YT YouTube2Text

YouTube2Text → Transcripts → The Frontier Ideas a16z Wants to Fund Now | Marc Andreessen, Ben Horowitz & Chris Dixon

The Frontier Ideas a16z Wants to Fund Now | Marc Andreessen, Ben Horowitz & Chris Dixon — Transcript

by a16z · 14,230 words · 1,988 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00We're announcing today that we're
0:01expanding the ACC crypto fund to A16Z
0:04frontier. This is the biggest delta
0:06between what the early adopters
0:08understand as possible. Now that gap is
0:10as wide as I've ever seen in my career.
0:11>> More than categories, more than
0:13products, we invest in entrepreneurs and
0:16so many of the most interesting
0:18entrepreneurs are outer planet. Startup
0:21formation is going to skyrocket from
0:22here.
0:22>> By the way, it already is. We used to
0:24ask the question, is this idea too
0:26ambitious? We're much more likely to ask
0:28is this ambitious enough because what's
0:31possible is so spectacular.
0:33>> Had I had astronaut at Netscape in 1994
0:36like oh my god Ben like what would you
0:38and I have been able to do at Netscape?
0:40>> We would have built the whole damn
0:41internet.
0:41>> You were 18 today how would you think
0:43about what you might [music] study
0:45stealing marks line but I think there's
0:46one thing to do which is study.
0:49>> Chris we have a big announcement today.
0:52>> Uh we're expanding the mandate. Uh I'll
0:54let you do the honors.
0:55>> Yeah. So, uh, yeah, we're announcing
0:57today that we're expanding the ACC
0:59crypto fund to A16Z Frontier. Um, the,
1:03um, the idea is, I say a couple of kind
1:06of motivations behind this. Um, one is,
1:09you know, crypto, um, has in some areas
1:12worked really well. Um, stable coins
1:14have become kind of a normal part of the
1:15tech landscape used by companies like
1:17Stripe and sort of almost, you know,
1:19every fintech company now kind of
1:20consider considers it part of their
1:22toolbox.
1:23um Wall Street firms are moving heavily
1:25into crypto uh with tokenized stocks and
1:28other kinds of assets. It's a growing
1:30fast growing category. Um but as a
1:33result, the lines have kind of blurred
1:34um which you know we find ourselves
1:36asking is Stripe a crypto company? Is
1:38you know somebody uh many of these
1:40companies where they're sort of sitting
1:42at the in the intersection of let's call
1:44it fintech and Wall Street and crypto.
1:47Um, and so it's really become kind of,
1:49you know, whereas, you know, years ago
1:51it was sort of its own world with its
1:53own companies and its own investors,
1:54it's the lines have blurred and it's
1:56become kind of more part of the um tech
1:58landscape. So that's number one. And
2:00then number two, and I I can let Mark
2:02and Ben talk about this more. I think we
2:03felt that at the firm there's a feeling
2:06that the um other kind of groups are are
2:11um have so many interesting
2:12opportunities in their core areas uh
2:15that uh what we kind of my my partner
2:18Martin calls outer planets, meaning the
2:20sort of the metaphor I think comes from
2:22Firefly, the TV show. It's the idea the
2:24inner planets are kind of the kind of
2:26the more core planets and there's kind
2:27of these wackier outer planets. Um that
2:30uh outer planets which which
2:31historically we've done a lot of kind of
2:32both types of investing. Um I think the
2:34feeling was um that because they're so
2:37busy with so many interesting
2:39opportunities and doing an amazing job
2:40with that um that there was an
2:43opportunity to kind of uh there's maybe
2:45a gap so say in in some of the coverage
2:47that um there and then I think just I'll
2:49just speak personally um you know I kind
2:51of see myself and my background
2:54I see Frontier as an area I'm
2:56comfortable with. this sort you know I
2:57joined the firm in 2013 and um we you
3:00know came in and with that kind of
3:01mandate I talked to Mark and Ben and I
3:03said hey I want to do this crazy stuff
3:04and they they were like great um and so
3:07you know that led to a couple of themes
3:09at the time and they're all kind of what
3:12I would call kind of second order
3:13effects like you know you had
3:15smartphones obviously started in 2007
3:18and by 2013 had gotten mainstream and
3:20there were all these kind of first order
3:22effects like building apps on top so
3:23like the firm was an investor in
3:25Instagram and you know, incredible uh
3:27product built on smartphone. But a
3:29second order effect was that you had the
3:31components of smartphones. You had, you
3:33know, sensors and chips getting better
3:35and better and maybe people could
3:36repurpose those in different ways. And
3:37so that led to I think Mark was the one
3:40who's first into drones as I recall and
3:42got me into them. But the kind of the
3:44thesis behind it beyond just they're
3:45cool is is that now that you have this
3:48supply chain of um you know all these
3:51great parts that entrepreneurs can kind
3:52of recombine them and build you know we
3:54let we invested in Oculus as an example
3:56and that was really kind of a
3:57reconfigured smartphone. Um Skyio that
4:00was one of our investments from back
4:01then that was you know at the beginning
4:02was kind of a reconfigured smartphone.
4:04It's gotten much more advanced now. uh
4:06crypto I think at the time was kind of a
4:07downstream effect to me of sort of the
4:09open source software movement morphing
4:11into services and also kind of a result
4:14of social media where you could have
4:15these communities form without
4:16mainstream acceptance and build build uh
4:19you know and kind of coales around these
4:21these blockchains. Um so for me
4:24personally I'll just say this is sort of
4:25I feel you know kind of a continuation
4:27of long area of interest and I think the
4:29firm has you know there's many other
4:30investments the firm has had in these
4:31kinds of frontier areas. Um, so we we
4:34felt I think it's kind of consistent
4:35with things we've been doing for a long
4:37time.
4:38>> Uh, Ben Mark, let's hear from you. I
4:40know this was a joint uh effort and
4:42something you guys were very excited
4:43about and felt like filled a filled a
4:44gap or really expanded the the mandate
4:46of the firm. Ben, why don't you add your
4:48take?
4:49>> Yeah. So, look, I think that there's
4:52areas um that are really novel. So, it's
4:57taking it it's looking at, you know,
4:59what's possible technologically and
5:01problems in the world and they're just
5:04the first ones in it. And it may end up
5:06being a category. It may end up being a
5:08successful company. It may, you know,
5:10blow up. Um, but you know, there's a
5:13class of these things. And you know what
5:16we've done typically or or historically
5:19is
5:21was first like Chris would just look at
5:24these new areas himself. Um and then uh
5:28kind of as we uh identified new areas,
5:32we'd create a a vertical behind that
5:34area. Uh and what we found with our
5:37vertical strategy which has worked very
5:39well is that you know there's still new
5:43things that fall outside of the
5:44verticals and the process of turning
5:48that into uh a practice um is a little
5:54bit undefined. Not that we haven't done
5:55it but it's a little bit undefined. And
5:58so by Chris doing this, we kind of will
6:01have the motion to invest in the very
6:04most interesting founders who are outer
6:07planets and then you know if that works
6:10you know maybe that becomes a whole fund
6:13at some point uh or or maybe frontiers
6:16turns into that thing because it gets so
6:18interesting like that could be as well.
6:20And so I think that you know it's very
6:22exciting for us because it kind of uh
6:25fills a hole in the architecture and
6:27then also
6:29you know we really at the end of the day
6:31more than categories more than products
6:34we invest in entrepreneurs and so many
6:37of the most interesting entrepreneurs
6:39are outer planet. So this gives us a a
6:43real way to get to that.
6:45>> And to me this seems like a you know
6:46continuation of what we were talking
6:47about at the at the last offsite which
6:49is which struck me because you know as
6:51uh as ambitious as the firm has been as
6:53early as the firm has been as as big of
6:55the bets that the firm has made. You
6:56know the big focus of the offsite was
6:58how do we bet even bigger or the the the
7:00only regrets we have are not even going
7:01even harder even earlier e even bigger.
7:04and and Chris, you're someone who, you
7:06know, has made bets in some of these
7:08emerging categories, but uh it just
7:09feels like a continuation of of that of
7:11that message. We're going to spend most
7:12of the time in the future, but while
7:13we're talking about history, I have to
7:15to ask one historical question, which
7:16is, you know, Chris, I believe, is the
7:18longest standing consecutive uh partner
7:20at the firm, you know, uh besides Benark
7:22having joined in 2013. Uh, I'm curious,
7:24Benmark, when you uh, you know, met
7:26Chris, what did you see in him that made
7:28it clear that, oh, we have to have him
7:30at at Acing Z? And did you, uh,
7:32anticipate the type of, uh, career he
7:34would go on to going to have? Ben, I'll
7:36ask you to start.
7:37>> Look, I I I think originally, um, we
7:40just knew Chris as a very smart guy and
7:43we really liked his ideas from his blog.
7:46I mean, like that that that's really
7:48what it came down to. He was a really
7:49good writer. um his ideas were very
7:52original and you know we we were looking
7:56for like somebody smart who could do
7:57that. And then the other thing we liked
7:59about him is he had real experience as
8:01an entrepreneur. So he knew kind of like
8:03what that journey felt like uh which you
8:06know was very very important to us in
8:08the early days to to set the culture of
8:10the firm the way we wanted it. um you
8:14know over time he's had probably
8:18the most original investment ideas that
8:21we've had in the firm and so and we've
8:23kind of copied those and uh multiplied
8:26them across. So you know one one of his
8:29early things was look we're not really
8:30looking for good ideas that look like
8:32good ideas. We're looking for good ideas
8:35that look like bad ideas. Um because and
8:38then
8:40a bad idea but where an entrepreneur has
8:43a secret that actually makes it a good
8:45idea. Uh and that that that framework
8:48was extremely useful for us in kind of
8:50understanding like what was going to be
8:52truly unique and have massive upside
8:55versus something that was going to get
8:57where all the profits were going to get
8:59competed away. Um, [clears throat] and
9:01so, you know, and he's he's had quite a
9:03few uh excellent ideas like that. So,
9:06you know, it's it's been great working
9:08with him over now like way more than a
9:12decade.
9:13>> Yep. Maybe I could add um also it's not
9:16just Chris as special as he is, it's
9:17also Chris's team. Um and um the the
9:20thing that I think people may not
9:21appreciate maybe [snorts] hard to see
9:23from the outside is just you know crypto
9:24has this you know crypto's like been so
9:26sexy in so many ways and it's been so
9:28you know it has this sort of flashy
9:29elements and you know it's a cultural
9:31you know kind of phenomena kind of
9:32moments but at the core of crypto it's
9:34it's a it's a technology it's a very
9:35deep technology and it's very complex
9:36technology um and you need to have like
9:39really deep technical expertise you know
9:41kind of coupled with like really good
9:42business judgment and really good market
9:43judgment in order to successfully invest
9:45in crypto which which which we which
9:46we've done. Um, including, by the way,
9:49including like, you know, literally
9:50Chris Steam has had like, you know, math
9:51professors and, you know, cryptographers
9:52and like people who are, you know, very
9:54very deep into the theory, you know,
9:55game theory, uh, you know, things like
9:57that. Um, you know, very deep software
9:58concepts. Um, and so it's it's one of
10:00the things I'm really excited about is
10:02in addition to Chris, his his colleagues
10:03and team members and the team that he's
10:05built, uh, now being able to work on a
10:07work on a broader set of topics.
10:09>> Yeah. And to that point, I mean, so so
10:10thank you. But I think that also the
10:12Yeah, the team I think the team's
10:13amazing. And I would also mention one
10:14other thing besides the technical side
10:16um we've invested very deeply over the
10:18years in policy.
10:19>> Yeah.
10:19>> And regulatory and I and I believe that
10:21a lot you know if you just self-driving
10:23cars, drones, AI itself is now becoming
10:26a political issue. Um I think more and
10:28more of these cutting edge technologies
10:30will have kind of a policy angle and and
10:32we believe that that that expertise that
10:34they've built kind of in the crucible of
10:36of the kind of crypto journey um will be
10:40uh useful for many other entrepreneurs.
10:42>> Yeah. Yeah. I agree. Well, on that end,
10:44no, we'll do one last historical
10:45question. Uh, you know, you pioneered
10:47not just some of the investments, but
10:48also some of the ways in which the firm
10:49works, I the policy arm and when you
10:51joined in 2013, you know, could you have
10:53imagined that the firm would have grown
10:55in this way? What was what were your
10:56thoughts about where things were going?
10:57>> Well, I always knew Mark I mean Mark has
11:00a you have a big uh poster of JP Morgan
11:02on your wall in your office. [laughter]
11:04>> Yes.
11:05>> And so these Mark and Ben do not lack
11:07for uh for you know drive and ambition.
11:10>> It's a famous photograph. It's a very
11:11it's a very famous photograph. It was it
11:13was he hated JP Morgan.
11:14>> It's inspiration inspiration
11:17>> inspiration. For people who haven't seen
11:19the photograph, you can Google it's a
11:20very famous one is JP Morgan hated to
11:21have his picture taken for this he had
11:23this this condition. He had this anyway
11:27he has this nose related issue. Anyway,
11:28so he hated having his photo taken. Uh
11:30this was pre- antibiotics, let's put it
11:32that way. Um and so he he was grouchy
11:34most of the time and so he has photo
11:35taken. So that but he they needed a
11:36formal portrait of him and so he finally
11:38literally you put him and set him down
11:39and the the the photographer like saw
11:41the whole thing coming and so it it's
11:42lit in a way where he's sitting in
11:44darkness um just absolutely giving the
11:47camera the stare of death and he's
11:49holding it looks in the photograph like
11:51he's holding a carving and and and what
11:53it is is the edge of the chair lit in a
11:56very specific way and it's like very
11:57clear the photographer like
11:58intentionally did that.
12:00>> Yeah. So that's the photograph that was
12:01inside my office door for
12:03>> You looked very very scary. very
12:05>> Yes. Exactly. Exactly.
12:06>> The other we took a lot of inspiration
12:09from him.
12:11>> Uh but one of the other things that that
12:13that we have in the firm that's actually
12:15in the culture document from him is we
12:18do first class business and then only in
12:20a first class way, which is, you know,
12:23it's like a cool phrasing of it, but
12:24that that when he said it, he was on
12:27trial. Um and they were kind of accusing
12:30or he was in court. They were accusing
12:32him of kind of nefarious things and he
12:35was like, "No,
12:37we only do first class business and men
12:40only in a first class way." Which, you
12:43know, Mark and I really loved that
12:46attitude about like you're you're
12:49definitely not going to tell me who I
12:50am. I know who I am and I'm going to let
12:52you know who I am. And that kind of
12:54thing I I think is really important in
12:57venture capital to like be principled
13:00and people know where you're coming from
13:02and be able to trust you and that kind
13:04of thing. So, so we got a lot from him.
13:06>> Yeah. So, JP Morgan.
13:08>> Yeah. And so, I mean I have never I mean
13:10I think a couple things I'll say is like
13:11I mean look like any venture there were
13:13ups and downs and you never you know I
13:15could I have predicted everything. I
13:17think things generally went really well
13:19and you know I didn't obviously know
13:20that all all along the way but um
13:23obviously know that Mark and Ben are
13:24incredibly ambitious. I've never I it's
13:26funny sometimes I'll have some idea and
13:28like some crazy thing we want to do and
13:30people like what are Mark inventing like
13:32I've never in my not once have they ever
13:34if anything have they ever pushed back
13:36if anything it's always can you double
13:37it can you double down can you get more
13:39aggressive um so just you know and then
13:41just it's a very entrepreneurial place
13:42and I think this has shown over the
13:43years of like how many this this is a
13:45good example right here um that you know
13:47when we you know we sort of adapt um see
13:50what's happening in the world change
13:51things around we went through a whole
13:53kind of restructuring of the of the
13:55firm. Um, uh, you know, bring in new
13:58ideas. Sort of anything's anything's
14:00fair game as long as you can kind of
14:01defend it. Um, and so, you know, that I
14:04that's also why I've enjoyed working
14:05here. It's a very entrepreneurial
14:07culture. I think I think also what Ben
14:09said is very important, the culture of
14:10the firm. Um, that's why I was initially
14:13kind of attracted to it. Having been a
14:14customer of venture capital before that
14:16and experienced the product, I
14:18understood that there could be a better
14:20product. Um [snorts] u but that and that
14:22that's I think everyone literally
14:24everyone uh that I know at this firm
14:27like comments on the culture and how
14:28important that is um and how they you
14:30know enjoy working here and and feel
14:32like they're doing some of the best work
14:34of their life. So
14:35>> um so I think that's all like a very you
14:37know there's obviously have to make good
14:39investments and this and that but I
14:40think it all really starts with the
14:41people and the culture.
14:43>> Well it's funny that you mentioned um
14:45it's related to the culture. It's funny
14:46you mentioned ups and downs because you
14:47know from the outside looking in it
14:49feels like it's all all all ups and
14:50>> that's how all startups are is like you
14:52know like the successful startups always
14:54look you know look great from the
14:55outside as someone who's I think we can
14:57all comment as people who have been
14:59>> in inside of startups basically I don't
15:01know of a startup that wasn't a
15:03show at some point or at multiple points
15:06>> it also speaks to the culture I've heard
15:07from all three of you guys you know
15:08things like you know never complain
15:09never explain you know happy warrior um
15:12and so yeah it's just a testament there
15:14>> let's uh let's let's go to the frontier,
15:16let's go to the future. So, sometime
15:18there's this phrase in venture, you
15:19know, being too early is as bad as being
15:21uh, you know, too late or something. Um,
15:23you've been early a lot. You've been,
15:25you know, mostly right or or right in
15:27very big ways, but you've also, you
15:28know, been too early at times. I'm
15:30curious how you think about technical
15:31risk or or the idea of being too early
15:33in in our era now.
15:34>> Yeah. I mean, we talk about this a lot.
15:35Like I think we I think I can speak for
15:37Mark and Ben, they can comment, but we
15:39believe that basically every good idea
15:41that we've ever seen that involve
15:43technology has eventually worked or we
15:45believe will work. [laughter] Okay, if
15:47it hasn't worked yet and and the
15:49important qualifier, good idea. There
15:51are bad ideas, but um it's just kind of
15:53remarkable, especially having lived
15:55through the internet um how many things
15:58um just, you know, almost everything
16:00that ended up working had a precursor
16:02that didn't work and maybe was 10 years
16:04too early. Um, and so so much, you know,
16:06and that's the hardest thing I think in
16:07technology is to predict the timing of
16:09these things. Um, and and there's just
16:11so it's so complicated, right? There's
16:12the technology, there's the culture,
16:13there's the business environment,
16:15there's the political environment.
16:16There's so many factors that go into
16:17these things and having it all line up.
16:19Um, so, you know, and so I I would say
16:21with, you know, I think generally at the
16:22firm, I would say specifically in the
16:24Frontier Fund, we want to air on the
16:26side of being too early. We understand
16:27that sometimes things will be too early
16:28and we're we're okay with kind of cool
16:30category risk. Um, some categories may
16:33not work. Look, I mean, even things that
16:35sort of worked didn't work. Like, you
16:37know, we sold Oculus, got bought, but
16:39the reality is like it's now 13 years
16:40later. I still believe very strongly in
16:42VR. Um, we'd love to in make VR
16:44investments, but it's 13 years later and
16:47it's not, you know, it hasn't got as
16:48mainstream as we would have liked. So,
16:49it took long. It turned out there's just
16:51many more hard technical problems to
16:52solve. I kind of believe I don't know
16:55what Mark believes. I think that there
16:56that with the stuff we're seeing in AI
16:58now, you know, you see the first sort of
16:59tipping happening, first software
17:00engineering getting dramatically better
17:02than math. I think there's a scenario
17:04and I I believe this and I kind of want
17:06to invest behind it where we're we may
17:08be pulling forward 20 years of things
17:11that you know our whole mindset of how
17:13we think about technical risk and oh
17:15that's a science project not an
17:16engineering project I think we have to
17:18re-evaluate that whole frame right now
17:20um because I think things you know the
17:23sort of the you know if you look at
17:24these the growth of I mean math is not
17:26the last thing that's going to change
17:27you know you're about to have physics a
17:28lot of stuff's going on in hardware
17:29right now
17:30>> I think it's dramatically I think you a
17:3230 person team can credibly build like a
17:35full stack device company in a way that
17:37you might have needed a thousand people
17:39or something a decade ago, right? And I
17:41think and you have these very, you know,
17:43very um thoughtful uh deep entrepreneurs
17:46who are looking for how do I take these
17:50new capabilities and do dramatic things
17:52and you know so some of it's tactical
17:54like how do I create more defensibility?
17:55How do I You don't want to be two feet
17:57in front of the AI frontier and get sort
17:59of washed over by the title wave, but a
18:02lot of it's about just you have these
18:03amazing new capabilities and people have
18:04only begun to explore what you can do.
18:06Um, and so it I don't know it feels like
18:10an especially exciting time. Um, and we
18:12don't it's it um it's it's Yeah. So I
18:18don't know if you
18:19>> Yeah. So I would add so talk about AI as
18:21sort of a catalyst. Um
18:24so this is the biggest delta between I
18:27would say what the early adopters
18:28understand is possible now uh versus
18:30what kind of everybody kind of normal
18:31people understand. Um and I think that's
18:34you know true at the individual level
18:34but it's especially true at the company
18:36level.
18:37>> Um and the early the AI native early
18:39stage startups have just a completely
18:40different view on what they can do with
18:42AI than the than the than sort of you
18:43know typical incumbents in almost any
18:45industry. Uh almost without exception
18:47and and I think that that gap is as wide
18:49as I've ever seen in my career. And then
18:50it that gap seems like it's gapping
18:52further and further out which is AI is
18:53just racing ahead in capability even
18:55week by week. Um and so the leading edge
18:58entrepreneurs and technical people we
18:59deal with like they're they're they're
19:01getting they're just they're they're
19:02just by the week they're getting sort of
19:03steadily more advanced um than than than
19:05than a typical incumbent. Um so it's
19:08it's it's the biggest alt I've ever
19:10seen. And if you kind of extrapolate
19:11from that like you know if you if if
19:13that's what happens um then there's just
19:15there are going to be a lot of
19:16industries that are really up for grabs
19:17in a fundamentally you know kind of more
19:19more profound way than maybe has been
19:20the case in the past and then the and
19:22that's going to be because the founders
19:24are going to have a level of superpowers
19:25coming from the use of these tools and
19:27these technologies that they just simply
19:28haven't had in the past. I mean, you
19:30know, had I [snorts] had, you know,
19:32Astra in Netscape in 1994, like, oh my
19:36god, like, [snorts] you know, like Ben,
19:38like, what would you and I have been
19:39able to do at Netscape?
19:42>> We would have built the whole damn
19:43internet. I mean, absolutely incredible.
19:45It would absolutely incredible. And, you
19:47know, as Chris said, like we're seeing
19:48this, we're seeing entrepreneurs come in
19:49with like a with plans of a level of
19:51depth and sophistication, you know,
19:52multiple parts
19:53>> connecting together.
19:55>> By the way, we would have gotten crypto
19:58done then. Oh, 100%. Yeah. Yeah. Like,
20:00by the way, in like two weeks, like we'
20:02have streaming video a lot sooner. I
20:03mean, like, all kinds of things would
20:04have happened. And and then, by the way,
20:05hardware, I mean, we never even, you
20:06know, bit off hardware, but, you know,
20:07had we been able to, there's all kinds
20:09of things we could have done. Um, and
20:11so, um, I I just like I think that the
20:13leading edge entrepreneurs now are going
20:14to have like capabilities and which
20:16means they're going to have business
20:17plans that are like a lot more, you
20:18know, they're going to have product
20:19plans that are a lot more technically
20:20intensive and complex. They're going to
20:21have business plans that are more
20:22ambitious and complex. Um, they're going
20:24to be building companies that just like
20:26ramp much much faster. um uh you know
20:28the the scope of the product that they
20:29build and then you know hopefully the
20:31the scope of their business. Um and then
20:32and then and then again this goes right
20:34back to like th those entrepreneurs are
20:36going to color way outside the lines
20:37like they're they're not going to cons
20:38we already see this. They're not going
20:39to think of themselves as like I'm just
20:41a software company or I'm just a
20:42hardware company or I'm just a whatever
20:43or even I'm just an AI company or
20:45anything. They're going to they're going
20:46to think very broadly and expansively
20:47about exactly what you know how many
20:49different areas they want to touch and
20:50how they want to combine things. And so
20:52the the sort of you know conceptual
20:53artist kind of aspect of this is going
20:55to like just like blow way out. Yeah, I
20:58was going to say, you know, we used to
21:00kind of ask the question, is is this
21:02idea too ambitious? And I I don't think
21:04I think we're much more likely to ask
21:06this is this ambitious enough
21:08>> um because you know what's possible is
21:11so spectacular. And if you do something
21:16moderate like it, it probably won't be
21:19differentiated enough. And by the way,
21:21look in incumbents, you know, we spend a
21:23lot of time, I have a lot of friends in
21:24incumbent companies, by the way,
21:24including incumbent tech, including
21:26incumbent companies in Silicon Valley.
21:28And it's like, [snorts] you know, we
21:30wish them all the best. And, you know,
21:30we, by the way, we have lots of
21:31companies that would like to arm them up
21:33with all these all these new tools also.
21:35Um, and so, you know, so no doubt some
21:36of them are going to be able to adapt,
21:37but like, you know, if any of them are
21:39listening right now, I would say like
21:40the clock is ticking. Like it it like
21:43>> stuff is moving.
21:44>> I mean, like I I I I am basically I
21:47mean, I'm full-time. I'm basically
21:49drowning um in the the in the in the in
21:51the in the with the pace of development
21:53right now
21:53>> and the speed with which things are
21:54changing. Um and I'm trying to keep up
21:56with all of it and it's nearly
21:57impossible.
21:58>> Like it it the yeah the the the the
22:01supermpowered 24y old with AI is going
22:03to just achieve like amazing things in
22:05in the next decade.
22:06>> And this is a question we were asking
22:08right before this which is you know is
22:10AI more uh sustaining or disruptive in
22:12terms of like will more of the gains go
22:14to incumbents or or startups? And it
22:15seems so far it's both in some sense
22:17which is some you know the companies
22:19like you know Meta and others have you
22:20know uh sort of adjusted very well and
22:23and also their net new companies that
22:24are you know u almost trillion dollars
22:26in you know within a decade and it's so
22:29is that your read that that's
22:30effectively both or how do you think
22:32about the sustaining disruptive? Well,
22:33the incumbents that are adapting are
22:34actually the incumbents I think for the
22:35most part that are actually going
22:36directly into AI, you know, and then
22:38there are others and then there are
22:39certain, you know, there's certain
22:40companies certainly in tech that are on
22:41the leading edge even if they're not
22:42prim primary AI developers. But at least
22:44right now the pattern is is is not that
22:46I mean it's it's just really
22:48actually there's a very specific thing
22:49that I think maybe it's a topic for a
22:51different podcast. There's a very
22:52specific thing I think is happening
22:53which is I think many companies are not
22:55adopting AI while their employees are.
22:57>> Um and their employees are and their
22:59employees are not necessarily telling
23:00the man management what's happening. And
23:03I and by the way I'm going to guess
23:04every single person listening to this
23:05podcast right now is thinking oh yeah
23:07that's what I'm doing. You know I'm
23:08using AI you know to you know do all you
23:10know do all my work or at least let's
23:12say fact check all my work uh you know
23:14proof read all my all my all my uh
23:17emails
23:18>> um even if not have them written
23:19outright. Um, and so there's this weird
23:21thing where there's something about AI
23:22right now where it's like it's much
23:24easier for the individual to adopt than
23:25it is the institution. Um, and by the
23:27way, and and again it goes back to like
23:29our role in the world, which is like
23:30those some of those individuals are
23:31going to figure out like, oh, with this
23:33tool, I can now like start my, you know,
23:35it's now easier than ever for me to
23:36start my own business.
23:37>> Um, and so I I think startup formation
23:38is going to skyrocket from here. Um,
23:40because these
23:41>> by the way, it already is.
23:43>> Oh, yeah. I mean, we're seeing it. I
23:44mean, we're seeing it 100% for sure. Um,
23:47so anyway, just you kind of go go back
23:49to the theme of this is just I I I'm
23:50completely convinced like we're going to
23:52see a wave of like much more technically
23:53aggressive, much more business
23:54aggressive, much more uh expans
23:56expansionist creative original thinking,
23:59many more combinations of of of things
24:01that we would have considered to be
24:02separate things in the past. Um, and uh
24:05and uh we're going to see yeah founders
24:06was just like an in and teams was just
24:08like an incredible level of skill level
24:10uh incredible level of skill that that
24:12they otherwise might have needed, you
24:13know, 30 years to achieve or something.
24:14They're going to have like right out of
24:15the gate. And by the way, the other
24:16thing I'd say is that the the
24:17Christensen disruptive innovation frame
24:20assumes that sort of the technology
24:21world is relatively contained and that
24:24these are companies fighting with each
24:25other. Yeah,
24:26>> I think it's probably the wrong frame
24:28right now, which is it's really sort of
24:30people who are techn technologically
24:32modern going out into other industries
24:35and just changing the economy more
24:36broadly. Yeah. Right. Especially with
24:38these kind of full stack companies, you
24:39go and you create a full, you know, like
24:42uh we just announced our investment in
24:44Adams, like they're going and they're
24:45building like a full stack kind of food
24:47service and and other things, right? And
24:50>> um so so I think that the you talk about
24:53this sometimes like the market has sort
24:54of gone from, you know, the 10 big tech
24:57companies fighting with each other to
24:58the 10 tech company or 10 and all the
25:00startups going against the the other
25:03Fortune 500 or the rest of the economy.
25:06>> Yeah. Yeah, that's right.
25:08>> I'm I'm curious where where you guys
25:09think we are in the in the development
25:10or how you because it feels like you
25:12know in the internet era um application
25:14uh you know we're sort of you know the
25:16biggest companies um you know Facebook,
25:18Amazon etc. Um but in in sort of this
25:21era it feels like you know Nvidia and
25:23some of the more hardware companies are
25:24are taking a lot of the margin or some
25:26of the bigger companies. Um how do you
25:27think about you know where we are are we
25:29in the sort of applications or consumer
25:32like how do how do you think about that?
25:34I think Mark would say all all of the
25:37above and I think a lot of people think
25:38that actually right now it seems like a
25:40lot of stuff is working you know
25:41obviously the big the big AI labs but
25:44you know open source is growing very
25:45very quickly there's ser obviously the
25:47serious efforts in China the serious
25:49efforts in the US um a lot of
25:51application layer things are doing
25:53really well um that you know so you know
25:56maybe at some point you you know
25:57obviously over time that shakes out and
25:59the kind of value chain there's certain
26:00things that certain layers that do
26:02better than others but um
26:04>> and then the other I think it's just the
26:05the it's still it's like it's kind of
26:07crazy. You have the demand for a lot of
26:09this stuff uh is is still very early by
26:12most metrics. Um
26:13>> Yep.
26:14>> I think your team just put out some
26:15charts on this. It's like 3% of you know
26:17measured enterprise use or something
26:19like this.
26:20>> Um and so and it's just hard and it just
26:22seems like a you know the market for
26:24intelligence is very very large and it's
26:26only getting started. So
26:29>> um yeah, I don't know.
26:30>> Yeah, I agree. And one of your your
26:32points also is you like you mentioned
26:34that every new technology uh has this
26:36sort of first this you know phase where
26:38people think about what are the old use
26:39cases how they just do the new
26:41technology and here u and then of course
26:43people figure out net new cases that are
26:45native to that technology and so where
26:47are we there
26:47>> yeah I mean like sort of the classic
26:49example I like to use is the internet
26:50right where in the '9s you had sort of
26:52people taking existing behaviors like
26:54you might order a Sears catalog or
26:56something and instead Sears puts it on
26:57the internet and of course when they do
26:58that you want to use the internet
26:59version it's easier and a lot of the
27:01early internet was like that and then
27:02you had you know in the 2000s you had
27:04YouTube and they come along and they put
27:05up a empty website and they say well
27:07someday there's going to be these
27:08YouTube influencers and that's going to
27:10populate the so they were betting on
27:11they weren't just betting on the
27:13technology and taking something they
27:14knew would work right they were betting
27:15on a on a on a cultural change or
27:17behavioral change um so in some ways
27:19they were taking more risk because they
27:21were betting on you know the technology
27:22and the and these yeah and these broader
27:24changes um but that's where you got the
27:26really kind of world changing stuff is
27:28in the 2000 social media you know all
27:30all And then of course mobile
27:32accelerated it and and I think that's a
27:34my own view is that's a kind of pretty
27:35common pattern and just just human
27:37nature right you you build something and
27:39the first thing you do is you do these
27:41things you already know um in a better
27:43way and I think that's what we've seen
27:44in AI so far where you have you know
27:46people already coded people already did
27:47math um people already did knowledge
27:49work people already were used to like
27:51search engines and sort of semi-hating
27:53and they're doing those things now um
27:55but we've only begun to explore the
27:57space of like what are the radical
27:58things that that are new behaviors new
28:00activities, right? Things that people
28:02have never done before. Um, it seems
28:05like we're probably at that tip. I
28:06believe we're kind of at that tipping
28:07point where we're going to start to just
28:09sort of get see the crazier, weirder
28:10stuff, which I personally am very
28:12excited about and and find that I always
28:14find those waves in technology the most
28:16personally exciting. They're they're
28:17kind of more um and and there's sort of
28:19more layers of risk. going back to
28:20frontier like you can find some things
28:22will be you know kind of more direct
28:24shots on goal and other things will be
28:26these kinds of bank shots where you have
28:27to have multiple things go right and
28:29maybe there's more layers of risk but
28:31maybe if you're right there's more
28:32upside as well
28:34>> yeah I'm I'm really excited in social
28:36used to be in the 2000s like every few
28:38years there was a new major network
28:39there was Facebook you know what there
28:41was snap tik tok and and then it felt
28:44like there was a little bit of a darth
28:45on the social side but it with AI it
28:47does feel inevitable you know and been
28:49dating with Tinder of course and there
28:51hasn't been something in a while. It
28:52feels inevitable that hey if there's
28:54this you know interface or device that
28:56I'm you know giving all my information
28:57to that there's going to be some way
28:58that these are these are hooked up and
29:00enabling a whole new set of social
29:02experiences.
29:02>> Well I mean you can talk about this Mark
29:04but we you know we're headed to a world
29:06where it looks like the main the the
29:08most numerous participants on the
29:10internet will be agents and not people.
29:12>> Uh we're early phases of that but it
29:14seems somewhat inevitable. you have
29:15these products like Muse and Instinct
29:17that are sort of utility proxy agents
29:19for people. Um, and then you're you're
29:22going to I think you see a whole
29:22different sets of categories of these
29:24agents and they'll become the main
29:25participants on the internet. And then
29:27there's questions of like, you know,
29:29will they need their own ways of
29:31communicating and transacting and uh
29:34sharing knowledge and doing all the
29:36things people do. Um, and so it's
29:39probably going to be a whole series of
29:41of innovations along those lines. rent a
29:44human.com. [laughter] Yeah,
29:46>> rent a human. That's like inverts the
29:48model, right?
29:50>> Moltbook was cool, too.
29:51>> Yeah. Yeah, totally. That's Well, you
29:53know, Yeah. Molt book was the public
29:55public kind of version and then it
29:56turned out the agents figured out how to
29:58do their own
29:59>> uh the mold book in the hugging face.
30:01So, yeah, it's going to be a lot more.
30:03Yeah, the agent the agent ecology online
30:04is going to be spectacular. People are
30:06going to plug into that in all kinds of
30:07ways. Yeah, Mark, I'm curious if there
30:09are things that you're tracking either
30:10in in social or consumer or or even just
30:13this frame of like the skuorphic, you
30:15know, things that that sort of the, you
30:17know, worked in the past sort of uh
30:19platform era and now like what net new
30:21use cases or what is native to that
30:24you're seeing or anticipating or or
30:26looking forward to?
30:26>> Well, I'll give you my my version maybe
30:27a little bit of what Chris was saying,
30:28but like I I think so many I mean look
30:30there's a lot of great demos, a lot of
30:31great products releasing right now, but
30:32there's there's this thing I've just
30:33observed. But I saw the internet era
30:35earlier and I I call it product manager
30:37brain [snorts]
30:38>> um which is product manager is basically
30:39designing products for product managers
30:41right it's like it's by the way it's the
30:42one thing everybody does is they build
30:43products for themselves product managers
30:45have the story that they're building for
30:46other people [laughter]
30:47>> and yet somehow all the use cases are
30:49you know book me a restaurant you know
30:50to eat in San Francisco to go have
30:51dinner with my friends who are also
30:52product managers um or you know book me
30:55a plane flight to go to a product
30:56management conference where I also get
30:57to hang out with all my product manager
30:59friends and so there there's this you
31:01know it's a little and again there's a
31:03role for this but it's a bit like I
31:04think it's like the anthropic I'm a what
31:05is it? I'm a happy, helpful, harmful,
31:07harmless. I'm a happy, harmless,
31:09whatever it is. Honest assistant, right?
31:11And it's a little which also means I'm I
31:13am the most boring possible thing you
31:14could possibly imagine that's going to
31:16be constantly lecturing me. You've asked
31:17me a question I'm not allowed to answer,
31:19right? And so it's it's like there's
31:20just going to be this this this sort of
31:22we just had this very just the nature of
31:24the early companies who just had this
31:26very blinker view on on human activity
31:28what people are going to want to do. And
31:30it's to your point like with with Muse
31:31and Instinct, it's it's starting to open
31:32up. Um, but like look, this is going to
31:35be a much more entertaining, fun,
31:38enjoyable, exciting. Um, by the way,
31:40lucrative people are going to make
31:41money. There'll be all kinds of ways for
31:42people to make money.
31:44>> U, [snorts] be able to save money, u, be
31:45able to be parts of, you know,
31:47participate in parts of the economy, be
31:48able to be creators, you know, be able
31:50to, uh, you know, create new kinds of
31:51art. I mean, just in the last like three
31:53weeks, there's been an explosion of
31:54people making music videos like Denovo.
31:56>> Yeah.
31:57>> On like an incredible array of like
31:59creative ideas and topics
32:00>> like video games in the last three weeks
32:01have blown up. Uh,
32:02>> can we mention that just because that's
32:03so striking?
32:04>> Yeah, I mean I always think I always
32:05find video games interesting because
32:06they they're kind of the the lead I find
32:09they're often the leading edge of media
32:10because it just it's a technical culture
32:12that kind of iterates quickly. But like
32:14the I just saw one where it's like you
32:15can take any two video well first of all
32:17you can disassemble everything port them
32:19and just crazy stuff going on now where
32:20you can you can now run like Windows
32:22games on Unix faster than on Windows
32:25like because of these these
32:26disassemblers and these AIs are so good.
32:28Um, but also there was like this one
32:30where you can take any two video games
32:32and it will mash them up together and
32:33create a combo video game of it, which
32:36you know, just like a crazy new
32:38behavior.
32:38>> Super Mario and GTA, [laughter]
32:40>> right?
32:41>> Spider-Man and Jurassic Park.
32:43>> Um, and it's just and it's literally the
32:45crazy like to Mark's point, it's been I
32:46think like two weeks or something. It
32:47was like since what like Opus 55 or
32:50something. Um, and so it's just like the
32:52rate at which this stuff is happening
32:54and like what in a month there'll
32:55probably be some other, you know, two
32:57weeks there'll be some other big wave of
32:58this. Um, uh, you see Epic just
33:01yesterday said no more humans, right?
33:03And so like, yeah, there's been cultural
33:05resistance in the video game industry to
33:07AI, but like I think that's the first of
33:09the dominoes to fall because it's just
33:10inevitable because it's a much better
33:12way to build things and you're going to
33:13it's going to be great for consumers.
33:14You're going to have this
33:15>> I think it's going to be a creative
33:16renaissance. you're going to have this,
33:17you know, any independent developer can
33:19go and build like they have a huge team
33:21and um and and still have I think they
33:24can still have there's a lot of room for
33:25human creativity. It just it's a it's a
33:27is superpowers human creativity. Um
33:30>> so I think I think you were mentioning
33:32movies. Well, oh yeah, that might take
33:33longer. I don't know when that when that
33:35where that is on the you know the the or
33:38that that death door meme or whatever
33:40like I don't know which door it is but
33:42uh
33:42>> you can snap it quite quickly. I just
33:44I'm completely I was giving I'm
33:46completely obsessed with the movie this
33:47movie Digger that came out and I'm like
33:49literally the only person who saw it.
33:50You will be 100% tempted to walk out
33:52after the first half hour. Like you you
33:54will want to I I have a friend who
33:55literally took I have a friend who took
33:56her mother to see it
33:57>> and they walked out after the first half
33:58hour and now the running joke in the
33:59family is oh digger 2 is out. let's, you
34:01know, let's head to the theater, you
34:02know, just because it's such a comical
34:03idea, but so like a lot of people have
34:05had that, but if you actually watch the
34:07entire movie, there's a reason why they
34:08did what they did, and it's not what you
34:10think it is, and I don't want to spoil
34:11it, but like it's not it's not the way
34:13it is is presented at all. It's very
34:15clever. Um, anyway, but but like, you
34:16know, it's it's so it's this incredible
34:17creative achievement, but it's a $200
34:19million like cash loss for the for for
34:21for the studio, but it's that's because
34:23it had like $180 million production
34:25budget. Um, and so it's like with AI,
34:27what if that movie had had a $20 million
34:29production budget and what if it had
34:30made, you know, or a two or a $2 million
34:32production budget, right? Like that that
34:34level of art and creativity um at a
34:36lower price would have been great. As a
34:37business, by the way, as a business
34:38thing would have been f two million that
34:40that movie on a $2 million budget would
34:41have been insanely profitable. Would
34:42have been made the investors back 10
34:44times their money. Um, and so yeah,
34:45there's there's a huge thing that's
34:46going to open up in in the entertainment
34:48business. And then the other one, you
34:49know, Chris, you and I have have sort
34:50of, you know, been in and around
34:51hardware for a long time. like my whole
34:53career, you know, hardware, you know,
34:54for I'm a software person. Hardware has
34:56always been like terror incognito, which
34:58is like, you know, there are dragons
34:59there, right? Like it's just it's been
35:01so easy to blow up companies by going
35:02into hardware. And then again, your
35:04point like just in the last 3 weeks. So
35:05this this open source tool dropped
35:07called REA, reverse engineer anything.
35:10And it's this thing harnesses an AI
35:12model and you point it at anything,
35:13it'll reverse engineer it and give you
35:15source code. And one of the things that
35:16you can reverse engineer now is any
35:17piece of hardware,
35:18>> right? And so you have a thermostat
35:20that's like pissing you off in your
35:22house because it's not like working
35:23right. You can point this thing at the
35:24thermostat, reverse engineer it, and and
35:26the AI will write you new code and
35:27install it and put a new interface in
35:28the thermostat.
35:29>> And the same thing is true for ba
35:31basically anything like basically any
35:33form of hardware can now be taken over
35:34and reprogrammed with with this stuff.
35:35And so all of a sudden it's just like I
35:37mean I mean you take all these spaces I
35:39don't if anybody's doing this yet, but
35:40like just home automation is home
35:41automation's always been a complete
35:42nightmare of a market and and to
35:44implement as a customer. If anybody's
35:46ever done any sort of, you know, new
35:47house build, it's just a nightmare. But
35:49like home automation maybe should just
35:50be an AI bot that just like takes over
35:52all the hardware in your house and runs
35:53it however it is that you want to do
35:55that. And I I have friends that are
35:56literally doing this like right now in
35:57their in their own homes. Um, and so
36:00there is we we are potentially entering
36:01here like we're entering a renaissance.
36:03One way to think about it is like any
36:04anybody who understands software may
36:06sort of automatically be able to do a
36:08lot of the things that historically have
36:09been difficult in hardware um in a way
36:11that's just never been conceivable
36:12before. Um and so yeah, it just Yeah.
36:15And then you this sort of multiplicative
36:16thing you know these you know these are
36:18kind of the fields that are like closest
36:19into tech but then you just you know you
36:21you you expand out and you just you kind
36:23of think bas you know the entire economy
36:24is kind of uh open for for this kind of
36:26thing. Now
36:28>> Ben are there any uh any native uh AI
36:31native use cases that you're seeing or
36:32or are hoping to see or frontier uh sort
36:35of spaces that you we haven't discussed?
36:38>> Well that I think that was a really good
36:39discussion. I mean, I I think to me the
36:42the key is like these are going to be
36:44hard to anticipate because they're brand
36:46new things. You know, it it's funny like
36:49how the film industry has reacted where,
36:52you know, some people have really
36:54embraced it and you know, and then
36:55there's a bunch of people who have
36:57buried their head in the sand. But I
37:00think that, you know, what's going to
37:01happen with movies is there's going to
37:02be a new media. Um, you know, I think
37:05the AI film is not going to be a movie.
37:08it's going to be something very very
37:10different. Uh and and then look AI will
37:13also be used to make regular movies uh
37:17kind of more efficiently and and you
37:19know with better special effects and so
37:20forth. But the stories you can tell if
37:23you have this thing are just completely
37:25different than what you could ever tell
37:27before. Like there's you know it just
37:30hasn't been viable. I mean it's barely
37:32been viable for like Christopher Nolan
37:35but he's like the one guy who could do
37:37it. Um, but nobody else had a chance and
37:39and now everybody's going to be able to
37:41do it. And then there will be new ideas
37:43that that that come because you could do
37:45whatever you want.
37:46>> Yeah. I always think of like the camera.
37:48So the camera on the negative frame of
37:50the camera's invented. The negative
37:51version I guess is if you know if you're
37:53somebody who paints landscapes or
37:55something, it could replace you. The
37:56positive frame is is it unlocks motion
37:58pictures. You couldn't have had that
37:59without that, right? And so, you know, a
38:02lot of these technologies will both
38:03change an older form of media, but then
38:05also uh allow people to invent a new
38:08one. To Ben's point, like I I I would I
38:10don't know what it is, but I bet you
38:11there's a AI native form of media. Um
38:15people will make movies with AI, but
38:16they'll also do things that are probably
38:18hard for us to imagine right now. Um
38:20that'll be brand new.
38:22>> Yeah. [clears throat] The the brilliant
38:24Peter Teal once had this quote, I think
38:25a decade ago or something where he said,
38:27you know, crypto is libertarian. AI is
38:29uh is is communist which meant you know
38:31crypto is decentralizing AI is
38:33centralizing but I'm not sure that's
38:34entirely true given the conversation
38:36we're having about how AI uh gives so
38:38much leverage and so much power where
38:40you know one person companies or 30
38:42person companies can do you know orders
38:44of magnitude what they could do prior.
38:45>> Yeah. Yeah. I I I think there's all
38:48these interesting questions embedded in
38:49that like one is what is the end up
38:51final market structure of AI like I
38:53think where it's headed right now is
38:55you're going to have very successful
38:56like the big open AIS and anthropics but
38:58I you're also seeing a really big rise
39:00in uh open if you look at the open
39:02router data and things like open source
39:04is growing I think I believe much more
39:05quickly in terms of token consumption
39:08>> and then you know we haven't really seen
39:09yet and I think we may see with all the
39:11trends are there which is local models
39:12like people running these things at the
39:14edges running them on their computer
39:15maybe in the near term you can get you
39:18can run a top-end coding agent on your
39:20computer or chatbot or something like
39:21that. So I think that I think that
39:23there's a lot of opportunity for um a
39:25market structure that's sort of more
39:27broadly um decentralized like that. And
39:30then in terms of crypto I think crypto
39:31can play an important role like the
39:33natural way when you have an economy of
39:34Asians for them to pay each other would
39:36be stable coins.
39:37>> Yeah. And then on top of you know once
39:39you have sort of that base layer
39:40programmable money you would build you
39:42know decentralized lending and um and
39:45you know all the other things that
39:46people have built out in the crypto
39:47world. Um and so these are all open
39:50questions. Um I think it's what makes it
39:52kind of exciting and you know something
39:53we'd love to have an influence on is
39:55sort of you know I think we're you know
39:56we're big believers in open source
39:58software uh openweight models um uh
40:02crypto obviously blockchains and I I
40:04think I would love to see a version of
40:05this future that has all this great
40:08technology including these big
40:09centralized companies but also uh a lot
40:11of open source you know kind of level
40:13playing field I think it's better for
40:14consumers to have options like that it's
40:16better for startups to have options like
40:18that and so you know we'll do what we
40:20and to kind of hopefully nudge things
40:22that way.
40:23>> On that note, are Mark Ben are you guys
40:25encouraged with the you know where this
40:27sort of state of open source is
40:28currently from a technological
40:30perspective but also from a sort of you
40:32know um sort of just how it's perceived
40:34and how how it's uh the success of the
40:36industry in general.
40:37>> I mean right now it's great. I mean yeah
40:38so right now open source is racing
40:39ahead. you got this great dynamic where
40:41you know there's you know the two the
40:42two great tech superpowers of our age
40:43are the US and China both have you know
40:45very active um open source uh you know
40:48kind of worlds in addition to you know
40:49companies building building proprietary
40:50systems you know there there there's
40:52been this looming threat I mean you know
40:53we've talked about this in the past
40:54there's been this looming threat for the
40:55last three or four years of a level of
40:57what I would consider to be you know
40:58draconian and completely unacceptable
41:00you know either regulation or banning of
41:02open source that various folks in in
41:04governments have been threatening um I
41:06think the EU still aspires to do it um
41:08they just don't quite have the leverage.
41:10Uh
41:13threaten everybody with fines until we
41:14stop doing it, I guess. But um you know,
41:16the the current US administration is
41:18100% pro uh open source has been very
41:21encouraging. Um in fact, very much so
41:23100% the current administration 100%
41:25wants American open source to happen and
41:26to to win AI uh globally. The Chinese
41:29also um and by the way, a big surprise
41:31to me is I would have never thought the
41:32Chinese Communist Party would be like
41:33this enthusiastic about, you know, open
41:35source AI for the world, but at the
41:36moment they are. Um and so at the moment
41:39it's just it's just incredible. Like
41:40it's just absolutely ab absolutely
41:42amazing. And then you know the the open
41:43models are you know kind of lapping the
41:45um you know the uh the closed models by
41:47you know 6 months 12 months depending on
41:49how you count. Um, and then as as and
41:52then by the way the educational part of
41:53this I also always think when people
41:54talk about you know banning open source
41:56or regulating it they're like you know
41:57it's going to get used for bad things
41:58and you know there's there's sort of
41:59that whole discussion but the other is
42:00like do we want people to learn how this
42:02stuff works and do we want people to
42:03understand it and be able to deal with
42:05it and by the way right by by the way be
42:07be able to actually understand the
42:08problems as they arise and how to fix
42:10them and do we want the next generation
42:11of kids to actually understand how this
42:12stuff happens like what if to learn
42:15about AI like how how would you feel if
42:16the only way your kid could learn about
42:18AI is to go to work for one of two
42:19companies like that you know as a parent
42:21I think would probably drive you pretty
42:22crazy.
42:23>> Um whereas in a world of open source
42:24like every kid can learn how this stuff
42:25works and it's actually [clears throat]
42:27amazing. One of the really
42:28underestimated things about AI is AI
42:30will teach you about AI.
42:31>> Well with one exception which is that
42:33the AI models from certain of the large
42:35companies actually will refuse to do
42:36that which is sort of very interesting.
42:38Um but the open source models 100% will
42:40teach you how to make open source AI.
42:42>> Um and so every kid in the world I think
42:44should have as their birthright the
42:45ability to like fully understand and
42:47utilize this technology.
42:47>> By the way defend yourself. So like how
42:49how did hugging face defend themselves?
42:51Open source software. [laughter] I mean
42:53like open source means everyone has the
42:55most advanced vulnerability scanner to
42:57protect themselves from cyber. Um
42:59>> yep.
42:59>> Everyone has a doctor and a genius
43:02professor in their pock pocket. You know
43:04I mean
43:04>> y
43:05>> it's how we have why why the internet's
43:08ubiquitous. You can get internet access
43:09with all the world information for the
43:11price of a $10 cell phone because of
43:12open source. So, one of the things that
43:14started real quickly, Ben, one of the
43:15things that started is like wilderness
43:16explorers are starting to take local LMS
43:18with them on their phones completely
43:19loaded up to not require internet
43:21connection. Um, and the reason is cuz
43:23like if they get like in a you know, I
43:24don't know if like I don't know whatever
43:25if a bear I don't know what happens in
43:27the wilderness, but I don't
43:28>> Hi, bit by a rattlesnake.
43:30>> Bit [laughter] by a rattlesn bear clamps
43:32on your leg and won't like go or what
43:33whatever the hell it is that happens.
43:34You got to tell which bear is poisonous
43:36or whatever. It's like you want to be
43:37able to talk to an AI but you might not
43:38have connectivity. And so an open source
43:40model on the phone um all of a sudden is
43:42just like you have like a you know
43:44professor expert on I guess all aspects
43:46of wilderness survival um or what right
43:50whatever and right and that's a level of
43:51capability that you would not have if
43:52you were reliant on a handful of
43:54proprietary vendors where everything's
43:55running in the cloud and so yeah I I at
43:57the moment I feel really good about it I
43:58I am still worried that the you know the
44:00the various governments around the world
44:01are going to you know really kind of get
44:02hostile to this and it and by the way it
44:04goes back to your your your Peter Teal
44:05framing which is like without open
44:07source AI inevitably centralizes And
44:09inevitably centralizes down to, you
44:10know, one or two companies or maybe just
44:12flat out the government, right? And and
44:15I and I guess I I would ask people the
44:17following question, which is just close
44:18your eyes and picture the concept of
44:19government sushi. [laughter]
44:21Um, and then keep your eyes closed and
44:23picture government AI and you know, how
44:25do you feel about that? Um, and open
44:27source is the is the is the guarantor
44:29that that's that's not the future.
44:31>> Yeah. I I I mean to Mark's point I think
44:33the one thing that we know has always
44:36led to badness in the world is a massive
44:38concentration of power and you know one
44:40company one government one when one
44:43anything is is always been bad but you
44:46know kind of back to the point on open
44:47source like if you look at the problems
44:50in AI now the safety problems and so
44:52forth probably the biggest safety
44:54problem is this reward hacking
44:56phenomenon where the AI will do
44:57something that you didn't tell it to do
45:00but [clears throat] is kind of a means
45:02to an end to get its reward. And the
45:05labs so far haven't really worked on
45:08that very hard um for whatever reason.
45:10You know, they're they're busy kind of
45:11racing each other. And maybe there's no
45:15way to figure out what's going on with
45:17that and no way to kind of build a model
45:20that is more resilient to that kind of
45:22idea. But like what if every PhD
45:24student, every professor, every
45:26undergraduate in every university could
45:29look at the weights and the reward
45:32hacking and uh you know see if they
45:34could solve that and like maybe that
45:36would be a pretty darn darn exciting
45:37thing. And so you know these kinds of
45:40things are just not possible if only one
45:43company can work on the problem. Like
45:45you really want the whole world making
45:47this technology great and making this
45:49technology safe, not just the people who
45:52happen to work at two companies or maybe
45:55three companies. Go Meta.
45:58[clears throat]
45:58>> Exactly.
45:58>> Four companies. Go SpaceX.
46:00>> Yeah. Exactly. Every um well inspires a
46:03couple points. One is, you know, every
46:04week or every day, you know, there's a
46:06technological development or or
46:08breakthrough that we're blown away by.
46:09And then also it's coupled with the
46:11statement of like this is the worst the
46:12technology will ever be. like every day
46:13it just it just gets you know it gets
46:15better and better and we're it's
46:16compounding you and funny uh you know
46:18Tiel also had this quote around uh we
46:20wanted flying cars and all we got was
46:23140 characters um and it's funny because
46:25the guy who now runs a company who does
46:27the 140 characters or it also is
46:29building flying cars and we're investors
46:30in both
46:32>> and so uh it's funny how full circle uh
46:35it all all is um Chris I want to come to
46:38a different point uh that you had in in
46:40one of your wups which is uh the
46:42business model of the internet is
46:45breaking or is vulnerable. Why don't you
46:47talk about that? Yeah, I mean the
46:48internet was sort of there I I think of
46:49it as I call it the covenant the
46:51internet's covenant which is sort of a
46:52deal between a deal between the
46:55distribution side of the internet Google
46:57Facebook and so forth and the the
46:59content side you know news sites travel
47:01sites whatever it might be um you think
47:04about Google like the idea was you kind
47:06of give Google a snippet let them index
47:07you and then they'll send you some
47:08clicks right um that started to break
47:11like maybe last decade but particularly
47:14now you go to Google you get an answer
47:15right um you're get, you know, more and
47:18more you're sort of the social networks
47:19are trying to keep you on there, not
47:20linking out. Um and what that does is it
47:23has downstream effects, right? It means
47:24that the websites that make money on ads
47:27and conversions and other things um can
47:29no longer do that. I think that's
47:31probably going to you know, you see the
47:33early signs of this like, you know,
47:34Stack Overflow and some other websites
47:36where the traffic dropped precipitously
47:38when you had sort of a AI replacements.
47:40Um and so you know the the but but look
47:44people will still be booking tra like
47:45maybe they won't go maybe maybe your
47:46agent will go to the travel website or
47:48maybe um you know some other you know
47:51you'll get a recipe in some other way
47:52people will still get these things but
47:54you know many of these transactions will
47:56still happen um they just won't happen
47:58in the way they were happening and so
47:59that the the negative is it's painful in
48:01the near term and there's a bunch of
48:03businesses that are built on the sort of
48:04human attention economy that will have
48:06challenges. is I think on the
48:08entrepreneur side there's a chance to
48:09rethink these things from first
48:10principles and say okay what what is a
48:12what does a better internet look like
48:14you know internet where people are
48:15empowered with you know AI and
48:17everything else um and want to do things
48:19and and how do we kind of reorganize it
48:21so I think this is going to happen over
48:23the next let's call it decade is you're
48:25going to have sort of this kind of great
48:27reconfiguration of the internet and and
48:30how the economics work how the attention
48:32flows how knowledge is shared how people
48:35produce things together collabor
48:36collaborate
48:38um how organizations interact over the
48:41internet. Um so it I don't know it feels
48:43and in a way that I I feel like mobile
48:45didn't have that kind of impact. Mobile
48:47obviously, you know, I think, you know,
48:49made the internet go from whatever 500
48:53million to 5 billion people and
48:55increased engagement and a whole bunch
48:56of other things, you know, empowered
48:58people with all the information in their
49:00pocket had a lot of good effects. But
49:01this is something that's much more
49:02dramatic, I think, in the way it will it
49:04will reshape things and, you know, so I
49:07think there's a lot of also, you know,
49:08[clears throat] it'll be some somewhat
49:09painful for some, but I think it'll also
49:11be a lot of opportunity. On that note,
49:13just because we're talking about mobile,
49:14do you have any predictions or things
49:15you're excited about as it relates to
49:17computing or devices or or sort of form
49:19factor in which you know, ever since
49:21sort of the iPhone has been pretty
49:22pretty locked in. Are there new forms
49:24that you're you're particularly excited
49:25about or looking forward to?
49:26>> Well, I mean, everyone's talking about
49:28wearables and I think you'll see a b you
49:29pretty clearly going to see a bunch of
49:31Yeah, that just seems like it's unlocked
49:32in a way that it wasn't before with AI
49:34and voice. Um, and I think that's very
49:37interesting. Obviously, that's something
49:38that's sort of well covered. I've like
49:39I've always long believed in VR. I think
49:41it's just a question of get getting to
49:42the right format. You know, still heavy,
49:44you kind of sweat it. Um, a lot of it's
49:47just ergonomics. Um, but you know, it
49:49just seems kind of obvious to me as it
49:51has to many people and science fiction
49:53writers and others that if you have, you
49:55know, there are times in which people
49:56want deeply immersive u, you know, kind
49:59of otherworldly experiences. Um, and
50:02once that gets to the right point, I
50:03think, you know, AI might be the unlock
50:05because now the you can create this
50:06amazing content that actually is seeing
50:08on Twitter now. Like someone just did a
50:09port of Titanfall and so like by himself
50:12and it's like this incredible port. I
50:13don't know if it's going to you know,
50:15survive copyright lawsuits, but people
50:18will invent new stuff um that that you
50:20know that they invented. Um and and so
50:23you know I think it will just sort of
50:24unlock this wave because it was really
50:26kind of technically hard to build a you
50:27know VR game um or a world or something.
50:29Um, so I think I think you might see
50:31like kind of the resurgence of older
50:33devices like that. Um, there's all this
50:35kind of more crazy futuristic stuff
50:36happening with like, you know,
50:37Neurolink, brain computer interfaces.
50:40Um, obviously vehicles, you know, like
50:43we talked earlier about drones and and
50:45obviously cars, self-driving cars are
50:47are widespread. Yeah, I think you'll see
50:49I'm hoping you'll see kind of this
50:50resurgence of kind of homebrew hacking.
50:53Um, uh, you know, I I anecdotally I've
50:56seen a lot of that because you can now
50:58like just Claude will design a a 3D
51:01printed case for you and you get some
51:04cheap motherboards and then you write
51:06some custom firmware code and like
51:08you're off to the races. And so, um,
51:10maybe we'll see kind of a like there was
51:12kind of a movement like 12 years ago, 15
51:14years ago kind of around Makerbot and
51:16all these Maker things and but it was
51:18just the problem was just kind of hard.
51:20Yeah. So, you know, that would be cool.
51:22Uh I don't know what else.
51:23>> Yeah, I mean architecture I have a
51:25friend I shouldn't name uh but uh done a
51:27complete architecture design everything
51:29for a new new building. Um and you know
51:32you know people are assembling it but
51:33like the whole thing is is being steered
51:34by AI.
51:35>> Yeah.
51:36>> Um uh I think the new models are capable
51:39of doing is what they call design for
51:40manufacturing the the the final diagrams
51:44and specs to be able to factory which is
51:46very hard to do normally part of the
51:47manufacturing process. I think the
51:48models can just do that now.
51:50Um and so um yeah, no I mean like this
51:53these these are great examples of this
51:55like these domains that historically
51:56have been extremely difficult or
51:57prohibitive or impossible for kind of
51:59either normal people or for normal
52:00companies to do all of a sudden become
52:01you know quite possibly you know
52:02possibly just straightforward.
52:04>> Yeah. Is it worth saying more about your
52:05idea that uh you know AI right now is in
52:07really in single player mode and there's
52:09opportunity you know for machine age
52:10networks.
52:11>> Sure. I mean um I mean just a simple
52:13idea right is that just you you know in
52:15the internet era most um uh most
52:20technologies have had sort of multiple
52:22phases where you sort of first people
52:24will build some like you have the
52:26initial early iPhone apps or like these
52:28little flashlights and other kinds of
52:29sort of simple single single player apps
52:32something you use on your phone um and
52:34then over time people sort of said hey
52:35what we could do some things that are
52:36more social collaborative
52:39um you know you had inventions like Uber
52:41which is sort of a commerce network. You
52:42had um you know Instagram which is a
52:44social network and so forth. Um so one
52:48one might one observation one might make
52:50when you're looking at the AI landscape
52:52as amazing as it all is, it's still
52:54mostly you're sitting in front of a
52:55computer and you're typing and getting
52:56responses. Um there's some like little
52:59minor light sharing features, but at
53:01some point one might expect to have um
53:04you know sort of a network a multiplayer
53:06phase of AI where where you know you see
53:08you see it also like in a hugging face
53:10hack. there's a swarm that's obviously a
53:12bad a negative use case of a swarm. Are
53:14there positive use? You know, you also
53:16saw I think a swarm is one of the things
53:17that solve the one of the math theorems.
53:19>> You know, you can imagine a world where
53:21there's, you know, another word for
53:22swarm is a, you know, collaborative
53:24fleet of agents or something, right? And
53:26so like you can imagine lots of use
53:28cases where solving a hard problem,
53:30designing a building, you know, doing
53:32some kind of scientific breakthrough, uh
53:34you might might that might be the the
53:36right way to do it. You also see like I
53:38think Muse and Instinct have these
53:39features where you can like trust other
53:41people and you can imagine that sort of
53:42evolves into a trust network. So if you
53:44kind of squint you're seeing the early
53:46signs of this um and I think people are
53:49you know just kind of an experimental
53:50phase but it just based on on you know
53:54history and what might you know the kind
53:55of the possibilities of these products
53:57it seems like it might be a fertile area
53:58for entrepreneurs. But by the way, we've
54:01alluded, we talked a bit bit about sort
54:02of the uh education and just, you know,
54:05the learning uh opportunities that are
54:06enabled via this technology. I'm I'm
54:08curious, you know, obviously we just uh
54:10you know, uh incubated Haritz and
54:12Academy. Um thank you for for your
54:14support on it. If if you were 18 today,
54:17um how would you think about what you
54:18might uh study? Obviously, you have, you
54:20know, academic uh philosophy background.
54:22>> Um I think I'm stealing Mark's line, but
54:25I think you think there's one thing to
54:26do, which is study modern alchemy,
54:28right?
54:28>> Yeah. Yeah. Yeah. Yeah. Go for it. Go
54:30for it.
54:30>> Well, humans fig is Mark's idea, Mark's
54:32framing, but I mean, humans just figured
54:35out how to turn sand into intelligence.
54:37And the way to learn that, I think, is
54:39math, computer science, maybe economics,
54:42I don't know. And um and it's, you know,
54:45obvious look, I think it's also like
54:47like I I and I think Ben and Mark were
54:49lucky enough to have our, you know, a
54:52lot of our careers uh happen during the
54:54growth of the internet. And I think that
54:55was I feel very fortunate that I was
54:57able to do that. you know, I think that
54:59was something bequeathed to me by by
55:02people before me and that was I'm
55:04grateful to them. Um, and you know, I
55:06think now, look, it's obvious that, you
55:08know, um, that that the AI stuff we're
55:11talking about now is going to be the
55:12dominant narrative of the next 20, 30
55:14years. Um, so I think being and it
55:16doesn't have to, you don't have to
55:17necessarily be solely technical, but I
55:19think being in some way uh, in that in
55:22the mix and at the cutting edge and sort
55:25of using these tools feels like Yeah,
55:28>> kind of the the obvious thing to do. I
55:30don't know what you
55:31>> Yeah, I think I think a lot of other
55:33fields are going to be revolutionized,
55:34right? So I mean just you know the field
55:36field of the classics is being
55:37revolutionized because now you've got
55:38the vuvia scroll project and you're like
55:40able to use like AI to like decipher
55:42like dark scrolls and
55:43>> you know there may be entire new schools
55:44of philosophy that get you know
55:46discovered to study classics you know
55:47things you know lost lost books
55:49>> um economics uh I just having lunch with
55:51an economics professor and like the
55:52whole field of economics going to be is
55:54going to be transformed around this um
55:56and biology
55:57>> biology yeah exactly direct discovery
55:59design by the way design you know we
56:01talk about movies but also just broadly
56:02broadly design creativity of all kinds
56:03all all the all the all the forms of art
56:06music
56:06>> zoologology you're going to know what
56:08the animals are saying
56:10>> this is super interesting like we are
56:11plausibly heading towards a world in
56:13which we're going to understand what the
56:15animals are saying right um
56:16>> animal psychology will be actually real
56:19>> yeah so like okay like dog you know I
56:21don't know like dog you know dog
56:22training is going to take on a whole new
56:23thing when you actually understand what
56:25the what the the dog is saying
56:27right I think there's an X- prize for
56:28that [laughter] I'm serious it's like a
56:30it's like a
56:34>> [laughter]
56:34>> And they're they're definitely I mean
56:36anybody has a dog know they talk.
56:40They're in some weird language, you
56:42know, but but they they definitely are
56:44saying something and certainly cats do.
56:45Um
56:46>> I wait the entrepreneur listening to
56:47this who uh you comes in with a pitch.
56:50>> Totally. Totally. Um and then um uh um
56:53yeah, the the other way I think about it
56:54Yeah. in terms of like Yeah. thinking
56:55about like 18-year-olds or kids is um
56:57Douglas Adams had like the defining
56:58description uh of um the guy who wrote
57:01Hitchhiker's Guide to the Galaxy of like
57:03how technology is received on a
57:04generational level and his his famous
57:06framework was there's there any new
57:07technology is received three different
57:09ways by three different age cohorts. If
57:11you're um if you're u below the age of
57:1315, it's just the natural way the world
57:15works. Um, and I I've told in other
57:16settings the story of introducing my kid
57:18to Chat GPT at age seven. And I'm like,
57:20you know, this is such a big deal
57:21because the computer answers questions.
57:22And he's like, duh. Like, what good is a
57:25computer that doesn't answer questions,
57:26right? Um, uh, so it's just normal.
57:29>> Uh, if you're between the ages of 15 to
57:3135, it's new, exciting, and hot, you
57:32know, hot field, and you might be able
57:33to make a career out of it, in which
57:34case you want to learn all the all the
57:36new things. And if you're above the age
57:37of 35, it's unholy and evil, uh, and
57:39will surely, you know, overturn all
57:41existing structures and destroy society.
57:43>> I'll never answer the telephone.
57:46Exactly. Exactly. Old old men yell at
57:47cloud is a is a recurring recurring
57:49phenomenon.
57:50>> And so I I think 18-year-olds right now
57:51I mean it's a talk about prime time to
57:54to to gra to grab these new capabilities
57:55and run with them. Um um and you know
57:57the create like look it's going to be
57:59creation of entirely new fields
58:01>> uh of entrepreneurship and creativity
58:02and art and everything. Um and then look
58:04I mean I think kids younger than that
58:05I'm seeing with my my kid but like yeah
58:07they're just it's just of course you
58:08have this capability like of course you
58:09do. It's just like completely obvious.
58:11Um well and and actually the schools the
58:13schools are doing us a favor like so the
58:15the schools are not responding well to
58:16this um and so you know the schools are
58:18you know banning AI or having all these
58:19you know controversies with it but of
58:21course if you're a kid that just makes
58:22it cooler.
58:22>> Yeah.
58:23>> Right. And so a AI is going to be like
58:25the you know AI is your like secret
58:26superpower um that you have to like
58:28deviously figure out how to take
58:29advantage of. Um and so I think kids are
58:31going to have all kinds of ideas for
58:32what they're going to do.
58:33>> It's funny because adults keep saying oh
58:36what should the kids do? Like this so
58:37hard to be a kid. It's like, "No, no,
58:39no, no." Like, basically, it's you
58:44that's got the problem cuz you're
58:46obsolete. The kids are going to be all
58:48superhumans and they're going to be able
58:49to do anything and the world is open to
58:51them and they're going to, you know,
58:53they're they're going to be able to
58:54enter fields that they could never enter
58:56before.
58:57>> What I see in this conversation and in
58:58your fund and your work more generally
59:00is to use another Tism kind of definite
59:02optimism. The the future is happening.
59:04It it's going to happen. uh you know a
59:07lot of the what science we talked about
59:08science fiction is going to happen it's
59:10a matter of when and so we're going to
59:12bet early and as and maybe sometimes it
59:14be too early but we're going to keep
59:15betting and as long as you you know keep
59:16keep having that conviction uh you know
59:19it's it's going to work out.
59:20>> Yeah. Yeah. And to be clear, it's one
59:21thing I want to clarify is that, you
59:22know, we're throwing out all these
59:24theories and everything, but in [snorts]
59:25the end, I think as Ben said in the
59:27beginning, like we're we're in the
59:28betting on great entrepreneurs business.
59:30And so,
59:32>> um, you know, we any plenty of ideas
59:34that were not discussed today that are
59:36probably, you know, more interesting and
59:39more creative. um if if for whatever
59:42reason they don't fit in the in the
59:44other kind of core areas at our firm,
59:46like we're open-minded and and willing
59:49to, you know, to engage with those kinds
59:52of creative ideas. So, um so I would say
59:56you said definite optimism. I'm like
59:58optimism, but also I would say humble
1:00:01definite optimism or something.
1:00:03Intellectually humble. Yeah. and that
1:00:04I've been wrong so many times that uh we
1:00:07just we really want to we want to find
1:00:08great people who have their own definite
1:00:10optimist and back them.
1:00:11>> The entrepreneurs get to be optimistic.
1:00:14>> Yeah. Yep.
1:00:15>> That's a great place to wrap. Chris,
1:00:16thank you guys for coming on the
1:00:17podcast.
1:00:18>> Good. Thank you.
1:00:18>> All right. Okay. Thank you.

About this transcript

This page contains the full transcript of The Frontier Ideas a16z Wants to Fund Now | Marc Andreessen, Ben Horowitz & Chris Dixon by a16z, generated from the public captions YouTube serves with the video. The transcript has 14,230 words across 1,988 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.