Will OpenAI Kill All Startups?

This is Michael Seibel with Dalton Caldwell, and today we're going to talk about how OpenAI is going to kill all startups. This is our last video; might as well pack it in, we're done. OpenAI is going to do this. They're going to make the videos—next video, next video—doing our video for us and do a better job than us.

Yes, what do they call it? It's not a deep fake; it's like a complete AI recreation. Yes, yes, we're done. Well, it's been fun, so let's talk about how that happened. Right, so I think to start, people need to understand that companies like OpenAI, Anthropic, etc., they're actually trying to build AGI. They're not trying to build the AI-powered CRM or better search; they're trying to build AGI.

And this is not a debate; we're not going to talk about whether they're going to succeed at building AGI. Yes, there are plenty of videos and plenty of experts that are better experts than us to debate whether AGI is about to be created. Yes, I kind of bring this up—Godwin's Law debate—one of those people arguing with each other, and one person calls the other person a Nazi; the debate is over. No good discussion is productive; it just means it's over.

And so where I'm going with this is, once we start talking about how AI could become God, and we have Skynet, and like take all the things and when we upload to the Matrix, I'm arguing nothing useful. Like, that conversation is over. We can't help you in this video; we don't know, you know, whether or not Skynet is coming. It's outside the scope of this video, and we would love to talk about it; we just don't have the expertise.

Yeah, they should listen. Yeah, someone else is.

What can we talk about? I think it's this: there's a lot of history to learn from about when major technological changes have come out. Yeah, and to the extent this is a major technological change as opposed to, "We’re all going to get uploaded to The Matrix," or whatever. Yeah, there's a lot to learn from history.

Yes, and so this is where when founders do ask us advice on what they should be thinking about with AI or how will AI affect their startups or what is OpenAI going to do to kill other startups, we have a lot to say about that, or we have a lot of suggestions to look from history.

So why don't you list some of the trends? So, I mean, when you look at Major Innovations—talk about farming or modern farming, you talk about electricity, talk about the internet—I think what's been so interesting as a trend is how many businesses it's enabled. Right? Like, the number of businesses increased, and then the second trend is how startups were relatively advantaged versus incumbents. Like, the greater the technology change and the shorter the period of time, the more the startups are advantaged.

The whole history of the internet is amazing; there's a whole industry around starting about funding startups because some innovations happen so quickly and so many companies are disrupted so quickly. So there is clear precedence to major new technology trends creating opportunities for startups. If history repeats, as it tends to, somebody who's smart is looking at these tools today, saying, "Holy crap, real problems can be solved right now that couldn't be solved six to twelve months ago."

I think it's interesting because you're seeing this in a lot of conversations with founders right now. Right? Like, let's talk about who is choosing to start a company in this phase and what we can learn from them.

Yeah, a couple of things. So one, there's a difference between cargo culting AI and actually using AI to build better features. Yes, and so let's talk about the difference. Cargo culting AI is to say we have AI, and it's like tangential to what you're doing. You're just saying it to raise money; it doesn't help your customers; it doesn't improve your product, whatever.

Yes, but we are seeing people add AI features that dramatically increase retention, that dramatically increase the quality of the product, that make it much easier to charge for it. And that stuff is real, and that is not hype. That's not bullshit—no, that's real.

Yes, this is like when people first started launching apps when the App Store came out. Yeah, apps were actually good; yes, there was a hype cycle around apps, and a lot of people were cargo culting apps. Yeah, but building a high-quality mobile app that increased the retention of your users? Incredible. Facebook almost died because they didn't do it fast enough.

Exactly. Yeah, I think if you look at—I don't know—we saw from our generation's founders, we saw open-source and cloud computing come out. And think about cloud computing—yeah, if someone was like, "Oh, this cloud computing thing's hype," or "We're not going to add cloud computing because it's just a thing VCs want to hear."

What I'm saying is, like, there were sometimes people—you remember these people—they were like haters on cloud computing because it was overly hyped. Yeah, but that's missing the plot. Just because it was overly hyped didn't mean you shouldn't put on your thinking cap. Well, it didn't mean you should be like, "Hey, should we be running some of this stuff from the cloud instead of running around servers?"

Yeah, well, and I think that there's a big difference, and we've now seen both. Right? I think that we've seen the kind of fintech boom and the crypto boom where the people attracted to them were slightly different. Yeah, and as tools, they just weren't as useful.

Whereas, you know, like giving everyone a new bank—like for every single flavor—whereas like we also saw, literally during our lifetimes, launching a website with like closed-source expensive tools would cost five million bucks, and then with the similar tools that were better and open-source would cost like fifty. We built our companies on MySQL, yes, like Postgres—but yeah, well, Ruby, Postgres—regardless, like that shift happened.

And then I remember when, like, you had to buy servers and then wait for them to come and then configure them and then have to get more space in the colo. Like, the speed of actually building software was incredible. Like, we're going to keep buying servers—like some people rejected that.

Yeah, and so there was a moment in time where it didn’t make sense to jump on the bandwagon—and we were on the bandwagon, but it was actually smart. It wasn't just bullshit hype; it wasn't bullshit; it was real.

And so I think it's obvious that now AI is clearly going to be a tool as impactful as mobile or open-source back-ends or cloud computing, and then we'll see from there. Right? Like LLMs are clearly that useful; and so it's stupid to not think about how you can make your users happier, more productive.

And again, I think what's funny about this is you and I were in the room when OpenAI was created because we were working at YC with Sam, and we had YC Research. Yeah, and it was a non-profit—it was meant to be an enabling technology.

Exactly; the vision here—you were in the room—it was to be a non-profit so that all these startup flowers could bloom. Yes, and all these people, you know, could be enabled to create value.

And so again, I think sometimes people are afraid of a narrative that OpenAI—maybe that will happen—but like the vision was about creating the technology. Yes, it'll be harnessed by other people, and ideally, creating an AGI was kind of what his goal was.

Yeah, as opposed to, "I want to create AI-powered CRMs, and I'm going to destroy every startup this degree I've ever known." That was not what we understood, right?

No, no, no, I think what's interesting as well when you think about this game is that, like, there are very, very well-paid people—and you're pointing this out—very well-paid people who are leaving their jobs and who are starting startups right now because they believe that LLMs are a powerful tool.

Yeah, the way I think about this is opportunity cost and life. Yeah, there are people that I've been funding, and they have a great job, and they actually like their job. Yeah, they're choosing to leave because they know at this moment in history, because they become domain experts at AI, yeah, this is the moment to do a startup. Like, this is that!

Yes, imagine if you worked on, I don't know, cloud computing for years, and then all of a sudden, Amazon Web Services comes out; you'd be like, "Oh, this is my moment!" Yes, yes!

And so it's really cool to see these people with all this vast domain expertise that they did before the stuff was cool realize that this is the moment in time to work on this stuff. Well, and I think that we see a second group of people too, and I feel like those are the people who got the first iPhone when the App Store came out and said, "I'm going to build an app."

Yeah, and it was weird because at that moment, everyone had zero years of experience. That's exactly—remember the Brex founders? What they did when they were like fifteen was jailbreak iPhones. Remember all the kids that were doing jailbreaking? Like, this was like a rite of passage to be into this stuff and learn about it because no one knew anything about it.

So I think there's like two sets of smart people who are really attracted to this, right? Really, really smart people who love to see us, who are like, "This is an amazing tool, and I'm going to learn as much about it as possible because most people don't know anything about it."

I can get on the ground floor, and then people who have been doing ML for a couple of years who were like, "Oh my God, all of the things I imagined were going to happen are starting to happen. I can predict the future; it's a perfect time to do a startup."

I think what's tricky, though, is that there are people who don't think that this is a good time—the people who are really into CS don’t really, aren’t really excited by the power of LLMs. And people who don't have experience with ML at all, and those are the people who just want to make money.

And I think what's interesting is, in the last couple of cycles, those people have been very attracted to the startup world—"How do I make money fast?" Yeah, those people are not being attracted to this.

I think that's what makes our job so much fun because, like, working with people who want to make startups fast—might make money fast—and startups, this is—it's not a great group. Those are harder office hours, harder, harder when we're like, "Yeah, it'll just take a decade."

They’re like, "No, no, why don't I do my ICO or something?"

Whereas, like this group of people, I mean, it reminds me when people were nerding out about Ruby back in the day. Yeah, like "Oh wow, like I'm just excited about what I can build!" And when you think about it, this is a perfect time where if you're a creative person and you're an ambitious person, yeah, you can do really cool stuff. You know, we saw this with the iPhone.

Yes, well again, Uber would not have made sense without the iPhone. And so to me, it's less obvious that everyone’s going to build OpenAI competitors. I am worried that all the low-hanging fruit ideas OpenAI will just be able to do well enough.

Yes, but if you go deeper, yeah, to second-level insights—I'll use Uber as a consequence of the iPhone. Yep, second-order effects like this from LLMs are going to be amazing. You know what I mean? It's like, it wasn't obvious to try an analogy on you. Right? Like fart apps—remember everyone was making fart apps? That wasn't the big business.

I would argue maybe the analogy here was that it was obvious that maps were going to come to the phone. Yeah, it wasn't obvious that Uber was going to come. Right? And so it was like, never would I have imagined.

And so I definitely think that if you can think second order— and I'll even be charitable. I think sometimes to think second order, you think first order. It's okay if your thing starts as a toy. Like when people are like, "Oh, this is a thin wrapper on OpenAI." I always laugh because I'm like, "Okay, is that the destination or the starting point?"

Yeah, if it's a starting point, it reminds me of the old Dropbox as a thin wrapper on top of AWS on top of S3. Right? "I can do this in a weekend." I've always been a weekend—it's a sure sign that the person doesn't know what the fuck they’re talking about.

So I will say that, like, you know, if you are absorbing—if your thing is a thing wrapped—then wrap it on OpenAI to start, and you're absorbing hate? Ignore those fucks—yeah, like that's silly.

Like understand that many good things started like toys and like make something the customers really love. Yeah, this is a time to be optimistic. This is a time to build cool stuff. Yeah, and this is not a time to cargo cult, as I said earlier.

So superficially copying AI stuff just because it's hot or raising money? Don't do that. No, but if you can really solve customer problems in an amazing way, yeah, this is the time.

Yeah, and I think the last point we'll give is that I think that until these big companies solve AGI, it's going to be their first, second, and third goal. And so using the existing tools to make people's lives better, their businesses better, it's not on their list.

Like getting AGI is one, two, and three most important goals. So that's a huge opportunity for everyone.

So look, in conclusion, unless you actually believe in the short term that AGI will be created by OpenAI that will do all the things, that's possible for us to predict. Yes, OpenAI is not going to kill all startups. And there's plenty of room to innovate!

And maybe we would say that the tools that exist now and the LLMs that are being made available—this could be an explosion of new amazing startups. This could be the moment that we've been waiting for since mobile, where people can actually do cool new things and they aren't just culting.

So, there should be a moment where everyone's excited, except for the people on Twitter that are very exciting.

Alright, thanks so much, Dalton!

Thanks!

[Music]

[Music]