Reacts: ChatGPT - A Billion Dollar Company with Only 3 Employees? w/ Jeremiah Owyang & Ben Parr

Ben Parr: I believe that there will be a billion dollar company built in the next five-ish years that has one to three people tops because you can automate almost everything else. It's going to happen.

Yaniv: The Startup Podcast is excited to partner with Until Now, an incredible product and brand studio with proven experience in developing successful businesses and brands, including the Iconic, Airtasker, Karma, Spriggy, and Path Zero. We are putting our money where our mouths are on this one. Until Now even redesigned our brand. We think it looks fantastic and we're beyond impressed with the process and people. Whether you're a startup scale up or corporate venture, Until Now runs a cross-functional approach to solving your product, brand and go-to-market challenges. Head to their website for more examples and to get in touch.

Chris: Hey, I'm Chris

Jeremiah: Hey, I'm Jeremiah.

Ben Parr: Hello, I'm Ben.

Chris: And that's right. Yaniv is away again. This episode, he's paying me back for all the days I was away. He's on vacation. And instead we have two amazing old friends of mine joining us today, Ben Par and Jeremiah Owyang to discuss generative AI and the future of artificial intelligence in terms of its impact on tech, startups, culture, and the future of work.

Ben, Jeremiah, you guys, as I said, are incredible people, old friends of mine. We've shared some amazing stories, some of which we can talk about, some of which we can't talk about. how about you tell the audience just a little bit about yourselves, before we jump into the topic.

Jeremiah: Sure. I'm Jeremiah Owyang. I am in my Airstream office in my backyard in Silicon Valley, and I'm focused full-time on the AI market. Before that, I've been involved in other trends like Web two before that Web one, the collaborative economy and Web three. I was a CMO by Andreessen back company, and now I'm full-time exploring the AI market, and I'm trying to figure out exactly, what is gonna happen in the space.

I'm also an investor in Octane.ai.

Chris: Jeremiah, you are always very humble, so I wanna let an audience know you are one of the best analysts, that I know. You, have the ability to take a very complex subject and, you've done this many, many times and boil it down to incredibly simple, delightful, and even visual, landscape analysis and impact studies and so on.

And so it's really great to have you on the show, and it's been awesome collaborating with you over the years.

Jeremiah: Thank you. That's very kind.

Chris: And speaking of Octane ai, you mentioned you're an investor in them. Ben, you are the founder and CEO of Octane. tell us about the rest of your career.

Ben Parr: So, back in the day, I became the editor of Mashable at 2324, some ridiculously young age. And, it was because I was a part-time writer for them when I was in Chicago. I wrote 2,446 articles in my four years at Mashable.

 Mashable did everything from write a besting book for Harper Collins on the science and psychology of Attention Co captivology. I did a column for CBS cnet. I joined the board of directors of samma, the AI data training company. and I co-founded Octane.ai. And so I am, president and co-founder with my co-founder, CEO Matt Slick, we are AI for the world of e-commerce. We've been doing it since 2016, though. And so, we've been in this space for a long time and I've been waiting for this moment in the AI world for a very long time. our company, we power AI data collection and, AI analysis tools. So our best known product is our quiz product.

So thousands of e-commerce brands use our quiz product to learn about their customers, ask like what kind of skin they're looking for, what kind of routine they need, what kind of, fashion style and uses generative AI to make the product recommendations. And then we have a suite of new generative AI tools that are coming out in the coming weeks for our customers that we've been testing internally.

I'm also, write a newsletter called the AI Analyst. It's been part subs.com. Easy to find where I talk about AI trends in that. And most recently, I have become the AI columnist for the information. And so I'm writing about AI trends once a month or so for that publication.

Chris: Ben, you've been doing this AI thing before it was mainstream Cool. So, very cool to have you on the show as well. as I said, we've been great friends and done some amazing adventures together So boys, It feels like AI has been coming since the birth of computers, right?

 Touring dreamed of a machine that can mimic human intelligence. and so in a way it's been coming forever also feels like it's come all at once. what was your first impression when you first saw chat g p t and you saw what it could do and how people reacted to it? Because for me, it felt like prior to that point, AI was this pointless buzzword that people used that was more or less useless in more or less use cases because you just needed lots of training data and at best maybe you could make some recommendations about a product catalog on Amazon or something.

And after that moment it was like a watershed. Change in my thinking and what I think of as the capabilities of ai. So for each of you, was your initial knee-jerk gut reactions?

Jeremiah: Well, I was still working in, web three space as a cmo, and I saw that and I go, oh, it's on, that was the pivotal moment like when you first put your hands on an iPhone? When I first was able to tweet and blog and then my first. Modem, dial up modem, like all of those sequence of events, like it's a, thing you remember.

And you know, the, stupid thing about when you try new technologies, like the first time you use Google Maps, you look at your house, of course I looked up how accurate am I, my bio and then that of my friends. And it was a joke because a lot of it was wrong, right? Hallucinating. But you could see the power in it once you learned to use some of the prompts.

But I'll bet Ben Parr wasn't surprised.

Ben Parr: No, first of all, Chris, said mimic UBA intelligence. I, think the goal in what's gonna happen is surpass. Intelligence, this is inevitable to me. The only question is timing, and how exactly it happens. Now, for context, octane and especially my co-founder Matt, was head of product use stream and is like our product engineering mind at Octane, we had first access to G P T three back in 2020, and we were playing with it.

And Matt personally built the first product we ever did on, open AI and generative ai, which was something we called Octane.ai. And the best way to describe it was like a, Jasper for e-commerce. And it was more experimental. but we got thousands of people to sign up for it.

That gave us an idea of like the power of generative ai, because we could see like, oh, this thing could write entire simple essays and things like that. What ChatGPT did was provide the user interface that made it clear that. This is not just a novelty for a small group of people, it's something Everwood could use.

And then GPT four made it clear that it can do what a heidt could do in terms of speaking voice. And, the result has been just a whirlwind. One of my first things it did after Chat GPT came out was I recorded a TikTok about how AI was gonna impact education. If it experimented with some TOS.

That one got a million and a half views and I got thousands of teachers sending me dms and I was like, oh crap, this is real. And so there was an old crap moment was when that video went viral the reality is like everyone thinks it's overnight success, but the technology underlying, it's been being built for years, but really was the user interface that unlocked it.

And now everyone's seeing all the possibilities. and it does feel like a very different world. From when we were at December, I think just purely people believing now, like, oh, the technology is here, it's ready. It's that strong, changes everyone's perspective from investors to the public.

Jeremiah: It feels real. Like I don't hear people saying, that's not gonna happen. Instead, people are saying, we're all gonna lose our jobs. Humanity's gonna die. So, I mean, there's no pushback that it's not real and that's different than from other technology trends that we've seen. Like web three, there was a lot of pushback.

The sharing economy, there's a lot of pushback. Social media, there's a lot of pushback. This time it's different. People see that it's real and they're concerned about the output.

Chris: So the pushback is like, holy crap, this could disrupt everything. This is, this is too real. right.

Jeremiah: that's right.

It's too fast.

Ben Parr: You know, I love my friends at Web three, but Web three has always had, some really core problems. The most core problem of them all though, has been the user interface and like the difficulty of setting up a wall, that the difficulty of getting started, importing money in.

And that's never been truly solved just because of the nature of Web three. You know, there's other issues, but that's the big one. chat is the most natural user interface that you can possibly have. Like chat is more natural than a website. my co-founder wrote a home blog post years ago, 2016, how all websites would eventually be chatbots.

And it's taken a couple years, but I do think actually we're gonna getting to that point in the next couple years, why have a website when you can just have a back and forth and have access to the entirety of human knowledge.

Chris: So I think the discussion of Web 3.0 crypto versus AI is, probably something worth going into in another day. I have a lot of thoughts about the efficacy and usefulness of, crypto and, blockchain, but, on the point about chat being the most natural interface, I agree with you to a point. You know, when, Apple first released the AirPods, I predicted that voice would be the next user interface, And if you think about her or you think about the Star Trek computer or what have you, that seems like a really incredible application of chat, meaning natural language input via voice.

In many cases, I think writing documents, designing diagrams, decks, collaborating in various ways. I actually think chat is a, shitty way to interact with, most products and workflows because actually very low bandwidth and requires you to type a lot of characters instead of just clicking or interacting.

You are right though, Ben, that the key difference between GPT and ChatGPT is obviously the chat part. It's the part that made it accessible and visible and visceral for everyday people and created that Exponential. It's the fastest growing product in human history, right? So this was a real game changer.

 This is the Startup podcast, right? So let's talk about the impact of ai, generative AI specifically on. Startups and founders.

What would you do as a founder of one of these startups in order to respond to this existential opportunity or threat?

Ben Parr: I wrote my first column for the information, a couple weeks ago on this exact topic. if anyone wants to go, Luca, Fred, it's called Don't build the wrong kind of AI business. there's a couple pieces to think about here. One, there's an opportunity to build real businesses with, lower cost if you use AI correctly I don't believe anymore.

First of all, like I just need a technical co-founder you don't anymore. You could use ChatGPT and some of the new AI tools, rep others to build your prototypes and to build working products really quickly. I was able to build complete games and I haven't programmed properly in 10 years.

I just don't believe it as an excuse anymore when, programming itself is becoming democratized. Now in terms of the opportunity, there's a lot of companies and a lot of ideas that'll be subsumed by big players. That's always true from meta to Openai.

There are two things that can make you really defensible. one is proprietary dataset, having your own dataset to fine tune existing models versus trying to create your own, makes for really powerful, unique, outputs that are not possible from an opening eye or others.

And so, like for Octane, for example, we have hundreds of billions, billions of unique data points from our quiz product on customer preferences and shopping and all sorts of e-commerce data points that makes our outputs really unique to the world of e-commerce. The second thing I said in the article was, vertically focused technologies will do better.

There was a company that's doing generative AI for personal injury attorneys, very specific, and they just had one of the most competitive rounds ever. I think went from like one and a half to 9 million in ARR or something like that. So like more the tripled revenue. And they had a bidding war benchmark, one at a 350 billion valuation, like, the old days of like 30, 40 x multiples.

Because they focus their AI in a specific vertical made it, hard for others to come in. opening ads not gonna build that. Med is not gonna go and build that. And so I would say focus again. And it's the same advice we've been giving founders forever. Focus on a very specific vertical if you can.

There are limited exceptions, but if you focus on a very specific problem, a very specific vertical, a very specific thing, and you build a data set to make yourself defensible, you could leverage these tools and build a huge company. with raising very little or no capital at all. because of generative ai, there are real opportunities.

There are real, waters to navigate. I try to do it as leanly as possible because investors are still trying to figure out, what companies to back and not to back. And while that's happening, if you could just build a profitable company or a company that doesn't need much capital, you will be in incredible shape as things change in the next few years.

Jeremiah: Great points. Ben. I've been attending many startup events in San Francisco, which has turned out no surprise to be the epicenter of the AI movement. I'll just share with you a few things that I've been seeing.

Last night I was at the YC event and met many of the startups, many of them who are in AI. But you know what? When they gave me their pitch, in most cases, they started off with what's the actual problem they're trying to solve? I met some startups at another event at the Amazon Web Services, who, by the way, are offering ML and AI.

Cloud services for startups. So you don't have to be an AI expert with their, growth product in the new bedrock product, which is, they have their own foundational, ai and to the Year Point Bend, or one startup, and she was focused on lawyers that work on m and a and all that they would do is ingest the legal documents and look for errors and patterns.

And then the, AI would solve those issues. But she was focused on the business problem. So there's a couple things that I'm looking for, for startups that are gonna stand the test of time here. when they're using ai, some of the rules have changed. Some of them are old. The first one is having that proprietary data set.

I am worried about any startup that is connected or dependent upon GPT. For example, a few weeks ago they announced the plugins. About half of the AI startups in YC had to pivot. When that announcement came out, they were not prepared. They were dependent. Chris, Ben, we saw this before with Facebook and Facebook apps and Twitter and Twitter apps, right?

Yeah. You get nodding your heads. So that happened. Again, these guys were young, they're Gen Z, this is like brand new to them. They weren't quite aware of these patterns. the second thing is creating your own unique dataset. I met a number of startups even last night that they were creating their own database of prompts for a very specific market, a very specific, role.

It happened to be, a very technical role. And so they had to create very specific prompts and that cannot be replicated cuz they're creating it news. And that's number two. A startup, if they're able to have a network effect where they're taking the, generative output or the prompts from one user, learning from it and then giving it to the next user.

You have a network effect, a marketplace, something Chris knows very well having worked at Uber. And then the fourth one is using autonomous systems for it to self-learn. And the fifth one, if you happen to have computing credits because you're tied to some of these companies, that's a big operational scale.

By the way, I'm quoting a lot of the learnings from the Blitzscaling methodology, which Reid Hoffman and Chris Ye put together. They've been teaching me a lot about this stuff. I've been listening to this stuff cuz it's still very much applies to this market. So those are the things that I'm looking for and seeing in the startup space.

Chris: So for me, I think there is still a big question around what Ben said around building niche tools. with AI specifically whether niche is gonna win. Or there's just gonna be a few broad generalized agents that can help you with everything. Like Google Assistant and Siri can plug you into everything else.

And with a few plugins and a few connectors and a few whatevers, maybe Google and, ChatGPT take the world. I think that's, a non-zero chance that this creates additional concentration of power in big tech, or even big players with proprietary data.

We've all mentioned proprietary data, but let's assume for a moment there's an opportunity for startups to operate and navigate and succeed. And it ends up being a niche outcome instead of a concentration of power at a big tech outcome. if you are running in an existing startup. Then you want to hope that you're solving a real problem, right?

As I touched on very early in the episode, you know, a lot of Web three projects are just not solving real problems. It's, kind of tech looking for a problem. But if you found a real painful problem and you still passionately believe in that problem, then the question has to be, how can AI accelerate the solution for that problem?

How do I add AI as an ingredient or rethink my product strategy with AI first or AI native to solve this problem in a better, cheaper, faster, more delightful way? That's the question you have to ask yourself in a post chat beauty world.

Now, if you are sitting on the sidelines and you've been thinking about starting a startup, I think the questions you have to ask yourself are a little bit different, the first thing I'd say is now is the time to start there is. With every new platform, a new reshuffling of the deck, new winners and losers, a cambri and explosion of innovation and excitement, now's the time to jump in. You have these moments, maybe one, every 15 years, 20 years, . If you're lucky, this may be your one.

And so get off the sidelines and jump in. And as we've talked about on the Startup podcast, many times, big disruption occurs at the advent of new technology, new regulations, and new cultural norms, So Uber was born on the back of the new technology, iPhone, and new cultural norms of like meeting people off the internet,

Jeremiah: And there was an economic downturn, right? People didn't wanna own cars as much as well.

Chris: That's right. it was fueled by an economic downturn, where people are looking for side hustles and so on. And so this is the exact same circumstance, right? There is an economic downturn, there is new technology with chat G P T, and there are new cultural norms emerging where people are starting to trust, Intelligence that is not human.

This is a, a perfect time for new innovation. And also when you're thinking about new startups, think about. AI native solutions. So if you are old enough to remember pre-phone, there was a photo sharing app called Flicker. It was the Web 2.0, one of the first Ajay, web apps that felt like an interactive surface.

And that was like the, all the rage, at least in Silicon Valley it was. but when iPhone came out, flicker didn't make the leap because it was not iPhone native, it was not mobile native. And instead, Instagram came out and blew by Flicker because it treated the phone as the first-class citizen. And so, if you're thinking about a new startup, or in fact you're young enough to pivot, think about what is the AI native solution to this?

Is it just a ChatGPT, plugin as Jeremiah mentioned? Is it something that rethinks the UX or the problem from scratch in some way? And so that's a big opportunity that I think incumbents will be less likely to do because they've got legacy investments in their old tools, user experiences, integrations, and so on. the last thing I'll say with this big change is surprise, surprise. I've been saying this since the beginning of the downturn. the FOMO back. these things happen in cycles and while investors are still kind of sorta waking up and still kind of sorta understanding how to invest in this space, the FOMO is well and truly on its way back.

You will be able to raise capital and you will be able to think about health evaluations if you are doing something that's on trend. Solving a real problem with good momentum, with good founders and good storytelling. I think we're heading back into a place where investors understand what's going on and, the chicken little is slowly evaporating from the market again.

Ben Parr: It is definitely one of those keystone kind of moments and I think I really would double triple down on AI native being really important. there's just so many examples of companies, not being able to make the transition to the next platform and companies built for the next platform being really successful.

And there are examples of companies who successfully made the transition because they treated like Facebook, treated the phone, and they switched to being mobile native first, and they succeeded at doing that and moving into the next era. So it's entirely possible, but you have to switch your mindset.

Now, if you have an existing company to being AI native, what would an AI native company do? What are the technologies you would be using? what would those interfaces look like? the reality is that, you can implement AI into your product so easily. It is so easy to work with, GPT4, API and like get started with that.

Not to mention once you've figured out how the basics of some of the models work, like implementing some of your own, if you want to, we're having a combination of both AI native, AI native, AI native. You're spot on, Chris.

Chris: This is hard for people to do, right? to pivot from the old platform thinking to the new platform thinking. For the listeners out there, you wanna let go of domain dogma, things that were historically true in your domain.

You wanna let go of cognitive inertia where you've been thinking down a certain line, a certain hypothesis, certain UX approach. You wanna let go of all of that. And you want to go back to first principles thinking you need to let go of legacy thinking and start from pretty much a blank piece of paper.

That's the only way to do this, and do it successfully.

Jeremiah: There's three types of AI companies that I've observed in this space. so you as an entrepreneur, need to think about which position you wanna play here. And they each have their upsides and downsides. One is an AI model or AI company.

You're actually training and building the AI model. And by the way, we will see versions of that in each large company. And you might build it for like the Coke, the Pepsi. We already know that Bloomberg built a version of that because they own their data, which is gold. So there are opportunities just having segmented data to do that.

The second type of company is an AI application company, and so these are sitting on top of those. They're connected through APIs, plugins, and eventually we'll see more SDKs as well. and there's risks there because you might be dependent upon the AI models that I just, mentioned.

And then the third type of company, which is picks and shovels, is the AI infrastructure. You're selling tools, analytics, data optimization. I was just at a, very big, ml, commons meetup. And the big thing that they're focusing on is there's not enough compute power. So if you can even build software to solve that.

So those are the three types of companies you need to figure out which role we're gonna play. They each have their different risk levels as well.

Chris: As usual, Jeremiah, you do an incredible job of summarizing the whole landscape, elegantly so thank you for that.

So let's talk about the impact of this, not on founders in a vacuum, but on the startup ecosystem more broadly. particularly, I'm interested in San Francisco, New York.

That's where you guys spent a lot of time. we've heard proclamations that big cities are dead. Covid is eviscerated the need for you to be in a physical office with, colleagues. and these tech hubs are dead or dying. There's homeless everywhere, violence in the streets. So what is the impact on these tech hubs?

Are these AI companies being built remotely? Is still worthwhile to move to San Francisco and New York or another tech hub? We just had an episode with Jack Bloomfield talking about him moving from Australia to the US and the benefits that he felt from that. What are you guys seeing as residents of these cities?

Jeremiah: Sure, So I'm in the SF area and , is the epicenter for AI right now. we've had a really hard time, San Francisco has the highest vacancy rate of commercial. Real estate right now. It's extremely high and I can see it happening. And even small businesses are suffering.

The problems that you mentioned with homeless and people that really need help it's still an issue and I don't know how it's going to get solved. violence on the street, anti-Asian hate, lootings, robberies, happening to just the CVS and target, all those things are still happening, but at the same time, the startup ecosystem has bloomed again, in person in particular, there's two locations that it's happening at.

The first one is Hayes Valley, which has now been rebranded as Cerebral Valley, and that's right next to, it's not in Selma or Mission, it's just north of that. It's near the Civic Center off NS and a number of startups. hacker houses have emerged in that location and they're all hanging out there.

Secondly, shack 15, which has been a co-working spot that launched right before the pandemic, which was focused on data science that's now become like the hangout where there's hackathons happening. and as you already know, open AI is from San Francisco, and so these things are just, Booming right now the San Francisco AI scene, there's three to five events every single day.

Even the weekends, the hackathons are on the weekends. I've been to almost five events and it's only Thursday. because there's just so much happening and people are flying in, the investors are milling about  I haven't felt this excitement since web two, but it's happening in a more concentrated space and a tighter time period.

I have not felt this, and I've been here since.com. I've been through many cycles, so this has been amazing to see.

Ben Parr: Look first. It's possible to build a big company wherever you live. I still firmly believe that. And you know, you don't need to go to events every day to build a big company. And in fact, it can be detrimental, for specifically founder trying to build large tech companies because you gotta focus on product.

You gotta put your notes to the grindstone With that set, the narrative that you don't need it. Like, there's no advantage of being in a bigger city is wrong. And we're seeing it right now. You're right Jeremiah. Like, there is a lot of energy back in San Francisco in the Bay Area. And if you're there then serendipity happens.

And look, it's especially true for the newer generation of founders who don't have those connections and need that like support network and like the kind of thing that the three of us had, I'm gonna call it a generation ago cuz it's true. That with the early connections and like early buzz and like building those relationships because some of those people will become VCs and invested you and some of those people will become top executives at companies and some of those will become billionaires.

And those relationships are built by in-person pieces. And so like, look, I'm in LA and I love LA and I could afford to be at LA from the perspective of like, there's a lot of tech, but also it's an easy flight over and I've already built my network over the course of years. And so, I'm hopping up, once every two weeks-ish, to go and see some of the events, talk to some of the founders, talk to some of the investors, go to the conferences.

I'm speaking at a bunch of these conferences. I have like a whirlwind trip at the end of May where I'm in. Dallas, Miami, DC and New York because there's different generative AI events and, things happening in those cities as well. So it isn't just a San Francisco phenomenon, but I would say if you are a new founder, you're a mid early twenties and you're trying to build something.

There is nothing like being in a city like San Francisco, building those relationships, having that like back and forth, the VCs, and with other founders who do eventually become VCs of the people that you work with. There's nothing like it. You can't replicate that, virtually.

Yaniv: I'm with Francesco, founder and director at Until Now, and a former colleague of mine. Fra, it's great to see you again.

Fra: Hi,

Yaniv: Tell us a bit more about Until Now.

Fra: We start until now with a believe that a senior experience team can increase the chances of success for a funder. Usually we work with early stages startups and scale ups, but we also work with copper ventures before. Our goal is to accelerate a journey from AEA to product market fit, and sometimes we are going beyond that.

Yaniv: Nowra. Every agency reckons they're different. Tell us what's so special about Until Now.

Fra: Our teams come from places like the iconic Tasker, Deloitte, and publicist. So there is a really valuable mix of, in our knowledge and agency expertise. We also combine skill sets on every project. For example, when we design a brand, we already consider how the dent will live in the final experience. Or when we work on a product, we consider what is technically feasible as well as go-to-market strategies.

We also have different engagement styles. We sometimes augment the existing teams and other time we work as a virtual cross-functional team. Hopefully univ, you got to experience a bit of this when we did a rebrand for a startup podcast.

Yaniv: Head to their website for more examples and to get in touch.

Chris: So we just talked about the impact of AI on founders who are running, startups. How should they think about AI as an ingredient to their products? And we've talked about how AI is affecting the startup ecosystem in San Francisco and around the world. so let's now talk about how does AI affect.

Culture, work and civilization. What are we excited about and what are we about in terms of the impacts on our livelihoods and on our culture?

My initial reaction to this was really, this is going to change everything. this is almost too real. it's going to infiltrate and disrupt every part of work and civilization in some way. the same way that every platform before it pretty much is done, you could think about the personal computer, the internet, the iPhone, or smartphone, social media, and now ai. These were fundamental breakthrough shifts in platform technology that then reshuffle the deck in terms of the winners and losers in tech and the winners and losers in terms of products, and creates this huge cambri and explosion of opportunity, innovation and change for the world.

And I think one of the other things it does, you know, we used to talk about content as king. Well, now content is free. and so what happens next? Is it. back to the question of attention is king, or maybe Providence is king, or authenticity is king. Jeremiah, you're saying data in order to train these models.

But I think if you're a company, then proprietary data is king. But as an end user experience problem, I think Providence, and authenticity starts to become a real concern. Like, is this made by human or is this made by computer? Is this a real photo of Trump being arrested or is this a made up photo?

Is this a real thing that Biden said or is this made up? that to me seems like the democracy breaking question of the time.

Ben Parr: A couple thoughts here. one, brand, becomes a much higher priority. And so look, if you're a content creator, like a toker with millions of views, AI is not going to replicate your style, your tone of voice, your videos. And so people are going to trust those, people more.

 AI's still not good at doing video. And so personal brand increases in value, and audience that you have increases in value. And this is like why I tell like every journalist who's just starting out, you need to build up some brand and some audience right now to future prove yourself.

Chris: I think I wanna disagree with you on that, Ben. If anyone anywhere can start to create any kind of content I would argue potentially that there becomes a swing back towards a smaller set of authoritative brands, because content on a masthead called c n n is by definition, vetted and authentic.

Now, I have my problems with CNN and, how they think and how they vet and what they say and what they do, but the source of your content becomes more important, not less. and so certainly there might be a few independent creators and brands, let's say like Mr.

Beasts of the World or the Ben Pars of the world who are like, okay, I know that brand. I trust that brand. He knows how to vet. But I think people start to question the authenticity of content from people they start to stumble on, on the internet.

Ben Parr: I do disagree with you on this. in part because, there are certain things that AI has. As difficulty replicating, which is, completely new points of view. So, as an example, journalists, a lot are screwed by AI but not investigative journalists.

Why? Because the main skillset of an investigative journalist is not, the writing, it is the sourcing and the breaking of dos and stories, and it is the point of view.

 It is really hard to predict how things will move because I do like even Sam Alt Minnow, CEO of Open AI talked about this, where, more authority will concentrate in a smaller group of companies that do have the AI models and the data, and like, there's gonna be some crazy consequences, I suspect over the next few years.

my next column and the information are maybe the one that's posted now, is on AI and jobs, but, at least in the short term brand will matter look, the AI cannot yet replicate the dancing in the TikTok video or that like, personal point of view. And at some point, you know, it will get there and the voice is starting to get there.

Jeremiah: Ben, of video deep fakes that I've seen, which they're basically just taking out somebody's mouth and replacing it with some other words. You can't tell the difference already. So I realize that's not always net new content. I understand your point, but this is closer than we think.

Chris: Yeah, I think you are conflating two key things here. Time and supply side versus consumption side. On the thing of like, it can't replicate, I think that's a matter of days, weeks, or months maximum. Looking at the exponential curve of innovation here, and you are also talking about the supply side of like investigative journalism becomes more important.

I agree. But what I'm talking about is on the consumption side, I think there's a real probability that people will now trust fewer outlets for where they get their news from, they can't trust the veracity of that news. Whereas a CNN has investigative journalists, they have an established brand, they have tools, techniques, and workflows to verify the veracity of content

Jeremiah: Who said AI can't do all that stuff? By the way, AI could reach out to people and email people and send dms and get information, aggregate it, and pull it into a document and publish it. That could absolutely happen within by the end of the.

Chris: Actually, what I thought you were gonna say is couldn't you have a personal AI agent to verify the veracity of the content you're consuming

Jeremiah: You could have, you could do both. You could do both. So I think the notion, okay, AI agents, so if you're not familiar with this, go read Matt's post on autonomous, AI agents. Basically there are bots that operate independently. We could start to see ones where a journalist or a non journalist trains it to go out and, pings people and interviews them through digital means, and gets in sources and creates net new information that can absolutely happen.

Ben Parr: Agents are insane, because those are the things that could, like one could become a star theoretically if, you gave it the goal and the directive to become a star on a specific platform. I could absolutely see that potentially happening. again, I do think brand matters more, so it might be harder to build a new brand.

I might agree with you on that one, Chris. I think people who've already built existing brands, have more like defensibility in terms of like, look, I've already built a brand, or CNN's already built a brand and so I already built the existing trust. A new brand has a higher bar potentially, to fertile over in order to gain the trust of the public.

Jeremiah: I mean, we should also remember that the values are gonna change, right? So, you know, there was a point where people didn't wanna put credit cards on websites, so do eCommerce, right? We might see that people are just totally trustworthy of a digital journalist or a digital brand. what's her name, McKayla, that digital virtual influencer.

Jen Alpha, gen Z already follows her, trusts her, and she defines the fashion for those people. She's not a real person. I mean, that's already the norm for that generation. So I think we have to understand that the values of what we have now are gonna change. Just like the filters on TikTok and, and Instagram are already common.

We've already accepted that is reality.

Ben Parr: one other thought experiment for everyone. there's already groups of people with very low trust. let's say government officials is a specific example. how long until people are like, actually I'd prefer the AI to make decisions cuz God, do our politicians just suck. I don't think it's that far away.

I already think that sometimes I'm like, oh, dear Lord,

Jeremiah: We already use AI to make decisions for who we're gonna date, what restaurants we're gonna have, what content we're gonna read.

Ben Parr: People are already using AI to do the dating side. it starts on

Tinder. It writes the best passive message,

Jeremiah: Who gets hired is already done by ai.

Chris: So I actually had a Gen Z guy reach out to me and said, he had a great idea, for a startup, and he started pitching me this thing. And he was talking about contributing all our, material goods to a common pool and having AI allocate the resources. And he thought this was a fantastic idea.

Jeremiah: Sounds great.

Chris: He was describing like iRobot or something, and I said to him, you do realize that that's. Basically fascism powered by ai, right?

He's like, no, no, no, that's just branding. And he's like, you should think about it this way and think about it that way. And so I do think, through the, last few cycles in the, US political system especially, this kind of reverence for democracy and reverence for democratic values and free speech have, really eroded significantly.

And there are young people who are enthusiastically endorsing the idea of some form of command and control from the top.

Ben Parr: To be clear, when I heard you describe that, it sounded a little bit more like communism, which I find to be, more common, among sum Z and like, look gen Z and frankly, millennials too, should be upset at the existing capital system, given how much greater inequality and wealth has happened over the course of the last several decades, especially in the us but the line between different types of government styles and economic styles is really thin.

 The flip side is like I think there's an idealism set, like, the Star Trek universe. Chris, you and I both love Star Trek, You. Have a system where there's no income because people are not driven primarily by the acquisition of wealth.

Resources are allocated, freely because you could replicate them. And a lot of the discussion of just like, what will the world look like once agi artificial general intelligence becomes a thing. there's so many people who are waiting for this world where AI disrupts everything because their existing lives in the existing system doesn't work for them.

And that might make AI much more accepted across society.

Chris: So I don't wanna, go down the rabbit hole about politics too much here, but I think where this connects back to the topic is, I think the Star Trek Utopian World works because Star Trek is an a post scarcity economy where everything is free and easy to acquire through replicators and transporters and, warp energy and all this kind of stuff.

And I think when you have an abundance or an infinity of, resources, scarcity and, resource allocation becomes a non-issue. but this connects back to AI when you realize that potentially AI turns certain parts of our economy into a post scarcity environment,

So again, content production becomes almost free. video, photos, audio doesn't matter. decision making of certain kinds becomes free and a whole bunch of labor is displaced. and so when we're talking about impacts on society, how do you guys see this affecting the future of work when whole sections of intelligence and labor are displaced from the workforce?

Jeremiah: There's going to be some temporary pain as people fall out of their role or they have the opportunity to upskill into something that might be more interesting to them. But in the net midterm, Humans can focus on things that they might enjoy And I think we'll see more joy returned to life.

Basically what automation does is it's very good at automating replicated tasks that are repeated. these are not things that many humans often enjoy at work. There's some things we do love to repeat, of course, in your personal lives,

Automation can manage all those things. So we are gonna see people go up Maslow's Pyramid maybe towards self-actualization. They have more time in the day. They have things to think about. Yes, we need to solve the economic unbalances that will be forced to be reckoned with for sure. but I see this as a net positive for humanity and for work.

Chris: Jeremiah again, I, I think I agree with you over the long arc of history.

But in, the short term, and when I talk short term, I'm talking even five, 10 years where work is still seen as part of self-identity and worth and value in, our economy and in an hour society. I'm conscious of being the old man shaking his fist to change, right?

Because at the advent of computers and the advent of every new technology, people are like, oh my God, this is gonna just destroy all work everywhere and we need to struggle and fight against it. But I think in the case of ai, that might actually be true. the demand destruction for labor might be so.

Exhaustive and rapid, that society's ability to digest and create new opportunities for work or to slide up Maslow's hierarchy of needs may actually be itself disrupted. And that, that's kind of scary to me.

Ben Parr: This is the topic of my next column and the information. And, I tend to side more with you, Chris, on this topic. look, if you have a 10 person marketing team, you can do the same amount of work with three people, leveraging AI now. so, it's not that it replaces every job, but if you can do that across knowledge work, this is the first technology that truly replaces knowledge work.

We haven't had that before.

Jeremiah: And creatives. We haven't had creatives displaced like this before.

Ben Parr: And when you have that sort of technology, it will disrupt a lot of times. Now, there's a couple pieces here. One, companies are slow to adapt to changes, and so it'll take a while until they're like, oh, I can like cut some people and like, adopt a new technology.

Instead, it takes companies time to do that, but once they've done a cut, they won't come back and add new people because they're like, I don't wanna hire new person can instead add this AI technology. layoffs theoretically should be widespread and look one of the smartest people in our space.

Again, Sam Altman, tends to believe that this is will happen and believes that one of the real solutions is, a universal basic income. Believes one of the solutions is changing the way we think about. Work overall. and so there will be real disruption in the course of the next 2, 3, 5 years. I don't exactly know how long it'll take and it's less about the technology being there.

It's already kind of there and more about how long it takes for these changes to shake out, with people, realizing how much of their companies they could automate. I believe that there will be a billion dollar company built in the next five-ish years that has one to three people tops because you can automate almost everything else. It's going to happen.

And so, yeah, what do we do as a society? the biggest thing is that we're not really society talking about any contingency plans, any solutions, cuz we're so in denial that this could be a thing. even if it isn't a thing, need a contingency plan.

Need it now.

Chris: So Ben, again, I think this is gonna happen much faster than we realize because you talked about companies being slow to adopt, and therefore slow to fire people right size their workforce. The thing is, a lot of companies have already fired huge chunks of their workforce and are not gonna rehire them.

So the firing has already started in large chunks. And the other reason I think this is gonna happen even faster than usual is. These tech companies are baking the AI into their core common tool chain. And so all of a sudden, you know, an older person, who's not used to picking up new tools they're using Excel and, PowerPoint.

They finally learned how to use those. Well, guess what? Excel and PowerPoint has AI baked in, right? There's no new tool to learn. And so the right sizing has already started. The tools are, familiar and the rate of change is accelerating. this could happen in a much more compressed time scale.

Jeremiah: The trades need workers more than ever, so we might see a shift towards that. Anybody who puts their hands on somebody, for medical reasons, health, wellness therapy, physical therapy, those are jobs that still need to be done in the trades as well. So we'll see people start to shift around.

I think the really interesting notion is that for the first time developers, coders, software engineers, some of those low level positions or entry level positions are actually gonna be disrupted. And there's some of the people that are actually creating these technologies. So that's kinda like slapping your own hand in a way.

It's such a wild concept to think about. And those have been the, hardest to fill most in demand, highest paid worker jobs to date aside, perhaps from some advanced medical. So that's really crazy thing to see.

Chris: When we've been spending the last decades saying, well, minors and whatever should just go get retrained. And we've been looking nose down at, blue collar workers going, well, you just go get retrained. And now it's the white collar workers who need to go deal with the retraining, right?

so it'll be interesting to see how we react, in that new environment.

Jeremiah: Last thing with that though, as AI becomes more popular, robotics becomes easier to manage, so this is gonna span other industries as well.

Ben Parr: Robotics is an interesting topic on this front because. robotics by its nature is harder than automating knowledge work because, bolts and screws of things can, break down and change. And look, there are some amazing robots that we started to see. They're still prohibitively expensive, but the price will go down.

Probably the next disruption that affects non knowledge work using AI is driving, and self-driving has gotten really good. It's still not yet there where people are willing to use it every day. But why would a truck company ever use a human when you could use an AI that could see faster, react immediately, not get emotional, drive 24 hours a day instead of the, like 14 maximum that most truck drivers are allowed to do in the US? It's gonna happen. It just takes more time for the more physical tasks to be automated and they'll happen at different speeds, but AI will be impacting them too, cuz the AI can make decisions like which way the wheel should turn or which way the screw should go.

It is interesting though that one of the recent reports was, the last job to be automated or the least affected by this AI boom, is plumber.

Chris: So we've talked about, all the possible ways that founders can engage with the opportunity of AI and how it's affected, Silicon Valley culture and how it's affected the culture at large.

So the three of us actually engaged in an exercise to try to classify and rationalize some of this excitement and energy and, new tooling that we call the AI classification framework. so. By way of history, I was listening to Reid Hoffman interview Sam Altman, and they were commenting on this idea that the touring test was too, simplistic a model in order to determine the value or efficacy of an AI tool.

They were saying it's kinda like a simple pass fail, intelligent, not intelligent, or sensient not sensient, test. And the reality is that these tools are emerging with different capabilities chewing test is just not a good enough model And so that sponsored a thought in my mind of like, well, what would be a different, better way of classifying these tools? I brought that to Jeremiah and Ben. I said, Hey boys, let's create, some kind of framework to plot these tools, on a chart and help investors. Develop a thesis around it, help founders develop a, a vocabulary for how their tool differentiates from other tools and to help end users understand, well, I need this mix of tools to fill in the puzzles of what I need from my AI tools.

And so we created this classification Framework. And it, borrows or steals heavily from this thing called the theory of multiple intelligences, which says that as a human we have, five or seven different kinds of intelligences. We have existential intelligence, logical mathematical, linguistic, verbal, visual, spatial music, rhythmic body kinesthetic, interpersonal and interpersonal.

And so we, took those plotted them out. And then for each, there is a kind of a zero to five scale. can this tool. Engage with, let's say, interpersonal intelligence at all. And if not, that's a zero order that exhibit maybe the capabilities of a teen or a college graduate, or an expert or even a master or a super intelligence at level five.

And the expectation, I think the implicit expectation here is that all of these tools will inexorably start to move towards level five intelligence and that there will be one or more tools for each of those. Types of intelligences, I think over time, over the span of, let's call it five, ten, twenty five years.

Right. it sparked a really interesting conversation. It was, launched on TechCrunch and, got a lot of buzz. I'd love to hear your guys' thoughts on why you chose to participate with this framework. what value you think it brings and any reactions you guys have had from the community as you've been engaging with, AI startups at these hackathons around the world.

Jeremiah: Yeah. Chris, thank you for leading the project and inviting. So I'd love to categorize and chart, it's my training as an industry analyst back from Forrester or the Altimeter group where I was a co-founder I liked bringing some order to the, thoughts out there and it was really helpful.

and I was working with you on these, maturity, rankings, right? So level four we consider that a master, somebody who, spent 10,000 hours or 30 years at her craft. And that's kind of how we were thinking of that type of person, like the top 1% in their field.

And then we're thinking, what is a super intelligence? And this is borrowing from Nick Bostrom's book, super intelligence around when, the AI becomes more intelligent than the best human, and they would rapidly double Triple and then exponentially could be a hundred times more intelligent, if not a thousand times more intelligent than a human.

So that's what a super intelligence would be. So that's just my comments on how we were thinking about that and looking at the modalities, which you went through all of them in the beginning of that. Most of them were on the far left in maturity. I think most of them were scored like one and two I think.

so we'll see. I think we should actually rescore it maybe annually and then republish it to TechCrunch to show how things are changing.

Ben Parr: Uh, you know, once, Chris was like, here's, this framework I've been thinking a lot about, I was like, oh God, what has Kris Deno really? it's been long needed and I'm glad you took such initiative and that you looped us in on the initiative. we just do need new ways to grade AI systems. I think we need to be able to grade it based off of how human-like it is.

Like we're trying to grade for AGI right now for artificial general intelligence, the kinda AI that can do the things humans can't. And, the framework that, we've developed, that you've developed and that we've contributed to, does that. And it's really important for us to track where things are going with different AI tools.

Some more strong than others, but the reality is we wanna see, an intelligence that could score a five at all the points and then at some point, you know, might score, a seven and eight, a 14. I know there's not a 14 on the scale, but still, you know, at that point then, maybe all human consciousness has merged into one super entity.

And we can all throw a party in our brains. Nice. I don't know. but we need to track it. It's a really detailed model, and I think we're gonna be seeing those numbers go up in the next couple of years annually. Might even be too slow.

Chris: I was going to say that exact thing, I think annually is too slow. We almost have to do this six months. it's interesting. I think we didn't frame it this way, but I think if, a tool or a model can hit level three on most of the types of intelligence, I think we could probably reasonably classify that as generalized intelligence.

Is it self-aware or self-conscious? could it be considered to have consciousness? I don't know, but it would definitely be generalized and it would definitely be intelligence.

Alright, boys, this has been an incredible conversation, far-reaching and, incredibly in depth as well as I expected from my two friends. how can people find you on the interwebs and, reach out to, collaborate with you or to learn more?

Jeremiah: Yeah, I'm most active on Twitter and my handle is my first initial last name, jowyang.

Ben Parr: I only exist in artificial intelligence clouds, so no. At ben Par, b e n p a r r on every social network, Twitter, LinkedIn, TikTok are my most, useful, interesting, active. And then for sure my newsletter, ben par.dot com, the AI analyst, and so bunch of new stuff I talked about when my new columns come out, everything.

Chris: Awesome guys. We'll include links to that in the show notes, of course. And for me, you can find me as at Chris Sar on all the social media. And of course, if you're interested in the topics we discuss on the startup podcast, including ai, I have carved out some time to work with a small handful of companies.

So feel free to learn more about that at chrissaad.com/advisory. and of course, don't forget the Startup Pact, which is our little, agreement with the audience. If you've listened to a few of these shows and gotten some value from it, the only thing we ask for in return is that you follow us on your favorite.

Podcasting app. Read us and review us, and also find us on YouTube. We're trying to grow our YouTube channel. You can actually watch this episode on YouTube and see Jeremiah and Ben's, pretty faces and we're trying to grow that channel. So it'd be really fantastic if you could, give us a subscription over there as well.

Boys, thanks as always. Hopefully we'll catch up in person soon. And, as always, we'll catch you in the next episode. See you guys. Bye-bye.