Reid Hoffman on AI's "crescendo moment," regulation and real-world applications | E1739

0:00I believe that we will have a personal assistant for any professional informational task, which is professional task. process information, you do something with it. You make an investment decision, write a memo, write a prescription, something like that. Two to five years for every professional activity. Now, what the adoption will look like, will it be useful to essential? There's a variety of things. But as an amplifier, as amplification intelligence versus artificial intelligence, it is off the charts amazing. This Week in Startups is brought to you by Release. Large enterprises pose unique challenges for SaaS startups.

0:37Unlock customers with unique needs for private and single-tenant hosting without the toil of DIY with Release Delivery. Get your first month free at release.com slash twist. The Microsoft for Startups Founders Hub helps all founders build a better startup at a lower cost from day one. Startups get up to$150 ,000 in Azure credits, access to free open AI credits, free dev tools like GitHub, technical advisory, access to mentors and experts, and so much more. There is no funding requirement and it only takes minutes to join. Sign up today at aka.ms slash thisweekinstartups. And Veed makes it super easy for anyone, yes, you, to create great video.

1:29Filled with amazing features like templates, auto subtitles, text formatting, auto resizing, a full suite of AI tools, and much more, Veed gives you the tools to engage your audience on any platform. Head to Veed.io to start creating incredible video content in minutes.

1:49I'll just say to everybody, it's just great to have Reid Hoffman back on the program. You know, read as one of the founding members of paypal co-founder and ceo of linkedin which has just become a juggernaut 875 million members almost a billion now and he's a partner at graylock board member at microsoft and now the co-founder of inflection ai so it's just great to have you back on the program the last time you were on god i'm trying to remember i came to see you at linkedin when i was just starting this program exactly and uh we just talked about life and all that kind of stuff so god there's so much to talk about uh just when i say linkedin and you look back on what you built there and i say 875 million members and going to a billion it's just quite extraordinary is it not the staying power of linkedin what do you attribute it to just the network effect well so by the way one of the great things about doing this interview today This is May 5th.

2:46It's the 20th anniversary of the LinkedIn public launch. We are doing this discussion on that. So, that's a cool little milestone in my life and the world history. I think a lot of the... LinkedIn is the turtle that made it. And it's partially because it stayed really true to a mission. It's like, this is the role we plan people's lives. This is the thing we do or about you know time efficiency and time saving not time spending or about amplifying what you're doing in your you know job search or your career or your work or information about that we're not doing anything else we just keep going at it even when in the very earliest days you know launching you know you'll remember this because we're both old men of the internet the um you know may 5th 2003 when we launched the hot thing was friendster yes literally the only way that i could get journalists to talk about us was to talk about us as friendster but for business which is totally nonsensical it's like there is no friendster but for business that's like it's yeah it's like saying you know like pizza but for driving your car you're like no okay not sure what that is sure it is just amazing also how they've just kept the feature build going like the if you want to advertise or if you want to hire people uh it just works amazing and i i have been playing with chat gpt4 and plugins and i don't know about you but if you were to look at how exciting this is for two old men of the internet a hundred years i think old between the two of us here we remember dial up we remember pc on everybody's desk office windows the gui cloud mobile and broad 300 bod like my first people don't have any idea what we're talking about my first modem was a ventel 300 bod and then i got the haze 1200 and 2400 man those were upgrades yes exactly but when you when you when you look at what has happened in the last call it 90 days uh with chat gpt4 plugins auto gpt how does this compare in your mind the tingly feeling you probably got when you got your first pc or used the first windows interface or used a mobile phone iphone or cloud computing or broadband even how does this compare to you you know i think it's the crescendo moment like when i say this is bigger than the internet or bigger than mobile or bigger than cloud it's because it's building on all of them it's the tsunami that you're surfing at the top of it and it's very similar to but i think your your lens is correct is like the first time you were you know playing with a you know a pc was like oh my god like there's all these things this can do and you just kind of like like your very first metaverse was playing with the pc it's like you were you were kind of virtually transported in this thing and and i think ai the current scale compute because it's really a kind of application of scale compute which given its cognitive functions is a good thing to call it you know artificial intelligence but uh it's really like this kind of scale compute discovering things that we had never we'd never gotten anything even close to this before and the way that it's kind of human amplification is you know just kind of super interesting and so um you know so that that that aha moment like i remember when i was playing with gbd4 uh last year uh july and august that was on the board of open ai and i started like doing prompts and questions and i was like oh this is good this is amazing yeah right you know and it's not to say you don't worry about the concerns and you know human amplification you also have i think all the risk stuff should really be human amplification of bad bad actors like what happens with bad actors using it we can get to that as it's relevant but it's like oh my gosh now there's superpowers like i have superpowers where i can create an image where i couldn't create an image before i have superpowers where I can say, I would like to create an epic poem about Jason and his quest to make poker the relevant lens into thinking about investing.

6:59And make that an epic poem, and I can do it now, because it's like this superpower, which is really amazing. Anyway, so it is the most significant moment of technology in our lives so far. and maybe in our lives on the whole story. Our lifetimes. I mean, I am exactly where you are. I can't stop playing with this stuff. I can't stop talking to people about it. I had Brian Chesky on on Wednesday. He's just totally all in, and he's rethinking the entire conception of Airbnb. I had Aaron from Box on. He's just turned the entire company around and pivoted towards that, and then here you and I are talking about it in the same light, which is this isn't a drill.

7:43I gave everybody on the company, my investment company, ChatGP4, had them all sign up for it. About five of the 20 in the last, call it 60 days, have started taking large swaths of work and offloading it and getting phenomenal results in a technology that's crashing constantly. The plugins don't work. The webpage surfing doesn't work. It's got all those clunky kind of things at it. So let me ask you this just because it's fun to have the crystal ball. for white collar workers this year and again it's not a scare tactic it's just pragmatic what percentage of work that you and i do as investors and entrepreneurs communicators and just people on our teams if they dedicate themselves to this what percentage of their workload could be offloaded to chat gpt4 in 2023 spends a little bit on the job yes and you know Because, for example, if you're report writing or you're notes minute taking, then the answer is probably in the 50 % to 80%.

8:43Correct. Although, by the way, you may still be doing it. You just may now be doing it a lot better within that time frame. Like, what used to take you three hours to do now will take you 15 minutes. And you may still say, well, I'm going to take another hour and make it a lot better. Put a little polish on it, right. Exactly. you know amplified add some creativity put some questions in other kinds of things um you know in our venture business i think very little um you know it's not to say that very little later but very little 2023 um and it's partially because like if you look at gbd4 what it is a stunning superpower on it's like a research assistant that instantly delivers a synthetic kind of a synthesis and summary result about a prompt that, you know, kind of in your pocket.

9:30Now, obviously, it's creative and it can generate a Star Trek episode and do other kinds of things as ways of doing it. So, it's not just like a report, but it really takes all this stuff and kind of creates a synthesis of it. Now, one of the weird things about what we do in venture is we are looking for that needle in a haystack. We're looking for the extraordinarily different things. So, you say, well, say I was investing in a new area, which I don't really do. and I was going to go, oh, now I'm going to start investing in mRNA application companies. Then actually, by the way, using it as a research assistant could be used, it was like, oh, what are the key issues that I should look at when I'm looking at this company?

10:05You got the world's smartest assistant associates sitting next to you, catching you up. Right. Going right now, right away. Now, of course, there's some hallucinations, so you have to pay some attention to it and occasionally, but it's immediate and right there. And so that's really helpful. Yeah. But of course, if I was turning to do like mRNA, you know, kind of, you know, drug discovery investment, then I would have to go, you know, to your guys podcast and all in. Right. Because like casually is a dumb way to be doing investing. Yeah. And so it would help kind of get into it. But once you had expertise, like, like one of the questions I asked GBD4 last August when I was playing with this, because I was like, okay, you know, how much is this job replacement on stuff?

10:48All right, well, how would Reid Hoffman make money by investing in artificial intelligence was my prompt. And it gave what was like kind of, I'd call it an MBA professors, you know, business school professors, analysis that doesn't really, very smart person, but doesn't really understand venture capital. So it was like, right, well, you would look at the areas that had the largest TAMs, and you would look at which products and services had the greatest substitution or alternative impact because of this AI technology, then you go find teams that were capable of addressing that, and then you would invest in those teams.

11:23And that's what you do. And you're like, well, that makes total coherent sense. It's a coherent pitch. And that's not what we do. It's like telling Tarantino how to make a movie. It's like, yes, you do need a camera, you need film, and you do edit it. Yes, thank you. Exactly. Thank you. Whereas what we do is we look for like the really kind of stunningly bold idea from an entrepreneur that says, look, I recognize this new chance of going to market, this new kind of product, maybe it's a replacement. The fact that it happens to be a large TAM or not now isn't really the relevant question. The relevant question is what future TAMs look like.

11:59Could you reduce the market? Yeah. Yes. So there's all this stuff about the way we look at it that is not that. And so you're like, great. Not yet. Not yet. A useful tool. It will be. I believe that we will have a personal assistant for any professional informational task which is professional task i process information you do something with you make an investment decision write a memo write a prescription something like that two to five years for every professional activity now what the adoption will look like will it be useful to essential there's a variety of things but as an amplifier as amplification intelligence versus artificial intelligence yeah it is off the charts amazing And that's part of the reason why, you know, as you know, I, you know, published a book with it, you know, six weeks ago or something.

12:49Developer talent is the most precious resource for B2B startups. You know that. And you want your developers focused on product, not on compliance, right? When you're selling B2B software to large enterprises, you need to jump through a ton of security and compliance hoops. And one of those hoops is large customers need you to host your software on their cloud. And you need to build that out on a per customer basis. Think about that. So B2B startup companies constantly face this dilemma. Do you keep developers focused on infrastructure, which could hurt your product velocity? Or do you keep them focused on the product velocity, which would then delay your ability to close large customers?

13:31Well, I have a solution for you and it's called release delivery. What release delivery does is it automates the creation of enterprise class app delivery for private clouds and single tenant applications. Basically, this lets you deliver your software seamlessly into any customer environment. This will unlock a ton of revenue potential for you and release delivery will put all the tedious stuff on autopilot for you. So you can turn your ideas into apps and deploy those apps quickly and flexibly into their clouds. So here's your call to action. Let release show you the power of release delivery and get your first month free at release.com slash twist what a domain name r e l e a s e.com slash twist it's up to ten thousand dollars in value at release.com slash twist i think the human augmentation right this amplification this becoming a mutant becoming the 10x developer that's here right now i mean it just feels like all of a sudden you can fly and it's like oh wow i thought only superman could fly and like it's like no no all members of justice league cannot fly it's and as but one example of that have you seen the code interpreter that they launched over the weekend and played with that yet no i haven't checked so code interpreter lets you open a csv file you use the code interpreter i just start taking random csv files i found on like public information websites one of them was like number of electric vehicles and hybrids you know vin numbers whatever it's on some public thing and i said tell me about this and it was like or tell me trends about it give me some charts and it just starts spewing spewing out charts and i'm like okay that's a data scientist job and i gave a speaking of the other day and there was a data scientist there who did the data science for this hundred person hospitality company and i showed him this live on stage at their off site and he said uh that's about half of what i do all day is these requests and i said guess what now you don't have to do the requests you're getting people can do them themselves and then come to you for the more sophisticated one he said oh my god thank you because i'm like a month behind on getting people this basic stuff so everybody is going to be able to be a data scientist now not to his level maybe they're 60 but boy does that change everything yes and and by the way it helps him be a lot better too because as he refines on the okay what questions you really ask what data should you really look at how do you look at this data the right way all of those kinds of questions it makes us all better improves our capabilities yeah it's it is truly fascinating i guess everybody wants to talk about the downside of it so i think we should hit that right now real quick because what the consensus with smart people i've been talking to is this is augmenting everybody making everybody more efficient and we got a hell of a product roadmap so what aaron levy said today is yeah this is gonna make everybody 30 that was his number brian chesky picked the same number i had picked the same number i picked 30 brian said 30 or 40 and aaron said 20 to 30 so you're like okay and you said something similar like depending on the job it could be 50 could be 80 or maybe you go back and put another hour into it everybody who is in the industry understands this is going to just make us get through the backlog of product roadmap customer service whatever the issues in your organization are you're just going to move 30 faster a year which means compounding everybody's going to be twice as efficient every two or three years maybe you could speak to what would be your most charitable hey don't worry about the ai uh it's going to just make us all amazingly better our jobs and there's so many more problems off and then maybe you could get steel man the other side which is hey maybe we got to think this through because i know that biden had some people over today at the white house to talk about this yeah i think it was yesterday but um was yesterday yeah Yeah, it was yesterday.

17:22And so, let's see, you know, I'm fundamentally a believer that we that this will be really great for the vast majority of people. Yes. And even to say, look, in the case where you say, you know, okay, customer service cost center will actually, in fact, hit a bunch of jobs. Even in that, you say, well, but you can make AI to give skills, upskill, retrain, other kinds of things. You know, like, you know, well, maybe I can join sales or maybe I can do other things as kind of ways of doing this. And that is part of, I think, the reason why I'm ultimately bullish. I think the positive of it, it is an amplifier to human capability.

18:00Just like you'd say, every smartphone has a medical assistant, has a tutor for everything, has a kind of a personal AI assistant for whatever problem you're trying to navigate and kind of help you with it. And I think that's enormously kind of amplifying. Like you could even like say, well, I was driving somewhere and my car kind of broke down and I can ask the personal AI about it going, okay, these are what I see the symptoms of, you know, what should I do? Right. Like what's going on? Hugely, hugely beneficial across all of these angles. And that's the positive. And a little bit of the reason I was in is like, look, you know, like you said, well, 50 % of the work of my job goes away.

18:47It's like, well, not necessarily as much as the really bad parts. And then you can amplify it to getting a higher standard on what you're doing. And, you know, and that's really good. Well, it's really interesting about what you said there. I just want to amplify two things. One, you're getting rid of your chores, the things that are the most arduous, which then there's, you're going to find as a human, more creative things to do. If you're doing customer support, well, maybe you're doing customer training. Maybe you're doing customer success. Now you moved up from just telling people how to log in or reset their password.

19:16The second thing you said, though, is actually the first person I've heard say this, well, if this is this good to replace your job or a large portion of it, it's good enough to retrain you for another job. Exactly. That's actually a very important insight. I think that people, you're the first person I've heard actually say that. Yeah, no, exactly. Because it's like, we can make it part of the solution. So then when you get to the downsides, you go look, I think the two most obvious downsides are it's transition. And the majority of human beings don't like transition. They like feeling comfortable where they are.

19:48Yeah. You know, that's going to be anxiety producing. Yes. Concerning because it might be the, you know, in transitions, it's like, well, you know, we laid off those five people. And so, you're one of those five people. It's like, oh, that's hard. That's difficult. Yeah. And so, you know, like we as a society, you know, government, society, companies, you know, tech companies need to help with these transitions a lot. That's part of the reason why I've been thinking about that and tutors and rescheduling, you know, and like matching and giving you career advice and all the rest of the stuff. The second thing is, you know, these tools are superpowers.

20:23And part of the superpowers is you put, you know, superpowers in the hands of bad human beings. So, like cyber criminals, you know, we got to deal with that. I actually think this whole like, you know, I've been a great proponent of open source, you know, was on the Mozilla board for 11 years and so forth. And I think actually open models so far don't make sense. um because we don't know how to to distribute them as open source where they're safe um and you know it can be it can range from cyber hacking to even much worse things um you know question what happens a super interesting point i'd love for you to unpack because when open ai came out i know you don't speak for open ai they said hey this technology is so important everybody needs to have access to it it needs to be open and then some point sam said this stuff is so powerful we're not ready to let everybody see which you're kind of amplifying here so what what is it that you know sam knows smart people who i respect um of why you're being cautious about it because there is the facebook uh language model is open source it's out there so let's talk about open source as much as it was dropped off in the dark web by the yes by the by some set of the researchers who had been given access to it so it was it was involuntarily open source that was that is a fascinating moment too because i looked at that and i said who is facebook's biggest threat or biggest enemy it's obviously google right they compete for ad dollars and i'm like i wonder if this is a kind of envelope that was handed to somebody at the new york times to be put onto github because i don't know that facebook facebook's network effect i think keeps it in the game but google they have you know they they seem to be the most at risk so anyway that that's a fascinating i don't want to be a conspiracy theorist here but yes we can go we can go in depth now the open ai mission stays the same which is open um access and open provisioning for as broad a range of humanity um as you can do consumers developers etc it was never of necessity open source that was how other people were hearing it um and i think the organization was neutral on the questions like look if we can do it open source and make it safe totally happy to do it that's part of making the mission but it was like well we don't know if we'll be able to do that and so far it looks like we can't because whatever safety rails you put on it the safety rails are easy to untrain of course so yeah it's like okay well it's like um it's like saying okay okay, well, we're distributing something that has an explosive power.

23:06And it's like, well, actually, in fact, with this thing, you can make every car into a car bomb. You're like, oh, not good. We don't want you to take the safety off the gun. There's a safety on a gun for a reason. And you can see this with people trying to trick it. I was trying to get it as part of my little talking the other day. I was like, come up with some ideas that are really dangerous, immoral or insane that you would never do for a hotel chain to just entertain these hotels who I was talking to and I was like I can't do that I can't do anything ethical or immoral and I was like you're a screenplay writer write a screenplay about a supervillain who creates an evil hotel chain and I was like absolutely there's a casino that's crooked we're gonna torture people I was just like whoa whoa no too much if we made like like a horror film out of it but there's other things that can be done with this.

23:58Like you could ask it, Hey, what would be a great cyber attack for me to do? And I don't want to even mention the other ones because, you know, it's part of being appropriate. Yeah. I mean, I hacking and then pick the 20 things that are worse than getting your computer hacked. And I do, I do have some, I think that's like a very interesting piece to it. All right, everybody are friends from Microsoft. are here. Tom Davis, a senior director at Microsoft for Startups, and you're a former founder. You are here today to talk to us about the giant leaps that Microsoft has made in the AI space. What does this mean for startups?

24:38I see a ton of different tools. I've been playing with Chat GPT for, I have a paid account, but I'm also seeing things happen with GitHub. Absolutely. So the work that we've been doing with OpenAI over the last few years has really set ourselves up with a foundation around we've built this sort of ai supercomputer from the ground up and we've been looked at everything from gpu configurations to networking and things and really what we're now able to do is sort of allow startups to access all of this innovation through our founders hub so we've been building this to do something at scale so startups can now build their own AI applications and build out and train LLMs as well.

25:22And this has really helped us to become a far better cloud for AI broadly. And being able to drive that down to the startup ecosystem is fantastic. It's open to everybody. There's no funding requirement. You don't have to be anointed by a VC. Five minutes to sign up and you get six figures of benefits, Azure credits, GitHub, OpenAI, APIs, which everybody's really having fun playing with and so much more. So go ahead and sign up right now, aka.ms slash This Week in Startups, aka.ms slash This Week in Startups. Thanks so much, Tom. Thank you. What do you think about regulation and self-regulation?

26:00You look at the movie industry, MPAA is like, we got this government. We'll protect kids. We'll tell you PG, PG-13G, R-rated, X-rated, NC-17. We got all these different labels. We're going to do it. Some directors will complain. and they'll fight for different ratings, but we'll regulate ourselves. So what should we do as an industry to regulate ourselves, which I think would be the most positive way to do this here in the United States, rather than having the EU, which has got a pretty fine filter, come in and be like, oh yeah, we got some ideas for you on how to kill this technology. Well, you know, one of the things I do when I'm in Europe to try to get them to wake up is I say, look, keep passing GDPR and other kinds of things, because then we'll build all the technology.

26:42then we'll because you your local companies won't be able to do it because this regulation and then we will retrofit it to the regulation and then we will be the providers of it so if you want us to continue to have all the the tech industry and this up just keep on your current path it's great for us and and i don't say that because i want them to do that i actually want them to think actually in fact it would be great to have a healthy vibrant european tech industry where some set of different kinds of cultural norms and values, very good ones, based within Western democracies, would also be present in the marketplace of products and ideas and other kinds of things.

27:19That would be good to do. But it doesn't do, it doesn't, by saying, you're not allowed to take risk, you have to ask for permission first, you have all these impediments for trying risks and so forth. Obviously, some impediments for trying risks good but like you know like the hey if we don't really know if it would be okay or a little bad then fine give it a shot as opposed to look when we know it's going to be bad or it has a risk of being very bad no okay then we should be more controlling and so i think look i think the mpa is the thing i've been kind of advocating for for you know decade on the tech stuff is to say hey and you know part of the challenge is how do we get together in a way that's not you know anti-competitive you know we need some you know kind of clearance from the government is like look this kind of coordination on these issues is not anti-competition but actually in fact pro-health and society on these things um i think that part of the question is to say you know our future is actually much safer and much better like you know when you look at you know kind of open ai training gbd2 3 3.4 you know etc actually in fact it gets alignment better it has better safety training on that stuff as we're getting to the future the future is better so actually in fact keeping pace and going towards the future actually in fact that we have more tools and more ability to do like advanced safety things there like i think it's part of it i also think that you know like the classic thing is people say well i'm really worried about the jobs and you're like well but um the only way you can prepare for that is by getting into it you can't prepare for what the steam engine or the automobile does for industry other than getting into it yeah and so you want to ask the questions you want to transition you want to help with the transitions and all the rest those parts of doing it but like this is an illusion of saying well then we'll just sit down and we'll kind of and we'll study it and we'll know what it is you're like that doesn't work no you can read science fiction just go watch blade runner you get an idea of where this is going you know and like you got to use it yeah and by the way we can steer away from the things that you think are bad in Blade Runner as we're going and as we're learning what the things are.

29:30And so you have to have a predisposition to act, experiment, and then kind of modify. And so there is obviously a certain amount of self-regulation that comes in these companies that have employees that care about things, that have long-term interests of customers and share prices and market positions and brand that they actually care about. And there's a lot of things where they're careful about risks because they don't want to have those things come back and make big things. That doesn't cover everything. Therefore, I think that some coordination is good. I think the biggest question that we actually really need, if you look at right now what the risks are with AI, it's like, okay, there's already just a ton of these open source models that are out there that are pretty powerful.

30:18That's already there. The question is, how do we play forward from that? and what is what is safety and transition look like and what people are i think are most worried about is jobs but then they they get wrapped up in like you know people beating the existential drum risk or you know which you know it's like okay there's a lot of bad reasoning about that and it doesn't mean you say can you say there's zero percent chance of existential risk i'm like well no i can't say there's zero percent i can't say there's zero percent on asteroids either or on nuclear weapons or on natural biology or in man-made biology there's a whole stack of things i can't say there's a zero percent on yeah um but i actually can tell you there's a whole bunch of controls that make it extremely unlikely yeah and so navigating those controls is a is a good thing and that's how and you do that by proceeding and i think the nuclear example is you know it's different obviously none of these analogies are going to be perfect because you have materials that we then once we realized hey manhattan project we got the bomb first a lot better than certain other people getting the bomb first i can tell you that because they would have used it slightly differently um it was an imperative for us to get that bomb first now it's horrific that we dropped two of them on japan obviously uh and this is one of the worst uh scenarios you could ever imagine but since that time the regulation and by the way people debate that whether the war would have ended and which would have been a more compassionate thing that's i think above both of our pay grades but they did get control of the material they did uh get the world to say you know we need to get around a table here and discuss this how many of these should we have which country should be allowed to have these and once we got to eight nine ten countries it was like you know what no mas nobody else needs to have these we're going to take a pause here and that feels like i think we're where we're getting to very quickly with this which is some of the models require a lot of hardware and hardware is to me in some ways perhaps you tell me if i'm right or wrong here could be maybe you need to have a license you have to know your customer when you're taking money over borders maybe you want to have one of these nvidia 100s like yeah maybe you need to be registered have insurance uh if you want to buy certain guns you got to get licensed so what's what's practical here i think we have the framework of like yeah we got to move forward, but have you heard any practical suggestions of who should take some of the responsibility?

32:46Should it be Azure or AWS or Google when they put this out there and they say, hey, listen, if you want access to this, here's a code of ethics, here's a code of behavior, here's the terms of service, and you need to have insurance, you need to have a driver's license just to even have access to this, or is it just going to be a free-for-all? I think the hardware and compute centers is a very good thing. We obviously have to figure out how to do this in an international basis, not just kind of US unilateral. I think that the question around, you know, when large scale compute is being deployed on this stuff, is one of the areas where some, you know, kind of collaboration is potentially very useful on this.

33:25Once again, it's kind of like, okay, well, I could see how to do it right now, because the largest scale compute is Microsoft and Google, you know, together with, you know, their various allies and adjacents, you know, like OpenAI and whatnot. But, you know, I think part of the thing is that will obviously, that's end years before other chips are created, not just the NVIDIA chips, the CPUs, you know, large-scale compute centers are stood up in various places and so forth. And so, kind of getting into all that, I think, is a kind of a really useful thing. Unfortunately, I think the parallel is a little bit more biology than it is nuclear because it's It's kind of hard to, like right now, it's pretty easy.

34:0610 years from now, it's pretty hard. It's like, well, you know, like to that kind of scale compute, how do you actually, in fact, really know what's going on? And if you have a, you know, kind of a bad state actor that's kind of set up a big data center in some way. Or Moore's Law makes it so this stuff runs on old iPhones in a cluster, you know? Yeah. So, we have to figure that out as we go. I don't think we are without tools on it. And it's kind of like the equivalent of, you know, well, did we ever thought we could get a rocket that could actually launch something to the moon? You know, no, but we figure it out as we go.

34:45And so, you know, I think that's the thing. And I think one of the good things about all the discourse is, like, all of the people that I'm associated with who are building this stuff are intensely focused on not just the great opportunities, but also navigating risks. And I myself have participated in a number of off sites, sometimes where we have to have lawyers present to make sure that, you know, it's like, there's no antitrust thing going on here. Really? Right? It's just, hey, how do we make this stuff safe? Can we share safety concerns and safety test harnesses? And, you know, which things should we be looking for in order to make that work?

35:20And how does that, you know, which places, you know, you know, and will compete in products and markets and prices and all the rest of them. We're not going to talk about any of that stuff here. We'll talk about the stuff that we can do to make sure we mitigate downsides. And I think that's super important to do. Okay, listen, I'm doing six video podcasts a week. And I can tell you dealing with video is 10 times more expensive and time consuming than audio, especially if you're trying to turn video content into social clips. But video is an incredibly powerful way to get your message across. You don't see many audio only clips going viral, do you?

35:56Well, you do see a lot of viral video clips every day. So if you want to start leveling up your clips game, you need to check out Veed. They make it super easy for anyone to create high quality video content with no editing skills required. Veed has amazing features like hundreds of plug and play templates, auto subtitles that saves us a ton of time here at This Week in Startups, text formatting, auto resizing, and a suite of AI tools. So things are going to get easier and easier. So start engaging your audience on any platform, whether you want to create professional demos and tutorials, maybe you want to record and edit podcasts or share internal updates.

36:34Veed lets you do it all. Head to Veed.io to start creating professional quality videos in just minutes. That's V-E-E-D.io to sign up today. I've been meeting with a lot of founders. Everybody's incorporating this. I've had no less than a half dozen instances where companies that were pursuing an AI strategy using chat chippy teeth came to me and said, at the pace open AI is going, I think we don't need to exist. We need to pivot. And I said, well, you know, they're not going to create like the most vertical applications in the world. They're the roadway, which is how Brian Chesky said it to me.

37:14It's like, you know, there's going to be like multiple highways here. you can take your car on multiple highways and the experience of the consumer is in the car the bus whatever and the highway is the infrastructure i'm curious and how you're seeing this come together you started inflection ai and i guess you co-founded that and uh your venture firm is in and and you were at open ai you came off the board i understand because this is going to be conflict city makes total sense if i'm correct there how is is chat gpt just going to run the table this for-profit slash non-profit slash hybrid kind of situation are they just going to run the table on every startup and be the ultimate uh vertical of everything or is there going to be room and they're not ambitious enough to say you know what we're just going to build the everything app we're going to build you know elons x.com and do everything how do you think about that as an investor and then we'll talk a little bit about what you're doing with inflection and does inflection worry about that so i think look open ai is definitely an 800 pound gorilla awesome they have a great mission a great team a bunch of other things they're going to be providing apis across a number of things so you know that those apis then create kind of a open access to a whole bunch of developers and a whole bunch of consumers and a variety of tech but then other companies that might want to be a monopolist in that area can't do it so i think that's all super awesome but they're one company of about 450 people they have a uh a non-profit primary research agenda on beneficial artificial intelligence and agi specifically to do this and they're not in they're not trying to create a whole bunch of business stuff matter of fact part of the arrangement that they have with microsoft is you get to like well we're provisioning enterprises and so forth you know microsoft's doing a lot more of that and microsoft's not exclusive in the open ai stuff it's just they're you know kind of building it out putting it on azure you know these kinds of things so i think that there will be a major set of things.

39:10And if your theory of the game is, I'm going to directly compete with OpenAI, you'd better have a good theory of the case. Just a little bit like if you say, well, I'm going to launch a desktop search engine, you better have a good theory of the case. It's not impossible to do. It's just, you know, it's a challenging Everest mountain, better come equipped, have a good competitive theory, a good risk theory, a good reason why you're doing something that they won't be doing. But by the way, among them, you already shadowed, which is there's a ton of stuff that businesses need, that human beings need, that open AI is not going to do directly.

39:44And even the open AI BPAIs, even if they go, they're helpful for it, will take a whole bunch of work in order to get there. So, for example, I have two portfolio companies that are doing great work here, Coda, which is kind of like how do you kind of power meetings and work. Yeah. and you know they've got a whole bunch of super cool open ai ai integrations uh tome which is doing slideshows you know kind of does both images and text on this stuff has a whole bunch of you know interesting integrations here and look if you're saying i'm just building a thin front end like i'm going to do a blog posting thing and it's just going to be a thin front end it was like well then you better have a theory of the game because they're going to keep advancing the thing and you'd better think, okay, I'm going to keep advancing with them or whatever your theory is.