it should be considered a existential risk
- 14:42along with climate change and so forth.
- 14:44The reason I didn't sign that statement is because when you
- 14:47look at climate change, pandemic, a bunch of things,
- 14:49those are just risks. They're just bad.
- 14:51AI might add some robot risk, but it also is how do we solve
- 14:57pandemic, how do we improve climate, how do we deal with
- 15:00asteroids, all these things. AI is in the positive comment
- 15:03and and the mistake that I think all the X risk people make.
- 15:06Is they try to treat them as each solo versus a portfolio.
- 15:10So my view about X risk for humanity is what is the, what is
- 15:14the portfolio of X risk for humanity And as you're doing
- 15:17things, are you improving net portfolio and I think AI
- 15:21improves the net portfolio. And so therefore I'm not one of
- 15:24the X risk to solve things. You know, we don't know which
- 15:28way it's going to go, right. I certainly see the argument,
- 15:31OK, sort of this next portion of the conversation into the where
- 15:34the business dynamics and the companies.
- 15:36You know, you were. Yeah, you were so involved with
- 15:38Open AI earlier, early you you've stepped off the board
- 15:42what what is And you know they just had Dev Day and I think
- 15:44it's come up at this conference already.
- 15:46Like there's a sense that, oh, is that going to like destroy
- 15:49every startup? Like is it going to hurt your
- 15:51investments? Like, what do you make of the
- 15:53power of Open AI right now, and where there's room to compete
- 15:56with them? So I think Open AI is obviously
- 16:00made a set of very smart bets about the scale application of
- 16:04large language models. And they have with a bunch of
- 16:08genius and bright moves and people have have created the,
- 16:13the leading edge of of the drum beat by which everything else is
- 16:17following. And that's awesome.
- 16:19I don't think it constrains competition and almost any
- 16:23level, you know. I think there's even people
- 16:26competing to offer Frontier Model AP is.
- 16:29Open AI is not the only only party doing that.
- 16:31I think that the question of there's going to be tons of
- 16:35different interesting bots. So yes, ChatGPT is one of the
- 16:38great bots. I think there's going to be
- 16:40others obviously with inflection of pie is as part of doing that,
- 16:43but I think they're going to be different in solving different
- 16:45needs. Now, if what your start up plan
- 16:47was, I'm going to be a thin wrapper on top of a company's
- 16:51API, that's a dangerous place, whether it's Open AI.
- 16:57Google, Amazon, Microsoft, any of these things, you have to do
- 17:01something more substantive, whether it's like an enterprise
- 17:04integration, a network effect, a a stack of technology that
- 17:08really adds in addition to the API, those those are the things
- 17:12you need to be doing. And by the way, of course, part
- 17:14of it is there's so much progress happening within these,
- 17:18these, these AI capabilities. That you can't say, Oh well, I
- 17:21built my thing on GBD 3. Oh shit, GBD 4 is so much
- 17:24better. Fine.
- 17:25You need to be, you need to be anticipating that's what's
- 17:28coming. Where do you see us right now
- 17:31and sort of maybe the S curves or whatever charting of AI
- 17:35improvement? Like do you think next year
- 17:37we're going to enter sort of a flat lining period or you're
- 17:39seeing a lot of still acceleration or how do you think
- 17:42about the actual underlying technological improvement?
- 17:46So two things. One, I don't think we're at S
- 17:49curve yet. But one of the things that
- 17:52people frequently misunderstand is that we get to each new level
- 17:55of scale. They think it's, you know, just
- 17:57put in 10X compute, just put in 10X data, press button, magic of
- 18:02urges. And he's like, that's a lot more
- 18:04work than that, right? And and part of it is you get
- 18:08these different levels of scale, you have to figure out kind of
- 18:10different set of techniques that cause it to work the right way.
- 18:13And so you can have like the very first training run of GPD 4
- 18:18failed. And then they figured out some
- 18:20techniques to go, oh, if we teach at summarization and we do
- 18:23this then we can make it work really well.
- 18:25And so I think we will have to do those things for GPD 54.55
- 18:31and the kind of equivalent. But I think they're there.
- 18:34I think that's capable to do that.
- 18:35Like I think that's that we found it's not science risk
- 18:39maybe sometime in the next 5 or 10 years, I think those things
- 18:41will happen. And so I think we will certainly
- 18:43see. Similar between 3:00 and 4:00
- 18:46between 4:00 and 5:00 next year or.
- 18:49I don't know about next year. That would be huge to me, right?
- 18:52Three to four. Four to five would be if those
- 18:54are. Equal, but even 4.5 next year
- 18:57will be significant. Wow.
- 18:59Have you started to see any of that or?
- 19:01I haven't yet, but I anticipate from the.
- 19:05Buzz that all of the the the very leaky industry.
- 19:09Exactly. I'm anticipating.
- 19:11Do you see you can dodge this one if you're not allowed to,
- 19:14But like, is Open AI Microsoft's strategy?
- 19:16I mean, you're on the Microsoft board, right?
- 19:18Like how much do you see Open AI as the strategy for Microsoft in
- 19:22terms of AI? Well, it's certainly one of the
- 19:26major strategies and obviously Microsoft has a number of
- 19:30different. Business lines that Open AI is
- 19:33not really in, that it's also doing these things in.
- 19:36But I think this the partnership between Open AI and Microsoft is
- 19:40going to be one of the epic partnerships that Business
- 19:44School classes will be taught on for decades, just like Windows,
- 19:48Intel from back of the day. It's a similar massive kind of
- 19:53alignment that's going to create all kinds of things.
- 19:56Or it could be sort of a Steve Jobs like returning to Apple's
- 19:59sort of situation. Is there a risk?
- 20:01That, like Open AI, represents the future and Sam Altman
- 20:04becomes sort of the future of Microsoft well.
- 20:06Both Sam and Satya have put a lot of energy into aligning the
- 20:11interests and so there that's reason it's much more like kind
- 20:15of the Wintel kind of period which is I think they will both
- 20:19more or less succeed together or not succeed together.
- 20:22I I think that's the way that that will play out.
- 20:26How do you think about your time?
- 20:27You know, you're stepping back, so.
- 20:29I think they will succeed. Just to be clear, I wouldn't.
- 20:32I didn't mean to express hesitancy.
- 20:33I just think it was well aligned.
- 20:35No, no, no. Actually, now you're making me
- 20:36go back the, you know where I'm going to have a note on stage.
- 20:39They ended up getting the first round.
- 20:41Do you feel any regret that you didn't get the sort of venture
- 20:44investment in open air instead of personal?
- 20:47So I did. I actually was the lead
- 20:52financier personally out of my foundation.
- 20:54I did talk to SO that has a lot of equity, so.
- 20:57OK. Yeah, I, you know, so I
- 21:03technically first money in. All right.
- 21:05So we'll give you your credit. That's fine, That's fine.
- 21:07Whatever, whatever, whatever. You know, now I did talk to
- 21:11Greylock about it and I said, look, they don't have a go to
- 21:14market strategy. They don't have a business
- 21:16model, right? They don't have and look, part
- 21:19of our job with our LP's is to invest professionally.
- 21:24Right on this. So look, I think the technology
- 21:26is going to be really great, but I have no idea, which is the
- 21:28reason I'm doing it from my foundation, right.
- 21:30Right. So but of course, if knowing
- 21:32what you know now, if you somehow look all investing, it's
- 21:36much easier with crystal balls 10 years in the future.
- 21:39So you know, the what does that foundation give to you?
- 21:42Is that part of your political giving or is that separate?
- 21:44No, it's a it's a 501C3. It's a it's a foundation that,
- 21:47you know, invests in opportunity, for example.
- 21:50How do we get the the various disadvantaged communities to
- 21:53have much more economic opportunity?
- 21:55How do you enable science? I mean, it's doing a whole bunch
- 21:58of different things. How do you think about your time
- 22:01and your what you're working on right now because you do so many
- 22:03things, wear so many hats. What's sort of the priority at
- 22:05the moment? So for me, the priority for this
- 22:10year and next year, well next year especially, is obviously
- 22:13artificial intelligence and making sure we don't fumble all
- 22:16of the really great things that can help elevate humanity.
- 22:20And with some regret 2024 election, because I think it it,
- 22:25it matters to us and to the world.
- 22:28I would rather just be building, right.
- 22:30I'd rather just be doing all the kind of investing and technology
- 22:33stuff. But you know, when the when.
- 22:35So you're going to call in on the election.
- 22:36Oh yes, 100%. Are you primary and some
- 22:38Democrats? Not currently that I'm aware of,
- 22:43right? I do have a whole political team
- 22:45and all the rest that also does other things for other people.
- 22:48Look, I'm fundamentally a centrist.
- 22:50What I would really like both parties to have is more of an
- 22:55incentive going towards the center like.
- 22:57So I think that's part of the corrosion that's affecting the
- 23:02our society. Now one of the things that is a
- 23:05very popular thing to say is that, well you know, both Biden
- 23:08and Trump are like Biden's a centrist, He's been a centrist
- 23:11for decades his entire political career.
- 23:15So that's part of the reason why I strongly pro Biden, Great.
- 23:19Thank you so much for coming on stage.
- 23:21This was awesome. In our second to last
- 23:25conversation of the day, joining us on stage, venture capitalist
- 23:28Vinod Khosla, he's talking with Eric Newcomer.
- 23:34Vinod Khosla, thank you so much for sitting down with me.
- 23:37I read in the Wall Street Journal that you guys are Khosla
- 23:41Ventures are raising a $3 billion fund.
- 23:43Is that right? Yes, it's right.
- 23:46We're almost done what, you know, it's a time where many
- 23:50venture capital firms are sort of struggling or downsizing
- 23:54their fund size. I assume the Open AI investment
- 23:57helped along. But yeah, what what was the
- 23:59decision to sort of expand and what was sort of the fundraising
- 24:02environment like? You know, as Warren Buffett
- 24:04says, when others are fearful, it's time to be aggressive.
- 24:08When others are optimistic, it's time to be conservative.
- 24:12You're one one of the odd things.
- 24:15About our fund, if you look back the last five or six years, our
- 24:19rate of investing hasn't changed.
- 24:21So 20/21/22, we didn't like double our rate of investing.
- 24:27Generally it stayed about the same.
- 24:29I do think in this new domain of AI, it's time to be aggressive
- 24:34and both thoughtful and aggressive.
- 24:37It feels like you're deploying more capital than many firms
- 24:40right now. Do you?
- 24:41I feel like there's concern that you know, valuations, you're
- 24:45just paying a high price for AI companies and that we don't know
- 24:48anything outside of AI. We don't know what the bottom is
- 24:51or what prices will look like. How do you think about valuing
- 24:54companies in such an uncertain environment?
- 24:56You know there's two styles of investing and you know,
- 24:59Instacart recently had an IPO that really showed up all the
- 25:03styles of investing. I'll get the series wrong, but
- 25:07the 1st 3 rounds? Were reasonable valuation
- 25:11stepping up and we stepped up aggressively invested Now we
- 25:16started with $1,000,000 for at a 10 million post.
- 25:18So we bought it at a pretty good valuation then it's very nice
- 25:23these days. Yeah.
- 25:24And then we invest in next series and the next series and
- 25:27the valuation got to a billion and we decided not to invest.
- 25:32But what was happening was investors were doing momentum
- 25:37investing. So they're looking.
- 25:38Oh, he's investing at a billion. He's investing at a billion.
- 25:41Our men that kind of follow the herd kind of momentum investing
- 25:47is different than when you're investing in fundamentals.
- 25:52And we were investing in fundamentals by the way.
- 25:54Same time we invested in Instacart in the same time
- 25:58frame, roughly. We also put $1,000,000 into
- 26:01DoorDash at the gain a $10 million post.
- 26:05Because not too many people were investing in these areas back
- 26:09then, right? And I mean to translate, you
- 26:11know, early Instagram investors made money and later ones less.
- 26:15So can you talk us through given, you know, the AI
- 26:20audience, the investment in Open AI?
- 26:22I mean, we had Reed Hoffman on stage.
- 26:24You know, he got in through his foundation, but you took the
- 26:27first venture around. What was sort of the dynamic
- 26:30there and how did you just have the conviction for what seemed
- 26:33at the time to be a science project?
- 26:35Well, you know, Reed is very forward-looking.
- 26:40So I'd say I of all the people in the venture business, he's
- 26:46high on my list of people I really admire and how he
- 26:49invests. So I admire him a lot.
- 26:53In fact, but a venture fund like Greylock probably wouldn't want
- 26:56to invest in what was a speculative round.
- 27:00But the math was very simple. If you lose, you lose 1 times
- 27:03your money. If you win, you make 100 times
- 27:06your money. So you could place 50 bets, and
- 27:09if 49 of the 50 lost, you'd still do OK.
- 27:14But it was much more than that. I started writing about AII
- 27:19think it was 2011, Christmas, I said.
- 27:23Wrote about do we need doctors and do we need teachers?
- 27:27With the idea that an AI tutor would do the right thing.
- 27:30And my wife is just a beautiful AI tutor that's free, by the
- 27:34way. It's in a nonprofit, so I don't
- 27:36much care whether something's in a for profit or a nonprofit.
- 27:40My son's working at QRI in an AI doctor.
- 27:43I wrote about that in 2012. What was clear to me by 2018, it
- 27:49was around this time we made the decision five years ago to
- 27:52invest in Open AI. We'd already invested in a
- 27:56couple of deep learning companies, actually one or two
- 28:01that didn't work out and got sold for essentially Aqua higher
- 28:07kind of prices. But the fact that the companies
- 28:10didn't work didn't dissuade us. Because we fundamentally
- 28:14believed in the thesis, How much was it a bet on Sam Altman or
- 28:17the team versus the technology? It was clearly a bet on Sam.
- 28:22We knew Sam and thought he was awesome.
- 28:26We knew Greg and Ilya and spent a lot of time with the team and
- 28:30we really liked the team there. But more than anything, I'd long
- 28:35held this belief. 2012 when I first talked about.
- 28:41Again, it replace essentially all expertise.
- 28:44If true, then the upper bound is unlimited and it's great for
- 28:49humanity if that happens and and so from my point of view the
- 28:54upside was huge and it was important to make it happen was
- 28:59the structure. I have to give you a hint.
- 29:01There was a article in the New York Times with a writer called
- 29:04Laura Holson in the year 2000. I said at some point in the
- 29:09next. I forget next 25 years.
- 29:12I forget what I said back in the year 2000.
- 29:16I said AI will be so powerful we will have to redefine what it
- 29:20means to be human. That was 2000, so I was already
- 29:24dreaming of less crazy Feeling now.
- 29:28The nonprofit structure of Open AI, were you worried about?
- 29:31Just like the structure of the company, it's still sort of a
- 29:33mystery to people outside how it works.
- 29:35Look, people get hung up on structure.
- 29:38That's the wrong way to look at it.
- 29:40If you're talking about changing the world, who freaking cares
- 29:44about structure? We'll figure that out.
- 29:47We did. You know, when when I worked
- 29:50with Sam that time, that reasonable proposal said makes
- 29:53sense, there wasn't a lot of negotiations.
- 29:55Is there some limit to how much money you can make off the
- 29:57investment? The limited aid because the for
- 30:00profit nonprofit nature of the parent company, which is fine
- 30:05How we make. 5 billion on our $50 million investment.
- 30:09Those are public numbers. I'm fine.
- 30:12That would be a good outcome. Seems more equally importantly
- 30:16in this matter to Sam too. He knew I cared about the impact
- 30:20Open AI would have or in AGI would have.
- 30:23So now now you want to go forward and invest in AI
- 30:26companies. You know Open AI just had Dev
- 30:29Day where it seems like it's coming for every startup.
- 30:32How do you, how do you think about where you can invest?
- 30:35You get more shots on go on. So I was.
- 30:37Just talking to an entrepreneur outside and I said what was the
- 30:40sessions like? He said some of the speakers
- 30:44were more like Chatjavidi could have done that talk brutal.
- 30:54It's true, people speak in generalities and Chatjavidi does
- 30:59that really, really well. But what the the question was
- 31:04yeah, how you what, what start-ups are you investing in
- 31:07sort of the AI world. Right.
- 31:09This is a very tricky time to invest.
- 31:12There's a lot of very high valuations, and I've written
- 31:15about the fact that the very high valuations are bad both for
- 31:19the investors and the entrepreneurs.
- 31:21But just because the valuation is high doesn't mean it's a it's
- 31:28not a good investment. So I'll give you an example.
- 31:32There's a lot of billion dollar valuations and we've looked at a
- 31:35lot of them, in fact most of them.
- 31:38But when adapt, sorry, not adapt.
- 31:40But Replet came along, I thought, because of
- 31:45conversations with the founder, great founder, and a fairly
- 31:50different mission for where they want to be.
- 31:52In two years, we had a series of conversations about the future
- 31:57direction of Replet. And it's pretty different than
- 32:00what replica was last year. And we invested at a billion
- 32:04dollar valuation because I thought they would create
- 32:07something that's now one of my I have 10 predictions I've made.
- 32:11I won't remember all of them, but one of them is within 10
- 32:15years there'll be a billion people on the planet
- 32:18programming. And and what I mean by that is
- 32:22pro writing code by using natural language.
- 32:26That's a large enough change. That it's worth a billion dollar
- 32:31valuation if it's successful. So it becomes more like is Rep
- 32:34let successful or not. And if it is, the valuation
- 32:38wouldn't matter. Most of the people who are
- 32:41chasing somebody else, their valuation matters a lot.
- 32:46We've we've been sort of having a.
- 32:47Run, play, chat. You PD wouldn't give you a
- 32:50nuance. Answer the We've been having
- 32:53sort of running conversation about existential risk.
- 32:56I'm curious what your view is of, you know, open AI, sort of,
- 32:59you know, causing some real problem in in the world.
- 33:02You know, I'm frustrated with the academics who have nothing
- 33:06to do but be academic. And they think about academic
- 33:10risks. The chance of a sentient AI
- 33:14going wild in the next 10 or 15 years is about the same as a
- 33:18chance of the asteroid hitting planet Earth in the next.
- 33:22Whatever years. I think the sentient AI talk is
- 33:27such nonsensical talk and sensible people, like when when
- 33:31I was talking to Fei Fei Lee, you know, she said.
- 33:34There's more immediate risks to worry about.
- 33:37There's real risks that we should worry about, like bio
- 33:41warfare. This morning I was at AAI
- 33:45Security Summit talking about bio risk.
- 33:48Now that's a real risk. There's cyber, cyber risks.
- 33:53There's a larger risk of falling behind China because President
- 33:57Xi has declared he wants to be the source of technological
- 34:02innovation in AI by 20-30. So they will put a lot of
- 34:07resources. Now, those kinds of dictatorial
- 34:11edicts don't tend to work very well, but he has declared that,
- 34:16and I do think. And I've written about this.
- 34:20We are in the techno economic war with China, and we should do
- 34:24everything to win that war. Yeah.
- 34:26So what do you think Biden should do?
- 34:28That? Your team.
- 34:29You know, my recommendation is open up immigration to anybody
- 34:33with talent in this area. Talent.
- 34:39You know, we have an advantage in the talent war and that's
- 34:43what we should do. We should absolutely.
- 34:47Really go after slowing down China as much as we can.
- 34:51So I like the restrictions we placed on China.
- 34:54Do you want to ban TikTok? I absolutely would ban TikTok,
- 34:58TikTok in the nanosecond. Wow.
- 35:01OK, we've got a real disagreement here, Reid said.
- 35:03The opposite. Well, you know, I'm very clear.
- 35:08In the US, companies influence politics in China.
- 35:15Politics influences companies with total control.
- 35:19It's a very different system and I'll be happy to debate Reid on
- 35:22it. There's no question it has
- 35:27available. Capability and it has the
- 35:31ability to be controlled by the Chinese government.
- 35:34Well, to go one round with you, I think Reid's response would
- 35:37just be what he really wants is American companies to be able to
- 35:40operate in China, and by it won't.
- 35:41Happen so you can dream all you want.
- 35:43There's no chance that happens. You know, we've seen this for
- 35:48last 30 years since we opened up.
- 35:51I'm persuaded. The think of the following.
- 35:56I feel like the audience is on your side on this one.
- 35:59The the biggest control point influencing everybody from
- 36:04little kids to adults. At the control point of consumer
- 36:09behavior is TikTok, and we know it works for the Chinese
- 36:13Communist Party. At the other end, with 5G
- 36:17networks from Hawaii, they have the ability to surveil about 60%
- 36:23of the globe's population because their equipment is in
- 36:28the networks. I think we should be very, very
- 36:31worried and not worry about sentient risks.
- 36:35You know, Max Stegmark can do that all he wants.
- 36:37He doesn't need to do anything real.
- 36:41I love it. I love it.
- 36:43Are you know my style? It's that's why we had you here.
- 36:48Are you donating? Are you going to get involved in
- 36:49this political cycle or what's your stance there?
- 36:53You know, I tend to almost always contribute.
- 36:56I've contributed over the last 15 years to both Republicans and
- 37:00Democrats. So I look at the candidates, not
- 37:03the party. And and I like this general idea
- 37:07of a No Labels party also, though I haven't contributed
- 37:12anything. Yeah, Are they going to call you
- 37:13up after this to bankroll? Have you talked to Mansion or
- 37:15anything like that? I'm not a fan of Mansion because
- 37:18he's so opposed to climate change and he's very parochial
- 37:21about Virginia coal, and so I'll never support Mansion for that
- 37:25reason. And and climate is a very large
- 37:29risk for the planet, just like winning an AI is the large risk
- 37:33for the planet. And a few weeks ago I wrote a
- 37:37blog about that about our techno economic war with China and why
- 37:41we came to AI regulation to slow us down.
- 37:45I'm happy to argue that, and I was glad to see much of the AI
- 37:51executive orders were. Considerate of this point of
- 37:54view and spent plenty of time in DC talking to everybody about
- 37:59you're, you're good at forecasting my questions.
- 38:01Yeah. Do you do you think the
- 38:02executive order on AI was OK or? I think it was OK, yeah.
- 38:06Yeah. Do you worry it sort of signals
- 38:08worse regulation or you're optimistic about the?
- 38:11State, you know, it's really hard to tell.
- 38:12We're coming into an election. Yeah, it's going to be about the
- 38:15election, not about what's right.
- 38:17Let's you know that's the pragmatic path that's the
- 38:20equivalent saying otherwise would be the equivalent of read
- 38:23saying hey American companies should be able to do X in China
- 38:26won't happen you know so I I love read and respect him
- 38:33tremendously but you know we don't have to agree on
- 38:35everything. I do think the next year is
- 38:38about getting elected in the next four years will be the
- 38:41important years and it's too early to tell and we have to get
- 38:44people like Lisa Khan out of. Or crazy left wing Kooky Nolina
- 38:52Khan at the FTC. What does that do You hinge any
- 38:56sort of Biden donation on something like that or?
- 38:59You can't do that. You know there's 300 million
- 39:03people in this planet on in this country that he's not going to
- 39:08say I'll donate on condition of X or Y.
- 39:11Those never worked. It's unrealistic.
- 39:14There's lots of. Pressures on lots of people
- 39:17maybe to be practical about it. You know you we love your spicy
- 39:21takes. You've been doing this for 40
- 39:23years. I I interviewed you, you know,
- 39:25many months ago and asked you a similar question.
- 39:27But how much longer do you think you're going to stay sort of an
- 39:30active investor or spearheading? Well, if I life extension
- 39:35efforts by its Peter Thiel or whoever else is doing it work.
- 39:40No, seriously. I have at this saying you grow
- 39:45old when you retire. You don't retire when you grow
- 39:47old. I've seen too many people retire
- 39:50and grow old. So I clearly plan to do and
- 39:54health permitting this for the next 25 years.
- 39:57And then I'll be Warren Buffett's age and he's still
- 40:00doing it. So look, I'll do.
- 40:03This is so much fun and so impactful and keeps me so
- 40:08engaged. I still work easily 80 hours a
- 40:10week easily. You're always telling me some
- 40:14random paper you've read or yeah the I wanted to sort of the last
- 40:18couple questions at the end of this talk.
- 40:20Just looking forward like do you think you know the GPD 5 will be
- 40:25a major improvement like we saw three to four or what?
- 40:28What do you see in terms of, you know, where we are on the S
- 40:31curve and how much improvement you think we're going to see in
- 40:33many of these companies? Look, one, it's very hard to
- 40:36forecast, but I think. What GPT 4 was to GPT 3, and I
- 40:43have no inside information, and frankly even the people there
- 40:47for at Open AI couldn't tell you what GPT 5 will be.
- 40:51But GPT 5I expect we haven't seen anywhere near the limits of
- 40:57AI capability. That's a reasonable assumption.
- 41:00And so when I'm working with start-ups, I try and look at
- 41:04what 5 might have and six might have.
- 41:09Now, what might happen when five helps design 6?
- 41:13So GPD 5 probably will help design GPD 6.
- 41:17You got this exponential effect, and the question for all of you
- 41:21becomes which startups become roadkill in this process?
- 41:26And being thoughtful about that, I spend an incredible amount of
- 41:30time saying which startups should be invest in because they
- 41:34won't become roadkill. In the Are there any categories
- 41:37you would suggest? Well, there's lots of
- 41:39categories. There's lots of very, you know,
- 41:41if we create a billion programmers, you going to create
- 41:44real value no matter what the market.
- 41:46That's a +1. But I've also forecast in 25
- 41:49years we'll have a billion bipedal robots that will create
- 41:53a massive industry larger than today's auto industry.
- 41:57And my bet is we'll have more than a million in less than 10
- 42:01years. And somebody told me I was being
- 42:02too. Pessimistic The last thing I
- 42:05wanted to ask you was just what? What is your stance on open
- 42:08source right now? You'd sort of been like, oh, I
- 42:10want a Manhattan Project, so let me.
- 42:12Finish my talk. I do think in 10 years we'll
- 42:15have free doctors, free tutors for everybody, free lawyers so
- 42:20they can access the legal system.
- 42:23I'm in the process of writing A blog that is good note to finish
- 42:27on. You will answer the open source
- 42:29question Will AI lead to dystopia or utopia?
- 42:33Too many people are looking at the dystopia angle of this ten,
- 42:381% probability of something bad happening and ignoring the
- 42:42benefits to humanity of AI and the 10,000 start-ups that are
- 42:46going to do truly wonderful things and my job is to help
- 42:51them navigate the path forward. Coming back to your open source
- 42:55question, I'm very much against open sourcing in AI.
- 42:59Keep in mind, keep in mind. We were the firm back in the 80s
- 43:06that literally started the open source movement at Sun.
- 43:09NFS was the first major piece of software that was open source.
- 43:16Linux came later. So I was very much an open
- 43:19source fan and what it adds to creativity.
- 43:23We are the investor first investors in GitLab.
- 43:25It's it's the only you know. People forget GitHub wasn't open
- 43:30source, it was for open source software.
- 43:32GitLab was both open source and for open source software.
- 43:36Huge fans of open source. But in this techno economic race
- 43:40with China, we will help them and we can slow them down by 6
- 43:44months or a year. I think it's good for America.
- 43:48Vinod Khosa always entertaining and join.
- 43:50Thank you. Thank you.
- 43:55That's our episode. Thanks so much for listening.
- 43:58Shout out to Max Child and James Wilsterman, my Strobel Valley AI
- 44:01Summit Co host. Thank you to Riley Kinsella, my
- 44:05Chief of Staff, Gabby Caliendo Ed Volley, who's been
- 44:08instrumental on putting the conference together.
- 44:11Thanks to Young Chomsky. For the theme music, please
- 44:13like, comment, subscribe on YouTube, give us a review on
- 44:16Apple Podcasts and please subscribe to the sub stack
- 44:20newcomer.co. Thank you so much.