Latest / Newcomer Pod / This Kicked Off With a Dinner With Elon Musk Years Ago (with Reid Hoffman)
Transcript
- 0:01Hi I'm Eric newcomer author reporter of newcomer the podcast
- 0:07the you're listening to is called newcomer.
- 0:09It's all about startups and Venture Capital inside the
- 0:13business of Technology launched a YouTube channel, relaunching
- 0:18the podcast, you can find it. Apple Spotify newcomer dotco
- 0:23wherever you find your podcast, you can listen ball on the sub,
- 0:27stack very excited for my first. Yes, I think it delivers on The
- 0:32Insider proposition, PayPal Mafia member, Microsoft board,
- 0:35member LinkedIn, co-founder Greylock investor.
- 0:38You haven't guessed? That's Reid.
- 0:40Hoffman. What do we talk about?
- 0:42You can guess it. Artificial intelligence, you
- 0:44just stepped off the board of open and I asked him about that.
- 0:47We talked about AI sentience whether when and we'll have
- 0:51feelings, what we should do about it.
- 0:52If it has feelings, some deep cut philosophy references, but
- 0:56we also talked about Amazon, Microsoft, Google the business.
- 1:00Of cloud computing and how that's affected by a.
- 1:02I asked him about his Paypal mafia members.
- 1:05Peter teal, Elon Musk, what they were thinking.
- 1:08We talked about Joe Biden read is a democratic Mega donor.
- 1:12So I asked him about the state of his donating to the
- 1:14Democratic party. I think it's a great episode, I
- 1:17really enjoyed it, I hope he comes back on before we go to
- 1:20the episode, we have a word from vanta.
- 1:23Thank you for being our launch sponsor.
- 1:25For the newcomer podcast newcomer is brought to you by
- 1:28vanta Too Close. And grow major customers, you
- 1:31have to earn trust, but demonstrating your security and
- 1:34compliance can be time. Consuming tedious and expensive
- 1:39until you use vanta, vanta automates up to 90% of the work
- 1:43for the most sought-after, security, and privacy standards,
- 1:46save time and money on compliance with Vance's,
- 1:49enterprise-ready. Trust management platform for a
- 1:52limited time. Newcomer listeners.
- 1:54Get $1000 off vanta, go to VA en.
- 1:58Ta.com Comer to get started. Here's the episode read.
- 2:04Thank you so much for coming on the show.
- 2:05I'm very excited for this conversation, you know, I'm
- 2:08getting ready for my own AI conference at the end of this
- 2:12month. So hopefully, this is sort of a
- 2:14prep by the end of this conversation will all be super
- 2:17smart and ready for that. You know, a, I will be at the
- 2:21heart of it AGI generative. A, I like, I feel like there's
- 2:24been so many rounds of sort of AI as sort of the buzzword of
- 2:29the day. This nearly seems the hottest.
- 2:32When did you get into this moment of it?
- 2:34I mean, obviously, you just stepped off, the opening I
- 2:36bored. You are on that board.
- 2:38Like, when did you start to see this movement that we're
- 2:41experiencing right now? Coming?
- 2:43Well, it's funny because my undergraduate major was
- 2:46artificial intelligence and cognitive science.
- 2:48So I've been around the hoop for multiple waves for a long time.
- 2:54And I think this kicked off actually, in fact, with a dinner
- 2:58with Elon Musk years ago, We're because you know about tennis
- 3:03years ago Ilan and I would have dinner about once a quarter and
- 3:06he's like all you paying attention to this AI stuff and
- 3:09I'm like well I majored in it and I know about the size of
- 3:12know you need to get back involved.
- 3:14All right. All right, you know, sure, this
- 3:15is part of how I operate is people.
- 3:17I did a smart people from my network.
- 3:20Tell me things, and I go and do things and so I dug into it.
- 3:22I'm not. Oh yes, we have another way of
- 3:26coming and this was probably about seven or eight years ago
- 3:29when I say I saw the beginning of the the way, where the
- 3:32seismic event starting it was a seismic event out at Sea.
- 3:37And I was like, okay, there's going to be a tsunami here and
- 3:40we should start getting ready because the tsunami is actually
- 3:42going to be amazingly. Great, and interesting is that
- 3:45just one to one with you and your honor that yes sir.
- 3:47Yeah, yeah. And that is that the beginning
- 3:50of open a.i., Opening eyes later, what I did is I kind of
- 3:56that and went and made connections with the kind of the
- 3:59heads of every AI lab and every major company.
- 4:02Because I concluded that, I thought that the AI Revolution
- 4:05will be primarily driven by large companies initially
- 4:09because of the scale compute requirements.
- 4:12And so you've talked to Dennis hassabis met, Mustafa Suleiman
- 4:16talk to you on the Coon talk to Jeff Dean, you know, all these
- 4:19kind of folks and kind of, you know, built all that And then it
- 4:22was later in conversations with Sam and Ilan.
- 4:28That is it? Look, we need to do something.
- 4:30That's a pro Humanity. Could have not just commercial
- 4:33effort and my price for participation because I thought
- 4:35was a great idea, but my price participation was to ask elom to
- 4:39stop saying the word robot Clips because I thought that the
- 4:42problem was is it's very catchy at it.
- 4:46Evokes fear and actually in fact one of the things I think what
- 4:49this whole area is that it's so much more interesting.
- 4:52Young and have so much amazing opportunity for Humanity a
- 4:55little bit. Like I don't know.
- 4:56If you saw the Atlantic article, I wrote that we evolve ourselves
- 4:59through technology and I'm you know, going to be doing some
- 5:03writings around describing a, i as augmented intelligence versus
- 5:07artificial intelligence. And I wanted to cut up build
- 5:10that positive optimistic case that I think is the higher
- 5:13probability that I think we can shape towards.
- 5:15And so it was like okay I'm in but no more robot Clips I
- 5:19appreciate the ultimate sort of network.
- 5:22In the you tell the story through people, I always
- 5:24appreciate when the origin stories of Technology actually
- 5:27come through. The human beings with Ilan.
- 5:30In particular, I'm sort of confused by his position because
- 5:33it seems like he's very afraid of AI.
- 5:36And if that's the case, why would you want to like do
- 5:38anything to get the ball rolling down the hill?
- 5:40Like isn't there a sort of just like stay away from a man if you
- 5:43think it's so bad or like how do you see his thinking?
- 5:46I'm sure it's evolved over time. Well, I think his Instinct for
- 5:52the good and the challenging of this is he tend to think a, I
- 5:57will only be good if I'm the one who's in cattle, a dangerous
- 6:02sort of? Yeah.
- 6:03Yeah. This is actually somewhat
- 6:04replete with in the modern day. I filled not everybody on Ilan
- 6:08is in the public enough figure that I think, you know, making
- 6:10this comment of him is not talking at a school other people
- 6:13would. There's a surprising number of
- 6:15Messiah complex has in the field of AI.
- 6:18And it's the The I am the one who must bring this, you know,
- 6:20Prometheus, you know, the fire to humanity and you're like,
- 6:25okay, I kind of think it should be us versus an individual.
- 6:30Now us can't be 8 billion people us as a small group but I think
- 6:34more or less you see the folks who are steering with a moral
- 6:39compass. Try to say how do I get at least
- 6:4210 to 15 people Beyond myself with their hands on the steering
- 6:46wheel in deep conversation? Missions in order to make sure
- 6:49you get there and then let's make sure that we're having the
- 6:51conversations with the right communities.
- 6:54Like if you say well is this going to, you know,
- 6:56institutionalize ongoing, you know, power structures or racial
- 7:00bias, something else? Well, we're talking to the
- 7:02people to make sure that we're going to minimize that
- 7:05especially over time and navigate it is a real issue.
- 7:08And so that's the kind of anti Messiah complex which is more or
- 7:12less the efforts that I tend to get involved in right?
- 7:15At least sort of oligarchy. Yes.
- 7:18Okay. I control instead of just
- 7:20dictatorship. Well yeah, it depends a little
- 7:23bit, even an oligarchy look. Things are built by small
- 7:26numbers of people. It's just a fact, right?
- 7:28Like there aren't more than you know, a couple of Founders maybe
- 7:32maximum five in any particular thing, there's reasons why when
- 7:36you have a construction project you have a head of construction,
- 7:40right? Writing cetera.
- 7:42The important thing is to make sure that like that's why you
- 7:44have, why you have a CO, you have a board of directors,
- 7:47that's why you have, you know, you Will do we have the right
- 7:50thing where person is accountable to a broader group
- 7:54and that broader group feels their governance
- 7:58responsibilities. Seriously.
- 7:59So, oligarchy is a, is a charge, which is a charged word and I
- 8:04there's a logic to it. I'm glad I'm not using to say,
- 8:07it doesn't make sense that you want to be able to really
- 8:09understand it or around it. I mean specifically with open
- 8:12a.i. I mean you just stepped off the
- 8:14board you're also on the board of Microsoft which is obviously
- 8:18a very Nificant player in this future.
- 8:21I mean, it's open. I get a little frustrated with
- 8:25the open and opening I because I feel like there's a lot that I
- 8:27don't understand. I'm like maybe they should
- 8:29change the name a little bit. But is it still a charity in
- 8:33your mind? I mean it's obviously it's race
- 8:35from Tiger Global, the ultimate profit maker.
- 8:38Like how should we think about the sort of core Ambitions of
- 8:42open a I won the board. I was on as a 501 c 3 and
- 8:46they've been very diligent About making sure that all of the
- 8:50controls including for the subsidiary company are from the
- 8:54501 c 3 and diligent to its Mission which is staffed by
- 8:58people on the 501. C 3 board with the
- 9:01responsibilities of being on a 501 c, 3 board which is being of
- 9:04service. The mission not doing, you know,
- 9:06private environment and in other kinds of things.
- 9:08And so I actually think it is fundamentally still a 501 c 3.
- 9:12The Challenge is if you kind of say you look at this and say
- 9:14well in order to be a successful player in the In scale AI you
- 9:19need to have billions of dollars to compute.
- 9:21Where do you get those billions of dollars?
- 9:23Because, you know, the foundations and the philanthropy
- 9:26industry is generally, speaking bad at Tech and bad at anything
- 9:30other than little tiny checks in Tech.
- 9:32And she said, well right, it's really important to do this.
- 9:34So, part of what I think, you know, Sam, and that group of
- 9:38folks came up with this. Kind of, clever thing, was to
- 9:41say, well, look, we're about beneficial AI.
- 9:44We're about AI for Humanity, we're about making it, I'll make
- 9:47a comment on In a second, but we going to generate some
- 9:51commercially valuable things. What are we struck a commercial
- 9:54deal? So you can have the commercial
- 9:55things or you can share the commercial.
- 9:57Things, you invest in us, I want to do this and then we make sure
- 10:00that the AI has the right characteristics and then the
- 10:03open, you know, all short names have you know some you know,
- 10:08simplicity's to them. Yeah, the idea is Open to the
- 10:14world in terms of being able to use it and benefit from.
- 10:17It doesn't mean the same thing as open source because AI is
- 10:21actually one of those things where opening where you could
- 10:25open source, you could actually be creating something dangerous.
- 10:28As a modern example, last year opening.
- 10:31I deliberately like Dolly to was ready for months before it went
- 10:36out. I know because I was playing
- 10:38with it. They did the four months to do
- 10:40safety training and the kind of safety training.
- 10:42In is, well, let's make sure that individuals can't belie
- 10:46bulb. Let's make sure you can't create
- 10:48as best we can child sexual material.
- 10:51Let's make sure you can't do revenge porn, right?
- 10:53And we'll serve it through the API and we'll make it unchanged
- 10:56while Matt. And then the open source people
- 10:57come out and they go ability to do whatever you want.
- 11:01And then, wow, you get all this crazy, you know, terrible stuff.
- 11:05And that's part of like, so open is openness of availability, but
- 11:10still with safety and still with kind of called Pro-human
- 11:13controls, and that's part of what open.
- 11:15A.i. means. In this context, I wrote in sort
- 11:17of a mini essay in the newsletter about tech fatalism
- 11:21and it fits into your sort of Messiah complex that you're
- 11:24talking about where if I'm a young or new startup
- 11:27entrepreneur. It's like this is my moment.
- 11:30If I hold back you know there's a sense that somebody else is
- 11:33going to do it to this isn't necessarily research.
- 11:35Some of the tools are findable so I need to do it.
- 11:38If somebody's going to it's easy, if you're sitting your own
- 11:41personhood. To say, I'm better than that
- 11:44guy. Even if I have questions about
- 11:45it, I should do it so that I think we see that over and over
- 11:49again, obviously, the stakes with a.
- 11:50I think we both agree are much larger on the other hand.
- 11:54With a, i there's actually, in my view been a little bit more
- 11:56restraint. I mean, Google has been a little
- 11:59slower, Facebook, seems a little worried.
- 12:01Like, I don't know, how do you agree with that sort of view of
- 12:04tech fatalism? What do we is there anything to
- 12:06be done about it? Or it's just sort of if it's
- 12:08possible? It's going to happen.
- 12:10So the best team should do it or how do you Think about that
- 12:13sense of inevitability on if it's possible, it will be built
- 12:16well, one thing is you like edgy words so what you described is
- 12:20Tech fatalism, I might say is something more like Tech
- 12:24inevitability or Tech Destiny and part of it is what I guess.
- 12:28What I would say is like for example we are now in a AI
- 12:33moment in Era, there's Global competition for it at scale
- 12:36compute. It's not something that even
- 12:38somebody like a Google or someone else can kind of have
- 12:42any Dove real ball control on, but the way I look at it as hey,
- 12:46look, there's there's utopic outcomes and dystopic outcomes
- 12:49and it's within our control to steer it and even to steer it at
- 12:53speed even under competition. Because for example, obviously
- 12:57the general discourse within media is oh my God, what's
- 13:01happening with the data and what's going to happen bias?
- 13:03And what's gonna happen with crazy conversations?
- 13:05With Bing Chad and all the rest is stuff.
- 13:09And you're like, well, what am I obsessed about?
- 13:11I'm obsessed about the fact that I have line of sight to an AI
- 13:15tutor and an AI doctor on every cell phone, right?
- 13:18And think about if you delay that whatever number of years,
- 13:21you delay that what your human cost is overlaying that, right?
- 13:25And it's like how do we get that?
- 13:27And like, for example, people say wow, the real issue is that,
- 13:30you know, a not completely trained, you know, being chat
- 13:34model is going to go off the rails and have a drunken
- 13:37cocktail party conversation because it's provoked to do so
- 13:40and can't run away from the person who's provoking it right.
- 13:43And you say, well, is that the real issue or is the real issue?
- 13:46Let's make sure that as many people as we can have access to
- 13:49that a eye doctor, right? Have access to that a I tutor
- 13:53not only, you know, because obviously technology because
- 13:55it's expensive initially benefits that leads and people
- 13:58are rich. And whether that's the natural
- 13:59way of how our capitalist system and all the rest Works.
- 14:02We're trying to get everyone else as quickly as possible,
- 14:04Right? 100% agree with that.
- 14:06Yeah, I don't want any of my sort of cynical tone.
- 14:08Oh, I agree that it's like, oh my God, this is urgent.
- 14:11I don't extend it, you know. You're sort of referencing,
- 14:14maybe this Sydney situation where you have Kevin ruse in New
- 14:17York Times. You know, communicating with
- 14:19Bings version of chat gbt and sort of finding this character
- 14:23who's sort of goes by Sydney from the origin story and I Ben
- 14:26Thompson sort of had a similar experience and I would almost
- 14:29say it's sad for the world to be deprived of that to, you know,
- 14:32there's like certain paranoia, it's like, oh, I want to meet
- 14:34this sort of seemingly intelligent character.
- 14:37I don't know. What do you make of that whole
- 14:40episode? I mean, smart Tech writers.
- 14:43Really latched onto. This is something that they
- 14:45found moving. I don't know.
- 14:47Is there anything you take away from that Saga?
- 14:50And you think we'll see those sort of, I don't know.
- 14:52Intelligent characters again. Well, for sure because I think
- 14:572023 will be at least the first year of the so-called chatbot.
- 15:01Not just because of chat gbt and I think that we will have a
- 15:06bunch of different chat Bots. I think we'll have chat Bots
- 15:09that are there to be, you know, entertainment companions Witty
- 15:13dialogue, which is been so I think we'll have a chat Bots
- 15:16that are there to be information like insta.
- 15:18Wikipedia kind of things, I think will have chat Bots that
- 15:21are there to just have someone to talk to, right?
- 15:24So I think they'll be a whole whole range of things.
- 15:27So I think we will have all that experience.
- 15:29And I think part of the thing is to say, look, what are the
- 15:33parameters by which you should say the botched should
- 15:37absolutely not do X, and it's fine.
- 15:40If these people want about that. It's like, you know, smack
- 15:44talking and these people want something that has goes oh heck
- 15:49right, you know right and fine you know like what's the range
- 15:53of that and obviously children get in the mix and and the
- 15:56questions around things that we already encounter a lot with
- 15:59search which is like oh we worry about you know, areas of.
- 16:02Well could a chatbot enable self-harm in a way that would be
- 16:05really bad. Really try to make sure that
- 16:07someone who's depressed Doesn't figure out a way to harm
- 16:11themselves either with search or chatbots.
- 16:14Is there a psychologically persuasive?
- 16:16So it's not just the information provided.
- 16:18It's just sense that they might be like, walking you towards
- 16:21something. Yeah, and they are in, on this
- 16:23is my serious. This is the thing that's
- 16:24amazing. And it's part of the reason why
- 16:26I like everyone should have some interaction of these, is that in
- 16:29some emotional tangible way, we are really passing the Turing
- 16:33test. This is the thing that I had
- 16:34visibility on, you know, a few years ago because I was like,
- 16:38okay, actually, in fact Is we kind of Judge, you know,
- 16:42intelligence and sentience like the Google Engineers are good.
- 16:44I asked if it was conscious and it said it was because we use
- 16:47language as a way of doing that, right?
- 16:49You're like well, but look that tells you that your language use
- 16:52is not quite quite fully there. And because, you know, part of
- 16:58what's really amazing about hallucinations.
- 17:00And I'm probably gonna do a fireside chat box with a gray
- 17:03matter thing on hallucinations maybe later this week where, you
- 17:07know, the Hallucination is on one hand.
- 17:09It says this. Amazingly accurate.
- 17:11Wonderful thing, very persuasive.
- 17:13Right and then it says this other thing really persuasively.
- 17:16That's right, little fiction, right?
- 17:18And you're like wow you sound very persuasive in both cases
- 17:21but that one's true and that one's fiction.
- 17:23And that's part of the reason why I like I kind of go back to
- 17:25the augmented intelligence and all the things that I see going
- 17:29on with his 2023 year. Right is much less replacement
- 17:33and much more augmentation. It's not zero replacement but
- 17:36it's much more augmentation in terms of how this play is.
- 17:39And that It is super exciting. Yeah, I mean, some degree, it
- 17:43reflects sort of the weakness in human beings, own abilities to
- 17:48read what's happening. You know, ahead of this
- 17:50interview I was talking to the sort of the publicly available
- 17:53chat CPT. I don't know if you saw, I was,
- 17:56you know, asking it for questions, and I felt like it
- 17:58delivered a very reasonable set of questions, you know, you've
- 18:02written about bullets scaling. So it's like, let's ask about
- 18:05that. It's, you know, asking the
- 18:06context of Microsoft. But when I was like, you know,
- 18:09have you Have you ever watched Joe Rogan?
- 18:11Have you ever been on a podcast sometimes?
- 18:13Maybe you should have a long sort of?
- 18:15You should have a statement like I'm doing right now where I sort
- 18:17of have some things I'm saying. Then I asked a question other
- 18:20times. It should be short and sweet.
- 18:21Sometimes it you know, annoys you and says oligarchy, you
- 18:24know, like explaining to the to the chatbot like oh you can't
- 18:28just ask like a list of like straightforward questions and
- 18:31like it felt like it didn't really even get that and I get
- 18:34that we're starting to have a conversation.
- 18:35Now with companies like Jasper where it's almost like the
- 18:38language prompting and I think Sam Altman was maybe say it's
- 18:41like almost a form of plain language like coding because you
- 18:44have to figure out how to get what you want out of them.
- 18:47And maybe it was just my failure to explain it, but I don't know.
- 18:50As a journalist replacing questions.
- 18:52I didn't find the current model of chat GPT sort of really
- 18:57capable of. No, that's actually one of the
- 18:59things on chat. Gbt I find is it like, for
- 19:03example, you ask what? Questions?
- 19:04Ask Reid, Hoffman, and a podcast interview and you'll get some
- 19:07generic ones. It'll say like, well, right?
- 19:09What's going on with? New technologies, like Ai and
- 19:11right and what's going on in Silicon Valley and you know
- 19:14Andre like okay sure but those aren't the really interesting
- 19:17questions that's not what makes me a great journalist which is
- 19:20right about a lens to something that people can learn from and
- 19:25and that will evolve and change. I will get better but that's
- 19:28again one of the reasons why I think it's a people plus machine
- 19:31because like for example like if I were to say hey what should I
- 19:34ask Eric about or what should I talk to Eric about and going to
- 19:37give me some generic stuff. Now, if I said, oh, Give me a
- 19:41briefing on un governance systems as they apply to a, I
- 19:50write because I want to be able to talk about this.
- 19:52I didn't do this, but then then it would give me a kind of a
- 19:54quick Wikipedia briefing and that would make my conversation
- 19:57more interesting, right. And I might be able to ask a
- 19:58question about the governance system or something, you know,
- 20:01as a way of doing it. And so, and that's what I is, I
- 20:04think why the combo is so great and that's what we should be
- 20:07aiming towards it. Isn't to say, by the way,
- 20:10Sometimes like replacement is a good thing like for example you
- 20:12go to autonomous vehicles and say, hey, look if we could wave
- 20:15a wand and every car on the road today would be an autonomous
- 20:17vehicle, right? Probably save.
- 20:19We probably go from 40,000 deaths in the u.s. /.
- 20:23You know, Year to, you know, maybe 1,000 or 2,000.
- 20:27Yeah. You're like you're shaving 38
- 20:29ice thousand lives a year in doing this.
- 20:31It's a good thing. And you know it will have
- 20:34positive effect on on gridlocks and their climate change and all
- 20:36the rest of stuff and you go okay that replacement as
- 20:40Navigate truck jobs and all the rest of that Replacements.
- 20:43Good, right. I think a lot of it is going to
- 20:46end up being, you know, kind of various forms of amplification.
- 20:51Like, if you get to journalists you go, oh, it'll help me ask
- 20:56figure out which interesting questions to add, not because
- 20:59it'll just go here. Here's your script line.
- 21:01Ask question, but it will you can get like better information
- 21:06to prep your thinking on it. I'm glad you brought up.
- 21:10Like the self-driving car case and your are you still on the
- 21:12board of Aurora? I am your.
- 21:14Yeah, you know I covered Uber. So I've been in there.
- 21:17I was in their self driving cars, very early and they made a
- 21:20lot of promises lift made a lot of promises.
- 21:22I mean, I feel like part of my excitement about this sort of
- 21:26generative, a eye movement, is it, it feels like it doesn't
- 21:30require completeness in the same way that self-driving cars.
- 21:35Do you know? And that that's been a barrier
- 21:37to self driving cars on the Upside, you know, sometimes we
- 21:41sort of wave away oh the inaccuracy and then we sort of
- 21:44imagine and I think that's what we were sort of talking about
- 21:46earlier. You imagine in some of the
- 21:47completeness, that could come. So I guess the question here is
- 21:51just, do you think this sort of like what I'm calling the
- 21:54completeness problem? I guess just the idea that it
- 21:56needs to be sort of fully capable will be an issue with a
- 22:00large language models, or do you think think it is?
- 22:03I mean, you have this sort of augmented model where it like it
- 22:06could sort of stop now and still be extremely useful.
- 22:09Too much of society, one, I think it would stop now and be
- 22:12extremely useful, you know, I got line of sight on current
- 22:15technology for a tutor for a doctor, for a bunch of other
- 22:18stuff. One of the things my partner,
- 22:20Psalm motamedi, and I wrote last year, was it within five years.
- 22:24There's going to be a co-pilot for every profession.
- 22:26The way to think about that is what a professional do.
- 22:29They process information they take some kind of action
- 22:31sometimes that's generating other information just like you
- 22:34see with you know Microsoft's copilot product for engineers
- 22:37and what you can see happening with Dolly and other.
- 22:40His generation for graphic designers, you will see this for
- 22:42every professional. That there will be a co-pilot on
- 22:46today's technology that can be built.
- 22:48That's really amazing. I do think that as you continue
- 22:51to make progress, you can potentially make them even more
- 22:54amazing because, you know, part of what happened when you move
- 22:58from g b, d 3 to 3.5, which is a chibi, he is all of a sudden, it
- 23:03can write sonnets right, right? And it really know it was going
- 23:06to be able to write sonnets. I didn't write and how that
- 23:09comes out. And by the way, that's giving
- 23:11people superpowers, most people glued myself.
- 23:13I mean, look I could write a sonnet if you gave me a couple
- 23:15of days, correct. And a lot of coffee, and a lot
- 23:18of good night to you to really try, but you wouldn't, but, you
- 23:22wouldn't. Yeah.
- 23:23But now I can go, oh, you know, I'd like to to write a sonnet
- 23:27about my friend, Sam Altman, and I can go down and I could sit
- 23:31there and I can kind of type, you know, did it out?
- 23:33And I can generate why, I don't like that one mole.
- 23:34But that, I like this one, you know, and so did I and, and that
- 23:37that gives you super powers. I mean, think about what, Then
- 23:40do for, you know, kind of writing a whole variety of
- 23:43things with that. And I think the more and more
- 23:45completeness is the word you were using is I think, also a
- 23:49powerful thing, even though what we have right now is amazing is
- 23:53GPT for a big Improvement are like, what have you?
- 23:56I assume you've seen a fair bit of unreleased stuff.
- 23:58Like, how hyped should we be about the Improvement level?
- 24:02I have, I'm not really allowed to say very much about it
- 24:05because, you know, part of the responsibilities of former board
- 24:08members and confident Reality side, but I do think it will be
- 24:12a nice. I think people will look at and
- 24:13go. Wow, that's cool.
- 24:15And and it will be another. Iteration another thing where is
- 24:19as amazing as chat CBT has and obviously that's kind of the
- 24:22last few months is too kind of taken the World by storm with
- 24:24like, opening up this this this Vista of imagination and so I
- 24:28think gbd4 will be another step forward where people will go.
- 24:32Oh that's that that's another cool thing I think.
- 24:35Yeah. Can't be more specific than
- 24:36that. But I think it is a, you know,
- 24:38watch this space because it'll be cool.
- 24:40Throughout this conversation, we've danced around the sort of
- 24:42artificial general intelligence question.
- 24:44I mean, starting with the discussion of EU on in the
- 24:46creation of eventually open I'm curious how close you think we
- 24:51are with a GI? And you know, this idea of a
- 24:54sort of I mean people Define it, so many different ways.
- 24:57It's more sophisticated than humans and some tasks, you know,
- 25:00many tasks whatever. Do you think we're far from
- 25:03that? Or how do you see that play?
- 25:04Now personally amongst a lot of the people who are in the field.
- 25:09I'm probably on On the were much further than we think staged
- 25:12some of that's because I've lived through this before with
- 25:16my undergraduate degree and the, you know, the pattern generally
- 25:20is, oh my God, we've gotten this computer to do this amazing
- 25:23thing that we thought was formerly the Providence of only
- 25:27these cognitive, you know, kind of human beings and it could be
- 25:32that. So then by the way, in 10 years
- 25:35it will be solving new science problems like fusion and all the
- 25:38rest and you know, if you go back to the 70s, all the same
- 25:41dialogue. I mean, it's an ongoing thing.
- 25:43Now we do have a more amazing set of cognitive capabilities
- 25:46and we did before and there are some reasons to argue that it
- 25:49could be in a decade or two because you say well but
- 25:54actually in fact, you know, these large language models can
- 25:57enable coating. And that coating can, I can then
- 26:00be self-reflective and generative and that can then
- 26:03make something go. But like for example, when I
- 26:05look at the coding and how that works right now isn't Write the
- 26:10kind of code that's like, oh, that's amazing.
- 26:11Encode, it helps with the oh, I want to do a parser for
- 26:14quicksort, right? You know, like that kind of
- 26:17stuff, it's like, okay, that's great.
- 26:19Oh, you know, our systems integration use of an API or
- 26:21calling in a, an API for a spell checker or whatever.
- 26:24You know, the couch really helpful stuff and Engineers.
- 26:26But it's not like, oh my God, it's now we're inventing, the
- 26:30new, you know, kind of training of Large Scale Models
- 26:33techniques. And so, I think even some of the
- 26:36great Optimist will tell you oh the great like Believers that
- 26:39Soon as I there's one major invention and the thing is, once
- 26:43you get to one major invention, is that one major invention.
- 26:45Is it three major inventions is a 10 major inventions.
- 26:48Like I think we are we are some number of major inventions away
- 26:52and I certainly don't think it's impossible to get there.
- 26:54Sorry the major inventions are us human beings build buildings
- 26:58into the systems or yeah. Like for example, you know can
- 27:01it do like for example, a classic critique of a lot of
- 27:05large language models as Canada to Common Sense reasoning right
- 27:08hand, you know, very much Curtis is very loud.
- 27:10Exactly. Jack right.
- 27:11Exactly. And and you know, the short
- 27:15answer right now is a large language models are
- 27:17approximating common sense reasoning.
- 27:19Now they're doing it in a powerful and interesting enough
- 27:21way that you're like, well, you know, that's pretty useful.
- 27:26It's pretty helpful about what it's right.
- 27:28But I agree that it's not yet, doing all of that and also you
- 27:33got problems, like, you know, what are called One-Shot
- 27:35learning? Can you learn from one instance
- 27:36of it because currently the training requires, lots and lots
- 27:39of Compute processing over data or in South play, can you have
- 27:45an accurate memory store that you update?
- 27:47Like, for example, you say, oh now now fact X is happened right
- 27:53now, you re continuing to your entire world based on fact X.
- 27:56Yeah. If there's a bunch of this stuff
- 27:58that all go and question is, is that one major invention is
- 28:01that, you know, five major inventions and by the way, major
- 28:05inventions are major inventions even all the amazing stuff we've
- 28:08done over the last five to ten years.
- 28:10Major intentions are agent. Hutchinson took things on the
- 28:12Age. Act one, I tend to think it's
- 28:14further than most people think, and I don't know if that further
- 28:17is it's 10 years versus five or 20 years versus 10 or 50 years
- 28:21was 20. I don't really know that in your
- 28:23lifetime. Do you think it's possible?
- 28:25Although I don't know, so it's really, I don't know, but let me
- 28:30give to other lenses. I think on the AGI question
- 28:33because the other thing that people tend to do is they tend
- 28:37to go, there's like this AI which is technique machine
- 28:39learning and it's totally just great.
- 28:40It's all going until just then there's a GI and who knows what
- 28:42happens with a GI and say, well, first is Agi is a whole range of
- 28:46possible things. Like what if you said, hey, I
- 28:48can build something. That's the equivalent of a
- 28:51decent engineer or decent doctor.
- 28:55But to run it cost. Me $200 an hour and I have a GI
- 29:00it does it right, right. But it's $200 an hour.
- 29:03You like, okay well that's cool and that means we can we can get
- 29:06as many of them as we need. But it's an expensive and so it
- 29:10doesn't like all of a sudden you know, Terminator or you know or
- 29:14preventing Fusion or something like that is a GI.
- 29:19Or potential version of AGI. So what is Agi is the squishy
- 29:22thing that people then go? And now magic, the second thing
- 29:25is the way that I've looked at the progress in the last five to
- 29:30eight years is we're building a set of iteratively better
- 29:33savants, write it just like the chess player was a savant and
- 29:38the savants are interesting Lee different.
- 29:40Now when does Savant become a general intelligence and when
- 29:44might Savant become a general superintelligence, I don't know.
- 29:48It's obviously Ali superintelligence already in
- 29:50some ways like, for example, I wouldn't want to try to play go
- 29:53against it and win. Try to win, right?
- 29:55It's a superintelligence what it comes to go, right?
- 29:58But like okay, that's great because in our perspective
- 30:00having some savants like this that are super intelligence is
- 30:03really helpful to us. So the whole AGI discussion I
- 30:06think tends to go a little bit hollywood-esque, you know, it's
- 30:09not Terminator. There is there's a sort of
- 30:13Argument that could be made. I mean you know humans are very
- 30:17human Centric about our beliefs and our intelligence, right?
- 30:20We don't have a theory of mind for like other animals.
- 30:23It's very hard for us to prove that other species, you know,
- 30:27have some experience of Consciousness like quality or
- 30:30whatever. So it's very philosophically the
- 30:33good use of the term by the way. Thank you.
- 30:36I studied philosophy though. I forgotten more than I
- 30:39remember. But you know, I mean the
- 30:41question someday will Figure out what it's like to be a bat
- 30:44probably not this year. Right?
- 30:45Right, exactly. Is that?
- 30:47That's - yes, that's - yeah. Yeah.
- 30:49So the machine, if the machines better than me at chess and go
- 30:52there there's a level of like, oh here I am saying, it doesn't
- 30:55have an experience, but it's so much smarter than me in certain
- 30:59domains. The question is just like, it
- 31:01seems like humans are not capable of seeing what it's like
- 31:05to be a bat. So will we ever really be able
- 31:07to sort of convince ourselves that there's something that it's
- 31:10like to be an AGI? System.
- 31:14Well, I think the answer is Yes, but it will require a bunch of
- 31:20sophistication. Like one of the things I think
- 31:21is really interesting, as we D, anthropomorphize the world a
- 31:24little bit and I think some of this machine intelligence stuff
- 31:28will enable to do that, is, what does it mean to understand X or
- 31:32no X or experience X or have qualia, or whatever else.
- 31:36Right? And right now, what we do is we
- 31:38say, well, it's all some kind of shadowy image from being human.
- 31:42So we tend to undercount like animals intelligence and people
- 31:45tend to be surprised He's like, look, you know, some animals
- 31:48mate, for life and every house they clearly the theory of the
- 31:50world. There's clearly stuff we're
- 31:51doing. We go oh, they don't have the
- 31:52same kind of Consciousness, we do and you like, well, they
- 31:55certainly don't have the same Consciousness but we're not
- 31:57doing a very good job of studying like, where it's
- 31:59similar, but it's different. And I think we're going to need
- 32:02to broaden that out some to start saying, well, when you
- 32:06compare us and an eagle or a dolphin, or a whale, or a
- 32:11chimpanzee or a lion, you know what are the similarities and
- 32:16differences, Says, and how this works.
- 32:19And I think that will also then be well, what happens when it's
- 32:22a silicon substrate, you know, do we think that Consciousness
- 32:25requires a biological substrate? If so why and you know, part of
- 32:30how, of course we get to understand each other's
- 32:32Consciousness. Is we we get this depth of
- 32:35experience where I realize is, it isn't?
- 32:37You're just a, you know, a puppet.
- 32:39You know I am. Yeah, I'm just a puppet.
- 32:42Well, we're we're talking to each other through Riverside.
- 32:44So you know, who knows? Right deep, thanks.
- 32:46All that but otherwise already ahead of you.
- 32:49I'm just it's already know. Yeah.
- 32:52But I think we're going to have to get more sophisticated on
- 32:54that question. Now, I think it's it's too
- 32:56trivial to say because it can mimic language in particular,
- 33:01interesting ways. And it says yes I'm conscious.
- 33:04That that makes a conscious that that's not what we use as an
- 33:06instance. And and part of it is like, do
- 33:08you understand like part of how we can understand each other's
- 33:12Consciousness? As we realize that we experience
- 33:15things in similar ways, we feel Joy and sometimes those who
- 33:18paints similar and that comes out.
- 33:21And that's part of how we We Begin understanding.
- 33:24I think it'll be really good that this may kick off kind of
- 33:28us being slightly, less kind of call it narcissistically
- 33:33anthropocentric in this and broader concept.
- 33:37As we look at this, I was talking to my therapist the
- 33:40other day and I was saying, you know, I did this like kind
- 33:45gesture but I didn't I feel like some profound like it just
- 33:50seemed like the right thing to do.
- 33:51I did it felt like did the right thing, you know, shouldn't I
- 33:55feel like more around it and you know her perspective is much
- 33:59more like oh what matters is like doing the thing.
- 34:02Not sort of your internal States about it, which to me would go
- 34:05to the if the machine can can do all the things we expect from
- 34:09sort of a caring type type machine.
- 34:12Like, why do we need to spend all this time when we don't even
- 34:15expect that of humans, too. He's feel the right feelings.
- 34:18I look, I think the real question is what you do.
- 34:21Now that being said, part of how we predict what you do is that,
- 34:27you know, you may not have liked at that moment.
- 34:29God, ha, ha. I think of myself is really good
- 34:32because I've done this kind thing which might do, it might
- 34:34be a better human thing supposed to like right, you know, the I'm
- 34:37doing this because I'm better than most people, you know?
- 34:39It's like no I'm doing this to this efficient grief that maybe
- 34:42I should have felt for the night, you know, you know what I
- 34:44mean? Yeah, but, Like it's the pattern
- 34:48in which you engage in these things and part of the feelings.
- 34:50And so we're this because that creates a kind of a reliability
- 34:53of pattern of do you see other people?
- 34:55Do you have the aspiration to have not just yourself?
- 34:59But the people around, you leading better and improving
- 35:02lives. And obviously, if that's the
- 35:04behavior that we're seeing from these things, then that's lot of
- 35:07it. And the only question is, is
- 35:08what is that forward-looking momentum on it?
- 35:11And I think, you know, among humans that comes to an
- 35:14intention, a model of the world and so forth.
- 35:16Amongst amongst machines that we just maybe they know where
- 35:19aligned well, like we've done a really good alignment with human
- 35:23progress. Do you think there will be a
- 35:26point in time where it's like an ethical problem to unplug it?
- 35:30Like I think I've like a bear, right?
- 35:32Like a bear is dangerous, you know, there are circumstances
- 35:36where pretty comfortable killing the bear, but if the bear like
- 35:40hasn't actually done anything, we've taken it, under our care,
- 35:43or like, we don't just like shoot bears in zoos, you know.
- 35:46No, and it costs us money to sustain the bearer to do.
- 35:49You think there are cases where we might say, oh man.
- 35:52Now there's an ethical question around.
- 35:55Unplugging some, I think it's a, when not an if ya, right?
- 36:00I mean, now it may be a when once again just like a GI that's
- 36:03a fair ways out, right? But it's a, when not enough.
- 36:06And by the way, I think that's Again, part of the progress that
- 36:09we make. Because we think about like,
- 36:10okay? So how should we be treating?
- 36:14Because we know like, for example, if you go back, 100 150
- 36:18years. The whole concept of animal
- 36:19rights doesn't exist in human, you know, it's like, hey, you
- 36:25want to, you want to torture Animal X to death?
- 36:27You know, you're allowed to do that.
- 36:29That's an odd thing for you to do, right?
- 36:31And maybe it's kind of like, like, distasteful, like grungy
- 36:34bad in some way, but you know, it's it's either like okay
- 36:40where's now you're like oh that person is like going out to try
- 36:46to go torture animals. All's we should like get them in
- 36:48an institution, right? Like that's not okay.
- 36:52And I think that, you know, we like what are what is that
- 36:55further progress for the rights and lives?
- 36:58And I think it will ultimately come to things that we think are
- 37:03when it gets too kind of like things that have their own
- 37:05agency and have their own you know kind of Consciousness and
- 37:09and and sets of existence. We should be including all of
- 37:13that in some Grand or elevated you know kind of right.
- 37:17Conceptions. All right, so back to my
- 37:20listeners who, you know, want to know where to invest and make
- 37:22money off this and, you know, they'll be different from
- 37:24quality and Consciousness, also white, I mean, who do you?
- 37:28Who do you think are like the key players in the models and
- 37:32then obviously there are more sort of?
- 37:34I don't know if we're calling them vertical Solutions or
- 37:36product oriented or whatever. However you think about them.
- 37:38But starting with the models, like, who do you see as sort of
- 37:41the real players right now? Are you counting out a Google or
- 37:45do you think they'll still sir? To show up.
- 37:47Oh no I think Google show up and obviously you know open a
- 37:51Microsoft has done a ton of stuff I co-founded inflection
- 37:54last year with Mustafa Suleiman. We obviously, you know, think we
- 37:57have at just you know, amazing team and you know I do see a lot
- 38:01of teams and that's to build sort of the foundational tool.
- 38:04Yeah. They're going to well, the
- 38:05building their own models and they're going to build some
- 38:07things off those models. We haven't really said, what
- 38:09they are yet, but that's obviously going to be kind of
- 38:12new models, adapt another Greylock Investments building,
- 38:15its own models care. Doctor is building its own
- 38:18models. Anthropic is building its own
- 38:20models. And anthropic is, you know, I
- 38:22think because, you know, Dario and the crew is smart folks.
- 38:26I'm open to either, they're doing stuff with in a kind of a
- 38:29similar research program that open AI is doing.
- 38:32And so, I think those are the ones that I probably most track
- 38:35characters an interesting case. And, you know, we're still
- 38:37learning more about that company.
- 38:39You know, I was first to report, they're looking to raise 250
- 38:42million dollars. My understanding, is that?
- 38:45What's interesting is, they're building the models, but Then
- 38:47for a particular use case, right?
- 38:49Or like it's really a question of Leverage or, like, do people
- 38:52need to build the models to be competitive, or do you think
- 38:55there will be, you know, what you might, you know, a skin or
- 38:58whatever, you know, building. Can you build a great business
- 39:01on top of stability or open a, or do you need to do it
- 39:05yourself? I don't like what?
- 39:06I think you can, but the way you do it is you can't say it's
- 39:09because I have unique access to the model.
- 39:11It has to be you know, I have a business that has Network
- 39:14effect. So I'm well, integrated and
- 39:15Enterprise or I have another Deep stack of technology that
- 39:18I'm bringing into. It can't just be, I'm a
- 39:20lightweight front end to it because then other people can be
- 39:24the lightweight front end so you can build great businesses.
- 39:26I think with it. I do think that people will both
- 39:30build businesses off. You know, things like the open
- 39:33Ai apis and I think people will also train models because I
- 39:39think one of the things that will definitely happen is a lot
- 39:41of you to like also not just will large models be built in
- 39:45ways that are, I think. Interesting and compelling, but
- 39:48I think a bunch of smaller models were built that are
- 39:50specifically tuned and so forth and there's all kinds of
- 39:53reasons. Everything from you can build
- 39:55them to do something very specific but also like inference
- 39:57cost, does it? Does it run on a low compute or
- 40:00low power footprint, you know, etc.
- 40:03Etc. You know, AI doctor a I tutor,
- 40:05You Know, da and on a cell phone and so you know, I think the
- 40:09short answer to this is. All right.
- 40:12Do you think we are in a compute arms, race still or do you think
- 40:15this is going to continue where Just if you can raise a billion
- 40:19dollars to buy sort of GPU access basically from Microsoft
- 40:25or Amazon or Google, you're going to be sort of pretty far
- 40:29ahead, or how do you think about that sort of the money and
- 40:31Computing race shaping out? So it's kind of two lines of
- 40:34Trends is one line, which is the larger and larger models which
- 40:38by the way, you say well, okay? So scale, compute and 1X flap
- 40:41goes to 2x of flops and as your performance function, go up by
- 40:45and it doesn't have to go up by a hundred percent.
- 40:46Or two extra plus 1X. If you go up by 25%, but
- 40:50sometimes it really matter is code, in doctors, you know,
- 40:53legal other things was like, actually impacted even though
- 40:56it's twice as expensive, a 25% increase in, you know, twice as
- 41:00expensive or commute. That's why 5% increase in
- 41:02performance is worth it. And I think you then have a
- 41:04large scale model like a set of things that are kind of going
- 41:07along, you know, need to be using the Large Scale Models.
- 41:11Then I think there's a set of things that don't have that that
- 41:14need and like, for example, that's one of the reasons I
- 41:15wasn't really. Surprised at all by the
- 41:17profusion of image generation because those are you know
- 41:20generally speaking trainable for a million to 10 million dollars.
- 41:25I think there's going to be a range of those.
- 41:27I think you know maybe someone will figure out how to do a
- 41:30hundred million dollar version, and once they figure out how to
- 41:32do 100 million dollar version, someone else will figure out how
- 41:34to do they, 30 million dollar version of The 100 million
- 41:36dollar version. And this should I think there's
- 41:38a second line. Capital is, yes, exactly.
- 41:41And there's this second line going on where all of these
- 41:43other smaller models will fit into interesting businesses.
- 41:46And then I think, A lot of people will either deploy an
- 41:48open-source model that using themselves train their own
- 41:51model, get a special deal with kind of a like, a model provider
- 41:57or something else has a way of doing it.
- 41:58And so, I think the short answer is that will be on both and you
- 42:01have to be looking at this from, what's the specific that this
- 42:05business is doing? You know, the classic issues
- 42:08are, you know, how to go to market.
- 42:09How do you create a competitive mode?
- 42:11What are the things that give you real enduring value?
- 42:14That people will pay for in some way and Us all of those
- 42:18questions to apply, but there's going to be a panoply of answers
- 42:21depending on the different models of how it plays out.
- 42:23You think spend on this space in terms of computing will be
- 42:27larger in 24. Yes, and then larger and 25.
- 42:30Yes. Unquestionably we're on, you're
- 42:32still on the road is unquestionably.
- 42:34That's great for a certain company that you're on the board
- 42:37of a. Yeah, yes well and it's not just
- 42:40great for Microsoft, there are these other ones you know AWS
- 42:43Google. But yeah it does feel like On
- 42:47somewhat sleepy. Here you have any view there?
- 42:50Well, I think they have begun to realize what I've heard from the
- 42:54market is that they've begun to realize that they should have
- 42:57some stuff here? I don't think they've yet gotten
- 43:01fully underway. I think they are trying to train
- 43:06some large language models themselves.
- 43:09I don't know if they even realize that there is a skill to
- 43:14training. Those large language models
- 43:15because like you know, Times people say well you just turn on
- 43:18and you run the large language model, the training regime that
- 43:21you read in the papers and then you make some, we've seen a lot
- 43:24of failures, right? People trying to build these
- 43:26things and failing to do so. So you know there's an expertise
- 43:30that you learn in doing it as well.
- 43:32And so I think if sorry your own a Microsoft is around, open Ai
- 43:36and Google is around anthropic is Amazon going to be around
- 43:40stability? That's that's sort of the
- 43:42question that I'll put out to the world.
- 43:45I don't know if you have, I certainly Don't know anything.
- 43:47And in the case of, you know, very, very, very politely said,
- 43:55anthropic and open a, I have scale with huge models civility
- 43:59is all small models. So yeah, interesting.
- 44:05I don't think I've asked you sort of directly about sort of
- 44:09stepping off the open a.i. board or what I mean, I would assume
- 44:12you would prefer to be on the board or, you know.
- 44:16Yeah. Well.
- 44:16So, Look, it was a funny thing because, you know, I was getting
- 44:19more and more requests from various Greylock portfolio
- 44:22companies because we've been investing in AI stuff for over
- 44:25five years, like, real, a, I not just the, we call it soft, we
- 44:28call software AI, but actual AI companies, and I was getting
- 44:33more and more requests to do it. And I was like, going, oh no,
- 44:35you know what I did before was I well, here's the channel.
- 44:38You like here is the guy who the person who handles the API
- 44:42requests go, talk to them. Like, why can't you help me as
- 44:44well? I'm I'm on the board.
- 44:46I have respond. I did not.
- 44:47Do you doing that? And then I realized that shit.
- 44:50It's gonna look more and more. Like I might have a real
- 44:54conflict of interest even as we really carefully navigating it
- 44:58and it was like a look. It's really important because,
- 44:59you know, various forces are going to kind of try to question
- 45:03the frankly, super deep Integrity of open my eyes.
- 45:06Like, look I Sam, I think it might be best even though.
- 45:09Okay, main, I fan and now I to helping, I think it may be best
- 45:13for open a.i. and generally to stop off the Avoid a conflict of
- 45:18interest. And, you know, we talked about a
- 45:21bunch of said, okay, fine. We'll do it.
- 45:22And, you know, I obviously like I had dinner with Sam last night
- 45:25and most of what we were talking about was, you know, kind of the
- 45:29range of, okay, what's going on? And what are the important
- 45:32things that open as you solve and how should we be interfacing
- 45:35with governments, the government's understand, what
- 45:37are the key things that they should be in the mix and what
- 45:40great future things for Humanity or really important not to
- 45:43fumble in the in the generally like everyone going?
- 45:46Well, I'm Ring, I'm wearing I'm wearing and that you know we're
- 45:48talking about all that stuff and I said, oh I got a question for
- 45:51you. He's like, yeah, okay, it's like
- 45:53now that I'm no longer on the board.
- 45:55Could I ask you to personally? Look at unblocking my portfolio.
- 45:59He's saying the API because I couldn't ever asked you that
- 46:02question before. Unethical.
- 46:04All right, how I'm not on the board.
- 46:06So can I ask the question is like sure I want to migrate
- 46:09right? And that's the substance of it
- 46:11which I never would have done before but that wasn't why?
- 46:14I mean I obviously love Sam. His entire team.
- 46:17So the fact that you're sort of a democratic super donor was
- 46:22that in the calculus or always I mean we are seeing romelo can't.
- 46:25Well I I didn't think that at all coming to this conversation
- 46:27but just hearing what you're saying, the government's looking
- 46:30at it. Now it feels like Republicans
- 46:31are like trying to find something to be angry about
- 46:35well, with these AI thing. I don't quite know the
- 46:38unfortunate thing about the most vociferous of the Republican
- 46:43media ecosystem, as they just invent.
- 46:46Right? Like they're hallucination fall
- 46:48out, right? Right.
- 46:50I mean, it just like, I mean, the amount of just like, you
- 46:53know, 2020 election denial and all the rest.
- 46:55Would you can tell from having their text released from Fox
- 46:58News? It like here are these people
- 47:00who are on camera going on where you have a question about you
- 47:04know, what happened election and they're texting each other
- 47:06going. Oh my God, this is insane.
- 47:07This is a coup today and you're like, okay anyway so so all
- 47:12look, they don't require truth to generate heat and friction.
- 47:17Right sleaze. So that was wasn't that no no.
- 47:19It's just really, it's kind of the question of part of the
- 47:23thing. Like it's when you're serving on
- 47:27a board you have to understand what your mission is very deeply
- 47:30and and to navigate it and part of a fire one c.
- 47:33Three boards is to say, look I contribute by being a board
- 47:36member and helping and navigate various circumstances and all
- 47:39the rest. But like for example I can
- 47:41continue to be a counselor and an aide to the company, not
- 47:43being on the board, right. And one of the things I think is
- 47:46going to Very important for the next X years, for the entire
- 47:50world to know, is it open a guy takes its ethics?
- 47:52Super seriously, right. As do I, how does that fit with
- 47:55having invest? I mean, obviously, there are
- 47:58lots of companies that do great things.
- 47:59They have investors, I believe in companies probably more than
- 48:02personally, I believe in like Charities to accomplish things,
- 48:05but I The Duality of open area is extremely confusing.
- 48:09Like degre lock itself, invests, a lot or you invested early as
- 48:13an angel. I was, I was just nice.
- 48:15I was the founding investors an angel.
- 48:16As a program related investment from my Foundation.
- 48:20Because like, I started, I was among the first people to make a
- 48:24philanthropic donation to open a, I just straight out, you
- 48:27know, here's a grant by Wednesday, 3:00, and then Sam
- 48:31and crew came up with this idea for doing this commercial LPG.
- 48:34And I said, look, I'll help and I have no idea.
- 48:37If, if this will be an interesting economic investment,
- 48:40they didn't have a business plan.
- 48:41They didn't have a revenue plan and ever product plan.
- 48:44I brought to Greylock, we talked about it and I said, look, We
- 48:47think this will be possibly really interesting technology
- 48:49but you know part of our responsibility to our LPS which
- 48:52you know, glutes whole bunch of universities or else we invest
- 48:54in businesses and the reason our business plan there is that
- 48:59they're on khosla did Coast we invested wild thing that.
- 49:02Anyway, we don't care, that's sort of what vinod wants to
- 49:04project anyway, so yeah. Yes, that's exactly.
- 49:07As I put them 50 and then I think it was the only Venture
- 49:09fund investing in that round. Yeah, but like there was no
- 49:12business plan, there's no Revenue model does not go to
- 49:13market. There was no, he looks.
- 49:15And basically says someday we're going to AGI.
- 49:17And we're going to ask it how to make a bunch of money, like is
- 49:19he that's a joke, right? Or like, how much is he joking.
- 49:22It's not a 100% joke and it's not a zero percent joke.
- 49:25It's a question around the mission is really abound.
- 49:29How do we get to a gir as close to a jazz useful and make it
- 49:32useful to humanity? And by the way Age, The Closer
- 49:34you get to a GI the more interesting Technologies,
- 49:37Fallout, including the ability to have the technology itself
- 49:41solve various problems. So if you said, we have a
- 49:44business model problem, it's like, well, ask the Ask thing.
- 49:47Now if you currently sit down and ask, you know, chat gbt what
- 49:50the business model is, you'll get something pretty vague, and
- 49:52generic there wouldn't get you a meeting with the Venture
- 49:55capitalists because it's like the we will have ad-supported
- 49:58you like okay, whoo, right, don't you have a company that's
- 50:02trying to pitch decks now hours? Oh, yeah, Tom.
- 50:04No, and it's awesome, but by the way, that's the right kind of
- 50:07thing because because what it does is you say, hey, give me a
- 50:11set of tiles together with images and graphics and things
- 50:14arguing X. And then you start working with
- 50:16the Try to improve it say, oh, I need a slide, the does this, and
- 50:19I need to catch your headline here and right.
- 50:22And you know, Derrida and then, you know, obviously can edit it
- 50:24yourself and do it. So that's, that's the kind of
- 50:26amplification now. You don't say give me my
- 50:29business model right here. Like, I have this business model
- 50:33like articulated, right? Exactly politics.
- 50:37I'm gonna be like Silicon Valley, you know, has like
- 50:39worked on everybody be able to, you know, everybody can get
- 50:42along, there's sort of competition, but then you sir to
- 50:44still stay close to him everybody.
- 50:46Like you you especially like are good.
- 50:49You know, you were in the PayPal Mafia with a lot of people who
- 50:52are fairly very conservative. Now, the Trump years broke that
- 50:57in some ways. And did you maintain those
- 50:59relationships? I see headlines that.
- 51:01Say, you're friends with Peter teal.
- 51:03What's the state of your friendship with Peter teal?
- 51:05And how did it survive? I guess the Trump years is the
- 51:08question. Well, I think the thing that
- 51:11Peter and I learned when we were undergraduate Stanford together
- 51:13is it's very important to, because we, you know, I was, I
- 51:17was, I was a lefty, he was a righty.
- 51:19We'd argue a lot to maintain conversation and to argue.
- 51:23Thanks, you know, no lie. It's difficult to argue on
- 51:28things that feel existential and, and, and it is ethically
- 51:33challenged is things around Trump.
- 51:37You know, the, you know, Trump feels to be a, you know, a
- 51:41corrosive acid Upon Our democracy.
- 51:43That is disagreeing us and staining us to the world.
- 51:46Old. And so a be a dispassionate
- 51:49argument about it is challenging and, you know, to you have, you
- 51:53know, it ends up with some uneven ground and statements
- 51:56like I can't believe you're fucking saying that as part of
- 51:59dialogue, but on the other hand, you know, maintaining dialogue
- 52:03is I think part of how we make progress has society and and you
- 52:07know, I basically am sympathetic to people as long as they are
- 52:11legitimately and earnestly and committed to the dialogue and
- 52:14discussion of Truth between them, right?
- 52:16And and not committed. Otherwise and so you know, there
- 52:20are folks from the PayPal years that I don't really spend much
- 52:24time talking to. Yeah.
- 52:26Right. And there are others that I do
- 52:29continue because that conversation ever discovering
- 52:32who we are and who we should be is really important and you
- 52:34can't allow your own position to be the definer.
- 52:38I almost goes back to what we were talking about.
- 52:39The AI side which is make sure you're talking to other smart
- 52:43people who challenge you to make sure you're doing the right
- 52:46thing. And that's I think a good,
- 52:48General Life principle. I feel like part of what my
- 52:52dream of like the Silicon Valley world is that we have these, you
- 52:54know, we have Twitter is like the open Forum where however
- 52:56being sincere sort of on the level debates, but then you see
- 53:00something like, you know, politics that's modern Seinfeld
- 53:03show I've got or not Seinfeld on Springer.
- 53:05Jerry Springer, you know, like fee?
- 53:09Yeah, that's your right. But at least I just feel like
- 53:12the sort of like, if the arguments are on the level.
- 53:14Issue is my problem with some of the Sort of yeah Peter teal
- 53:19arguments that he's not actually publicly advancing his beliefs
- 53:23in a sincere way and that that's almost more corrosive though
- 53:26that's totally gross of and as much as that's happening.
- 53:28It's terrible. And that's one of the things
- 53:29that I, you know, in conversations.
- 53:31I have I push people including Peter on a lot.
- 53:33Yeah. Are you still, are you still
- 53:36gonna donate a lot? Or what was which your are you
- 53:39as animated about the Democratic party and working through sort
- 53:42of doing our channels at the moment?
- 53:44Well, what I would say is I think that we have a
- 53:47responsibility to try to make like with its kind of a
- 53:50Spider-Man ethics with with power comes responsibility with
- 53:53welcomes responsibility and you have to try to help contribute
- 53:56to what is the better society that we should be living in
- 54:01navigating in. And so I'm I stay committed on
- 54:03that basis and I do think there are some really amazing people.
- 54:07The administration I think Biden is kind of a good everyday guy.
- 54:11Yeah I think is actually in fact good for trying to build bridges
- 54:14in the country. I think there are people like
- 54:16Such a raimondo and security Buddha judge, who are thinking
- 54:20intensely about technology and kind of what could be done in
- 54:22the future. And I think there's other folks.
- 54:24Now, I think there's a bunch of folks on the Democratic side
- 54:27that I think are more concerned with their demagoguery than they
- 54:32are with the right thing in society.
- 54:34And so I tend to be, you know, unsympathetic to, you know, kind
- 54:41of the, you know, Michael Moritz is Sequoia.
- 54:44I think just wrote an op-ed. Sort of criticizing San
- 54:47Francisco government, you know. And there's, there's certainly
- 54:51this sort of, I don't know, get out the woke critique of the
- 54:55democratic party. Or yeah, I'm curious if there's,
- 54:58there's a piece of it sort of, outside of the governance that
- 55:02you're well, I definitely like, you know, now I find the kind of
- 55:06the interesting thing on woke is like, well, we're antiguo.
- 55:08Can you like, well, don't you think being awake is a good
- 55:10thing? I mean, it's a funny thing,
- 55:11right? But the answer of the
- 55:13ill-defined nature of woke is like key to the Because it's
- 55:17like, oh what's the substitute thing?
- 55:19You're saying there and you know, I mean we're seeing you on
- 55:22tweet about race right now which is a sort of a terrified anyway.
- 55:25Yeah. And so I think that the question
- 55:28on this stuff is to try to say look, people have a lot of
- 55:31different views and a lot of different things.
- 55:33And some of those use our are bad, especially in kind of
- 55:37minority, and need to be advocated against in various
- 55:39Ways. Part of why we like democracy is
- 55:43that discourse I'm very concerned about the status of
- 55:46Public discourse and obviously most people tend to focus that
- 55:49around social media which obviously has some vitamin
- 55:52things that we need to talk about.
- 55:55But on the other hand they don't track Like These, like opinion
- 55:59shows kind of like Fox News that represent themselves implicitly
- 56:03as new shows and saying, this is the following thing, like
- 56:08there's election fraud in 2020. And then when they were sued for
- 56:11the various forms of defamation, they said we're just an
- 56:13entertainment show, we don't do anything like news, so we have
- 56:16that. Within that we are already
- 56:18struggling on a varieties issues within society and we I think we
- 56:21need to sort them all out. Is there anything on the AI
- 56:24front that we missed or that? I should ask that you want to
- 56:28make sure to talk about I think we covered so much great ground.
- 56:30Yeah and we can do it again right gonna tell ya.
- 56:33I love it. I love it.
- 56:34I mean I do think Yeah. And this was a great, all the
- 56:38things are interested in, I'm interested in so is great.
- 56:41I really enjoyed having you on the podcast and thank you,
- 56:44thanks, likewise, and I follow the stuff you do and it's cool
- 56:47and keep doing it. Thank you.
- 56:50That's our episode. Thanks for listening.
- 56:51Shout out to our editor, Tommy Heron, my chief of staff Riley,
- 56:54kinsella, Young Chomsky who made the wonderful Music tune in next
- 56:59Tuesday for another great episode.
- 57:01If you have feedback, send it to me at erekat newcomers dotco, I
- 57:04always loved Scoops and tips. And please rate, like, comment,
- 57:09subscribe, do all the things that is a brand new podcast and
- 57:12I sincerely would appreciate your support all over the
- 57:16interwebs. Thank you, goodbye, goodbye
- 57:19goodbye. And please rate, like, comment,
- 57:09subscribe, do all the things that is a brand new podcast and
- 57:12I sincerely would appreciate your support all over the
- 57:16interwebs. Thank you, goodbye, goodbye
- 57:19goodbye.