Hard Fork Live（第2部分）：Stripe 的 Patrick Collison、Skip 的 Kathryn Zealand + 听众提问

Hard Fork Live, the podcast's inaugural event, was sponsored by premier sponsor IBM, associate sponsor Invesco QQQ, and supporting sponsor Intuit QuickBooks.

Well, Casey, what was your favorite part of Hard Fork Live?

There are so many things to choose from, Kevin. One thing comes to mind: when I showed up at the venue, they had no idea who I was and didn’t want to let me in. I went right up to security and said, “Hi, I’m Casey. I’m here to co-host the show,” and they said, “Who?” I said, “Oh, gosh, am I going to have to pull up a website on my phone and show you my face?” And I did, and then they let me in.

That’s good.

What was your favorite part?

I told them not to let you in. That’s why that happened.

Oh, dang it, Roose, you got me again. You scoundrel.

My favorite part was definitely this bit where we were changing out of our regular pants into some mechanical robot exoskeleton pants—

Mm-hmm.

—for a demo. We had practiced this the day before during rehearsal in sort of an empty backstage area. I’m taking off my pants, and I look up, and there’s Patrick Collison just looking at my fleshy, pale thighs.

I know, yeah. One of Silicon Valley’s great thinkers and leaders—and there we are, our trousers dropped, and he’s thinking, “What kind of show is this exactly?” Suffice to say, if you weren’t backstage at Hard Fork Live, you didn’t see the whole show.

I’m Kevin Roose, a tech columnist at The New York Times.

I’m Casey Newton from Platformer.

And this is Hard Fork.

This week, it’s part two of Hard Fork Live. We’ll talk to Stripe CEO Patrick Collison, and try on some robot pants with Skip’s Kathryn Zelen. Plus, members of the live audience ask us anything.

Anything?

And they did.

Well, this next guest I’m extremely excited about. I have known Patrick Collison and his brother John since the early 2010s. They had recently arrived in the city, and they were starting to build a company called Stripe. In the years since, it has grown into a juggernaut, and Patrick, I think, has become one of the most interesting and influential thinkers in tech.

Outside of his day job, he and Tyler Cowen wrote a piece in The Atlantic about the need for a new science of progress. He’s also the co-founder of the Arc Institute, which is doing a bunch of really important biomedical research. And, just to top it off, in April, he joined the board of Meta, which is currently reshuffling its AI team in ways that might be fun to talk about. So I’m extremely excited to welcome to the stage Patrick Collison.

Hey.

It’s real.

Hi, how are you?

Hey, good to see you.

Ah.

All right.

Patrick, thank you so much for coming.

Thanks for having me.

How are you?

I’m doing well. It’s wonderful to be here, where I’ve come to so many Long Now talks, and now it’s you guys.

There is no Long Now like a podcast, is what they say.

1. The Abundance Agenda

So, Patrick, you are a big proponent of what they call the abundance agenda. I think I can say that without paying a royalty to Ezra Klein and Derek Thompson. What is the most urgent abundance-related need right now? Is it housing, land use, fast-tracking approvals for new drugs? What is it?

Building.

Hmm. And why is that?

Well, I think, look, there are so many domains where we need more progress than we’ve garnered. The rate of improvement of life expectancies has really diminished, and you still get colds, right? That’s crazy. We’re on a trajectory, supposedly, to invent nuclear fusion and AGI, but we haven’t cured even the common cold. There are a lot of challenges we face in many domains. There are many pursuits in science and technology where we should be going much faster.

However, I think the place where the constraints are most unnecessary, most self-imposed, and easiest to—or at least, in principle, easiest to—address is in any kind of physical construction. By the way, California and San Francisco used to stand for this, right? We have the Kaiser signs around the Bay, and Marine Ship was in California. Those stories of the ships that got built in a day—they happened here. California is this funny superposition where, in the ’40s, ’50s, ’60s—I mean, actually, my favorite example of this is Treasure Island, which we built to celebrate the construction of the Golden Gate Bridge.

We were so drunk on the fervor of having finished the Golden Gate Bridge, caught up in the momentum of it, that we were like, “Well, let’s just go build another thing,” and so we built an island. And so California and San Francisco used to stand for this. And now California, my favorite example of this is that we passed a ballot prop for California high-speed rail. They used to have on the website—you’d go there—and the forecast completion date was 2033. I just checked recently, and they’ve taken it down. It used to be a 37-year project, and now who knows?

Yeah, it’s so interesting. I was interviewing someone once about their AI forecast and how everything was going to change in the next 2 or 3 years.

Construction.

We were going to have superintelligence and robot factories, and we were sitting in Berkeley, looking out over downtown Berkeley. I said, “Do you think any of that will get permitted by zoning?”

No.

“No.” So even the people thinking about the wildest possible futures cannot imagine land-use reform in the Bay Area.

Yeah, well, another funny example of this was the announced aspirational city project in Solano County a couple of years ago. You guys covered it on the podcast.

Yeah.

Thousands of times in its history—and America’s history is only a couple hundred years—America has decided, “Hey, let’s build a city.” That is a thing that has, in fact, ensued, and there is a city there today. It’s happened even recently, with Irvine and so forth.

But the response to that was astonishment, and people were aghast. That kind of audacity was offensive. When somebody says, “We’re going to create a superintelligence to populate the cosmos,” we’re like, “Yep, seems pretty reasonable.” But when someone says, “We’re going to build a new city in California,” it’s like, that’s just ridiculous.

2. Tech Faces A Trust Problem

I do think that there is an element separate from that, which is that with a lot of tech, some folks just feel like they’ve gotten a bad bargain, right? Maybe with social media, they don’t quite like the bargain that they’ve gotten, or they worry about the environmental impact of some of these technologies.

Sometimes I do think technologists come along and say, “We want to take a crack at this,” and people just think, “I don’t actually trust you to do that.” I imagine you have done some thinking about this as somebody who is trying to bring your influence to bear on public issues. So how do you reckon with that?

Well, you’ve got the Homer car problem.

Yeah.

Nobody wants that.

In many domains, I think it’s the case that if you go and talk to the insiders and the people who’ve been plowing some lonely furrow for a long time, they are brimming with ideas for how it could be done better. It’s not a matter of somebody careering in from outside with delusions of grandeur and some sense for how it could all be done differently. I think in many cases it’s actually about unleashing the ideas inside the system that exist but are not yet manifested.

One of my favorite examples, and one of the most striking examples of this, is that we ran a survey a couple of years ago of practicing top-tier scientists. We asked them not if they had more money, but if they could spend their current funding dollars however they liked. Today, funding dollars come with all sorts of restrictions, and they’re allocated by committees and by consensus, with tight field definitions and so forth.

Anyway, we asked these scientists, “If you could keep your current funding level but spend those dollars however you want, how much would your research agenda change?” I thought the results might be striking.

Maybe a third of scientists would actually like to be doing something else. 79% of respondents in that survey said they would change their research agenda a lot. These scientists are so self-sacrificing. They could go and make a bajillion dollars or, these days, 100 bajillion dollars, by doing something more lucrative, and they've decided to try to better humanity by working inside the academy.

They spend decades training, and then 48% are telling us that they would be doing something very different—not, again, if they had more money, but if they had fewer strictures attached to their current money. So anyway, I think that's one striking example. That was one of the influences for Arc. But I think in many of these cases, whether it's urban design and urban planning or any kind of construction or science or whatever, the insiders often know how to do it a lot better.

You mentioned the $100 million figure, so I have to ask: You are on the board of Meta, which is currently reshuffling its AI teams, building a superintelligence team, and reportedly offering people $100 million to come work for them. Have you been helping with that, and what kind of advice are you giving Mark Zuckerberg about that?

You add the payments guy to the board, and then all of a sudden... I've just had 1 board meeting, so I shouldn't comment.

3. Agents Take Over Commerce

Okay. Well, let me ask you about something you can comment on then, which is this idea of agentic commerce, which I've heard a bunch of times recently, and I think it means something like robots buying stuff. This is a big idea in the tech industry right now: You're just going to have these bots going out and buying things on your behalf, or agents transacting with each other. Right now, Stripe is a platform to help people, human beings, pay for things. But I can go on Operator and have it buy me a pizza or a burrito.

If you have 2 hours for it to complete that process.

Yes.

Yeah.

I imagine you're planning for a world in which this kind of thing is happening more. What does that look like, and is Stripe still necessary in that world?

Well, look, Stripe enables the transactions and money movement beneath the surface. So we're, to some extent, agnostic about and enthusiastic about all the different modalities in which that might happen. When Stripe started, mobile wasn't nearly as big a thing as it is today, stablecoins didn't exist, and so forth. So we're enthusiastic about the reinvention here.

From my standpoint, the variety that one can benefit from on the internet is tremendous. I think it's amazing that you can have so many niche businesses serving these preposterously large, global audiences and customer bases that couldn't exist without the internet, right? Because if you're only restricted to the hinterland that you locally serve, you get much less; there's much less that's possible.

But very few people say, “You know, the thing I love about internet commerce is that after I click on the thing, I'm brought to sort of a—”

Checkout flow.

Yeah, like a 30-field form to fill out and then fax us parts of it in Latin. It just feels very antiquated.

I think the promise for these—I mean, I don't know, I think “agents” sounds very highfalutin, but these critters—to go and be the paperwork minions for you, to dispatch one and say, “Hey, go handle that, and then I will resume whatever else I was doing”—I think that's very promising, and I think it's going to be better for everyone.

Will the agents use dollars or crypto?

I know I'm here with a well-known crypto booster, Casey Newton, but I think they will use amounts denominated in the local currency. I think in the U.S., things will be in dollars, and euros in Ireland, and so forth. I think there's kind of an irony where, with the invention of crypto, the initial conception of Bitcoin was, “We're going to throw off the shackles of the monetary tyranny and oppression that we're subjected to and liberate ourselves with this new currency.”

You look at Bitcoin, which has obviously garnered a lot of traction, but a majority of cryptocurrency transactions today are in stablecoins, which are dollar-denominated. I think it might be the case with stablecoins that what actually happens is, rather than overthrowing the dollar, it enables people outside the U.S. to access the dollar in ways they couldn't previously.

Yeah.

I think this is the biggest thing that's actually underappreciated by people here in America. For outsiders, lots of currencies are not great things to hold savings in, or not great things to transact in. I don't want to name names and single any poor currency out, but there are many currencies where, when you go to give them to your counterparty, they look at you funny and ask you, “What is that?”

Kohl's Cash comes to mind.

Yeah. And there are a lot of currencies that have lost 75% of their value in the last 5 years. Providing the ability for people outside the U.S. to hold dollars could, I think, be a big deal. So anyway, I think there's kind of an irony in crypto where one of the most important and consequential things it may end up enabling is the dollar's greater success.

4. Biology Gets A Virtual Cell

You are a co-founder of the Arc Institute. You mentioned it a moment ago. You also mentioned the persistent problem of the common cold. Do you think there is something that can be done about the common cold, and are you going to work on it?

We have a big announcement tonight. No. That mist on the way in was actually a vaccine.

Humanity has never cured a complex disease. This was only a penny that dropped for me a couple of years ago. There are some diseases—infectious diseases: the cold, COVID, what have you. Then we have monogenic diseases, Huntington's, things that are the product of some specific genetic abnormality.

But then you have all these diseases that are some combination of the environment, whatever you do with your life, and the genetic risk factors you might have. So this is most cancers, most cardiovascular disease, most neurodegenerative disease, most autoimmune disease, and so forth. Anyway, we've never cured one of these, and it's too complicated because figuring out the genes, the environment, the durations that elapse, and so forth is just... it's kind of intractable.

So we set up Arc Institute to try to—I mean, time will tell whether we're actually able to make any progress here. The idea was, “Hey, can we try a different strategy for going after these complex diseases?” One of the things that has been a real boost for us over the last 2 years is actually AI.

The LLMs, as we interact with them, are obviously wonderful and enable all sorts of productivity enhancements and so forth. But there's another language: DNA, the language of life. Hitherto, we haven't really figured it out. It's billions of base pairs, and we can't at a human level understand what's going on there. It's kind of beyond what any individual can comprehend.

What we've seen with some early work—I mean, we published our first virtual cell model yesterday, actually, by coincidence—

Yeah, tell us about that. What is a virtual cell?

Why haven't we cured cancer, or any complex disease? On some level, the problem is that experiments in biology are slow and expensive. You're dealing with actual physical cells, and you have to wait for them to grow. So if you have some hypothesis or some idea, you do things to the cells, and then you wait 3 months, you wait 6 months.

You have to purchase the reagents—the ingredients. They’re expensive. The supply chains span multiple countries, so it’s laborious. In software and engineering, you have an idea, write a quick command at the command line, and 10 milliseconds later, you have your response. In biology, that loop takes months to elapse.

So the idea behind a virtual cell—and this is not Arc’s idea; others have had this idea—is that if you actually had a useful, accurate way to perform computational experiments, where you try the thing that, in accordance with your hypothesis, you would try in silico, as the biologists say, because they always like to make things complicated and fancy, then, if you just computationally do your thing and that was actually accurate, it would be an enormous accelerant to biology. So anyway, that’s the idea.

We haven’t had that until recently. I mean, is this too much biology? Should I stop?

No, we like it.

Okay. All right.

You guys like biology? Woo!

One more thing, then I’ll stop. In the last couple of years, we’ve got technologies in 3 different domains that are really important.

A computer is composed of the ability to read, to think, to compute, whatever, and then to write. Those are the core 3 operations. In biology, in the last couple of years, we’ve got the ability to sequence individual cells, figure out what’s going on in just 1 cell, which is a big deal. We’ve gotten deep neural networks and transformers and all that stuff, so we can think over data of very significant complexity. And then we’ve had huge advances in functional genomics and CRISPR and the ability to make individual edits to individual cells.

Those are all big breakthroughs in their own right, but when you put them together, there’s this shimmering potential promise that maybe you could, for the first time, enable accurate computational predictions very quickly and very cheaply. And again, if that worked, it’d be huge in biology.

Mm. Drink well.

Yeah.

Drink well.

No, no, no.

Yeah.

You guys don’t type if it works, but—

5. Reading Beyond AI Summaries

You are a voracious reader. What is your case for reading the whole book instead of an AI summary? Or is that not a case you would make?

Gosh. I think reading the AI summary can be useful. I think it can have a powerfully inoculating effect.

I think TV is a waste. Long-form TV—lots of episodes—is a waste of time. But it’s very tempting, right? You watch an episode and you’re like, “Oh, geez, I really want to know what happens here.”

So I discovered a couple of years ago that the solution is you immediately go to Wikipedia and read the entire summary of the season. And then you’re cured of the desire to watch any more episodes. There is a therapeutic benefit to AI summaries.

That is the most San Francisco thing I’ve ever heard.

But—

I love it.

Yeah.

But one does not derive any deep pleasure or betterment from it, so I’ve never written an AI book summary.

Yeah. All right. Patrick, can we do some lightning-round questions?

Okay.

When do I get to ask for the Stripe feedback?

I am a Stripe customer. Actually, very quickly, you’re doing automated disputes now. Very excited to try this.

Yes.

Yeah.

Okay, good.

Yeah. Yeah.

And everything else is working well?

Everything’s otherwise working great.

Okay, great.

Yeah, yeah.

All right, lightning-round questions. What age will the median child born in 2025 in the U.S. live to be?

So, 25 today?

Yeah. Mm-hmm.

Yeah. 95.

Oh, great. What is 1 thing you wish an AI agent could do for you reliably that it can’t quite yet?

Reason over the scientific literature.

Mm. Mm.

It gets tripped up on the paywalls.

Well, we know how—

And I think, why hasn’t anyone who’s faced that done anything about it?

Yeah. We love paywalls in this house.

But imagine if you had critters paying at them on behalf of unsuspecting people.

Yeah, I love that.

That sounds beautiful.

That’s a business model for the media.

Yeah.

I’m excited to present the critter strategy at the next meeting. What trait do you hire for that most companies undervalue?

For Stripe, we really care about long-term thinking. Infrastructure can’t happen overnight. We’ve been working on it for 15 years, and we still feel like we’re just starting. Our products don’t have TikTok trajectories, and the super successful ones still take a decade to fully play out.

I like the analogy that those kinds of technology companies are building cars and roads. We need cars, and cars are super cool and beautiful and all the things, but someone’s got to build the roads. Stripe looks for the kind of people who are predisposed to working on roads.

Is there a book you read recently that made you say, “Everyone I know needs to read this”? Or even not that recently.

I’ll give you 2. The Dream Machine—hands up here, who feels like they really have a good understanding of the history of the internet?

Shame on you.

Well, it’s funny. Alan Kay once commented that computer science is a pop culture. He thought that we were recapitulating the same ideas in blissful ignorance of our history. And I think there’s something funny where we as technologists aren’t as into and fascinated by our own history as people are in other domains.

The Dream Machine is the best history of the internet and the thinking around it, and the motivations around it. It’s also published by Stripe Press.

Then, 1 I read reasonably recently that kind of blew my mind was The Demon Under the Microscope.

Ooh.

You all know that penicillin was the first antibiotic. Except it wasn’t. There was 1 before it, and it changed the world. The story behind its invention is very interesting. So those are my 2.

Very cool.

Great. Patrick Collison, thank you so much for joining us.

Thanks, Pat. Thank you so much.

Thank you.

Thank you.

Thank you.

When we come back, even more of Hard Fork Live.

6. Exoskeletons Hit The Stage

Well, Casey, we’re wearing what I would call robot pants.

Yeah, I think that’s fair.

Why are we wearing these robot pants?

We are wearing these robot pants because they give us a great excuse to talk to Kathryn Zealand.

Yeah, so Kathryn Zealand is the founder and CEO of Skip. They’re the company that makes these pants. The company spun out of X, the Google experimental research division—not the social media network owned by Elon Musk—in 2023. And since then, she’s been working on this exoskeleton movewear, like these pants, which have been called e-bikes for walking. Let’s bring on Kathryn Zealand from Skip.

Kathryn. Thank you, Kathryn.

Of course. Thank you. Good to see you.

Ah. All right.

You guys didn’t see, but backstage we just did the best quick change in theater history.

I just took my pants off in front of Patrick Collison. So, Kathryn, in a few words, tell us who your products are for and what problem you’re hoping to solve.

I’m sure I’m not the only one here who’s got a loved one who’s struggled with movement, or maybe you’ve got knee pain or had an injury. At Skip, what we want to do long-term is help anyone achieve whatever they want to do, regardless of any kind of mobility impairment.

This is our first product. This is the MO/GO. And this is actually aimed at people who want to go hiking.

But often, they may be struggling a little either with knee pain or with a kind of reduced mobility. An e-bike is a really great analogy. It particularly helps with the inclines—steep hills and stairs, things like that. But then eventually it will help, hopefully, everyone.

So tell us: What is in these pants? Folks who may be sitting a little farther back may not be able to see, but in addition to these pants, which I believe you partnered with Arc'teryx to make—

Yeah.

There’s also some pretty heavy machinery on here. So what does this stuff do?

Yeah. So the real magic is these lightweight motors. Here’s one I prepared earlier. Really, it’s only in the last decade or so that robotics has gotten efficient enough, affordable enough and light enough that you can have something this small that can do up to maybe 40% of a healthy person’s muscle forces. There’s a motor providing physical oomph, and built into the pants—those near the front can maybe see this—is a little cuff situation to help transfer that force to the body.

And I know that one of the things these pants can do is help you stand up, which you were explaining to me is so helpful for folks who are older and need some help with that. And I thought, “I need that help now.” So I think your market may actually be larger than you first suspected.

Yeah. I’m actually having a similar problem, because I don’t know if you noticed, but I’m quite pregnant.

I wasn’t going to say anything.

Thank you. It’s 8 months. Kathryn joined us at 8 months pregnant, so please do give her a round of applause.

All of a sudden, I also need help walking upstairs and standing up out of chairs.

Now, do you envision a world where lots of people are wearing exoskeleton-type movewear, or do you think this will primarily be for people who have mobility issues?

I think a bit of both. If you’re truly able-bodied, a product like this is probably just not worth your time necessarily. But there’s a huge range of people who might be recovering from an injury. They might be pregnant. They might have joint pain. They might have had surgery. I think probably most people at some point in their life will use one of these products.

We also do R&D in things like Parkinson’s disease and people who have more severe issues as well. But for sure, it’s much broader than the first set of products that you’re seeing coming on now.

As this technology develops and you’re able to shrink it down, do you see a world where more able-bodied people might just choose to wear it because they like feeling a little bionic when they walk around town?

I was going to say, are you trying to dunk a basketball? Is that what you’re getting at?

I’ve been trying since 1992.

For sure. But I would say, if you just want to make your exercise harder—because this is something that some people ask, they’re like, “Oh, can I get resistance mode?”—you could also try ankle weights. I think they’re going to make a comeback.

That’s good.

Now, exoskeletons are a staple of science fiction. I remember watching movies as a kid that had exoskeletons, and lots of other stuff in those movies has become commonplace, but not exoskeletons or this kind of hardware. So why is this taking longer than other types of futuristic technology?

I think there are 2 reasons. One is that hardware is hard, and people are really sensitive to what’s on their body, right? You’re wearing an early version of the prototype, and it’s probably not super comfortable. The fit’s not always perfect. You might tolerate an uglier and imperfect robot if it’s just in a factory, but when you’ve got to wear something, it’s important that it’s comfortable, that it fits really well and that it looks good. So the bar is really high for anything that’s worn.

I think the second reason is actually around AI. Historically, exoskeletons and all robotics have done very repetitive tasks, right? In a warehouse, doing lifting was one of the first places you saw exoskeletons actually become used because it’s the same task again and again. But when you think about maybe an older adult who wants to move about their community, they’re doing lots of different movements, and they’re moving in weird ways.

I’m a fidgeter. In one of our early prototypes, I was sitting at a very important meeting with my boss, and I was tapping my foot because I was a bit nervous. All of a sudden, it thought I was trying to stand up. It turned on and threw my leg up. The whole table went, “Oh,” and everyone’s water spilled, and there was chaos.

That was actually the plot of M3GAN 2.0.

Yeah.

That happened there.

Okay. Well, we’ve talked a bit about this product. Anything else we should know about it before we see what it can do?

You signed the disclaimer backstage, right?

No. We’re just going to wing it.

Excellent.

So let’s stand up.

Start by turning it on, even while you’re sitting if you want.

Yeah.

There’s a little button at the back here. Press and release, and the LEDs should change color.

Mine is already blue. Does that mean it’s on?

Yeah.

Now you’re on. Oh, hello.

And now you can try standing up and sitting back down. That’s a very common movement.

Okay. Yeah, this is giving mecha.

You can do some squats.

It actually is helping me stand up and squat. No, this is great. I feel like, you know, when you have a spotter at the gym?

Mm-hmm.

You wouldn’t know. It’s like that feeling when you’re just getting a little boost.

Yeah.

Yeah.

Exactly. But the problem is, you both strike me as quite fit, healthy people.

Oh, thank you. We’re always hearing that.

Exactly. So we were trying to think of a way that we could make this a little bit more challenging for you, so you really can appreciate the technology. Have you noticed that there’s a StairMaster behind you?

The StairMaster! You know, I did. Can I try?

Yeah.

With my—okay, please clap. Yeah. Wow. We’re really doing it. Okay, let’s put the stairs on.

About a level 3 or 4, maybe.

I’m a 5. I like to live on the edge.

Great.

Okay.

Great.

Okay, it’s good. I feel like what I’m doing is not climbing stairs. It’s like climbing stairs adjacent. It’s sort of like the physical equivalent of vibe coding. I’m vibe walking.

Right.

Okay.

One problem we have is that your brain is amazingly good at getting used to a new feeling, and so sometimes people aren’t even sure how much it’s helping. If you want, we could turn it off, and then you might see the difference.

Oh, you’re going to remotely disable it? Yeah, please do. Oh, God.

I’m saying we can if you want.

No, do it. Do it. I want to experience this on my own two feet.

All right. We’re going to turn you off in 3, 2, 1.

Off. Oh, yeah, that’s much harder. I don’t like that. Can you turn it back on?

You want it back on?

Yes. This is how you get your customers. They can’t go back.

But this is genuinely how we often get customers.

I’ve heard of the hedonic treadmill, but this is ridiculous.

Okay, you’re back on now.

Okay.

So is that a bit easier?

That’s good. Thank you. I think that was all right. I don’t know that it could’ve been done better. Oh, would you like to try? I would. Would it be okay if I tried?

Woo.

Hey, it’s Casey just cutting in from the studio here, and you’re not going to hear it in the podcast, but I do want to tell you what happened immediately after this, which is that I challenged Kevin to see if I could do better on the StairMaster wearing the robot pants, and I did while accompanied to the song “Work, Bitch” by Britney Spears. For copyright reasons, we’re not going to include that in the show, but you can just kind of hum it to yourself or maybe play it on Spotify later today. I think when I kicked my legs up, the StairMaster stopped because it was afraid for me.

It’s always a good sign when the worst part of our demo that goes wrong is the StairMaster. So that’s a win.

Absolutely. That was incredible.

Now, Kathryn, tell people where they can find out more about these pants if they’re interested.

Yeah. So we’re Skip, so skipwithjoy.com. I presume that’ll be in the show notes. And you can pre-order them now. We’ll be shipping next year. And we also do, like, rentals and experiences, especially for folks who really want to achieve a dream hike, but they’re not able to at the moment. We’re trying to help them do that and get feedback.

And how much for the pants?

A bit like an e-bike, so $5,000-ish.

Okay. And what would it take to bring that price down a little?

It would either take a discount code if you know me—which I feel like we’re all now friends—or improved international supply chains, which is a different conversation.

All right. Well, Kathryn, thank you so much for joining us. We really had a great time.

Yeah, of course.

Thank you. This was so fun.

Thank you.

Thank you. Thank you. Well, Kevin, as we say at the end of so many podcast segments, it’s time to get out of these pants. Casey, what’s next? I know what you’re thinking, which is: Can the show possibly go on longer? And yes, it can.

7. The Audience Takes Over

Because here’s the thing—and this, I mean from the bottom of my heart—the very best thing about this show is hearing from you all. Look, when 2 men decide they love the sound of their own voices enough to start a podcast, you never know if you’re going to hear anything back. But you guys email us each and every week, and we just wanted to talk to you directly.

So here’s what’s going to happen. We have some people with roving microphones. They’re going to be scurrying around, and if you want to ask us something, you can ask us now. We’re going to let it go on for a little bit, so let’s get the energy up. Ask fun questions. If you have more of a comment, maybe save that for an Apple review.

But we’re excited to hear what’s on your mind. And can I say, those of you who are wearing your Hard Fork hats, you look so beautiful. So good. What a memory you’re creating for me.

Okay, great. Now, does anyone have a question for the Hard Fork program? We have 1 right here, sort of in the middle. Vamp a little while she’s going up the stairs, Kevin. Nope. All right, we’re good.

Okay, this is a very hard question.

Yeah.

Last week, you introduced the nickname Chapatiti.

Chapatiti, yes.

Which I have been using frequently. And I would like to know what nickname you would use for each of the popular models.

Oh. Let’s see.

I’m going to let you handle that one, Mr. Improv.

Well, of course, I feel like Geminini is kind of right there. Bing-a-ling-a-ling.

Sometimes I do say Clauud, just—

Clauud.

—because I think it’s like—

Make it 2 syllables.

Yeah.

Thank you for asking that.

Yeah.

I’m serious. That was the right energy. Okay, what else would you like to know? Let’s see, I see 1 back here.

Oh.

Oh. Oh, I see 1 right here.

Who’s pointing here?

You know what? You point this one.

Okay, let’s go right there.

Hi. I love listening to you guys, so thank you. It’s a highlight of my week. I feel like you have a better, or a broader, view of what’s happening in the industry than maybe a lot of us do, so I might bring the vibe down a little bit. What are you the most concerned about when you look at these evolving technologies and how they’ll impact our lives?

I think I’m most worried about the pace of change. I used to feel like, when I left San Francisco, I was going back in time about 18 months to anywhere else in the world, just in terms of the stuff we have here. You know, the cars drive themselves, blah, blah, blah.

Now, I feel like that gap is opening up, and I’m really worried that society as a whole is just not prepared, either in terms of safety nets for people who may lose their jobs as a result of AI or other technologies, or just in terms of how much energy it takes to absorb this much change all at once. I’m not sure people fully grasp what’s happening.

And we’ve been around the block. I know that even the best-intentioned technological projects have unintended consequences. So we saw this happen with social media. All these companies said, “We’re going to save the world and connect the world, and everything’s going to be great,” and we’ve seen what’s happened with that.

So I hope that the folks in Silicon Valley, including some of the folks we’ve had on the stage tonight, are really thinking through some of the unintended consequences beforehand.

Yeah. To the extent that anything that happened over the past decade made you concerned about the ability of lawmakers to regulate technology, we’re now heading into an era where I do believe the technology is going to be even more powerful than Facebook, Instagram, and Snapchat.

We still have not answered a lot of basic questions about, for example, how children use this product safely, when they should start using it, and when they should stop using it. We ran out of time. I really want to talk to Sam and Brad about the fact that the same week that they’re signing a military contract, the same week that The New York Times is publishing a story about people experiencing these extreme mental health challenges with ChatGPT, they’re signing a deal with Mattel and saying, “We’re going to put it into toys.”

Now, I talked to OpenAI about that, and they said, “Well, we’re going to build toys for families first and kids 13 and up,” but there are just a lot of unanswered questions here. If I learned anything from the past 10 years, it’s that you unfortunately cannot count on lawmakers to do any meaningful pushback.

Honestly, 1 of the reasons why we make this show is because we like talking to you about problems that no 1 actually has a plan for. So we’ll keep doing that.

You call on the next 1.

Great question. I will call this person right here, sort of in the dead center. This will be challenging to get the microphone.

Just toss the mic.

No. This’ll be fun. Everybody can—how does the microphone get to this person? Okay. We’re going to pass it down, and that’s community. Beautiful.

Hi. I’m Hendrick. I love this show. I listen to it every week with my family. I’m a 13-year-old student, and I’ve always felt that I don’t see much AI in my classrooms or in my friends’ classrooms.

I was curious to get your guys’ opinion on how much AI you think we should have in the classroom and how we should integrate it, because I can see it from the teacher’s perspective: it’s plagiarism. There needs to be a balancing act. But from my perspective and from how I’ve used it, I feel like it’s actually really useful. So I’m just curious to get your view. Thanks.

Yeah, it’s a great question. I think this is really crucial. A lot of schools reacted to ChatGPT and other AI chatbots by just trying to stamp them out—banning them, detecting them. I think that was a mistake, and many of them have now come around to trying to say, “How do we teach with this technology and about this technology?”

I think AI literacy is a really important thing for schools to be teaching, because these are tools that people are going to be using in their jobs when they graduate and go out into the economy, if there still are jobs. I think it’s really naive to think that you’re just going to take it out of the classroom.

I do think that teachers and schools need to revamp the way that they teach to assume that people are going to have access to this stuff, and maybe use class time more for discussion and face-to-face interaction, and maybe assign homework differently. But I don’t think this stuff is going away, and I think the enlightened schools are the ones that are planning around its existence for the future.

Yeah. Thank you for the question. Thank you for listening. If I were 13 in this moment, the main thing that I would want to make sure I took away from school was the ability to think critically.

The challenge is, when I was in high school, I didn’t know how good was good enough at critical thinking. I do believe that the more you use AI as a substitute for reading and writing, your critical thinking skills are not going to develop to where they otherwise might. So I think that’s a good reason to make sure that you are actually writing the essays that are assigned to you and reading 1 out of every 3 books that’s assigned to you.

And I’ll say that while so much focus is on the school stuff right now, I think very soon the conversation is going to shift to AIs as companions. I have to say, I am so much more worried about folks your age who feel like ChatGPT is a better friend to them than anyone they know in their real life. That is such a dark path, and nothing is preventing us from walking down that path right now.

So, real friends over AI friends is what I would say to that.

Okay, great questions. You call the next one.

Let's go up here, the balcony. Do we have a mic up there? No. I picked the hardest possible spot. Okay. Oh, wow, you need the robot pants.

Thank you very much, Kevin and Casey. Love the show. You guys are awesome. Of any of the stories that you've had in the last 18 months, I think about this a lot: which one would make the best Black Mirror episode?

The best Black Mirror episode to me is just: you're trapped inside Italian brain rot. You have an objective to accomplish, but you have to get Crocodilini Bambolini on board. Yeah, that comes to mind for me.

Yeah.

How about you?

I think the AI friend stuff was—I liked it when you had the conversation with Turing, my AI friend. He couldn't be here tonight, by the way, but he sends his regrets. It was a configuration error.

Tragic.

So, yeah, that would be mine. Let's do a couple more. Casey, your call.

Oh, how about right here? We didn't realize how big this venue was until we started asking people to run around with microphones.

Like a T-shirt cannon—

Yeah.

—for microphones.

Hi.

Hi.

So you've had at least 2 different speakers tonight say that hardware is hard, and you showed us a partial, expensive exoskeleton. That's promising, but I really want the AI to get into a robot, a personal assistant, so that it can load my dishwasher, it can do the laundry, it'll fold the clothes, and it'll put them in the right drawer. Have you talked to anyone who has a realistic timeline and roadmap for this?

It's funny. One thing we knew about this show was that there was going to be a damn robot on stage, okay? And so we're thrilled with Kathryn Zealand bringing her rig. But there are a number of other companies in San Francisco who are working on stuff like this. One of them is called Physical Intelligence. They've raised billions of dollars over the past several years, and the reason is that you can actually take a large language model, similar to ChatGPT, put it inside a robot, and it's much more useful than it was before.

So that's kind of the reason that people think we're about to see this step change in function. But to your point, hardware is hard. I would say that in 2028, you'll probably still be folding your own laundry. But I could be wrong. I could be wrong.

All right. Let's go right here.

Less of a light question, but twice you mentioned that OpenAI has these new defense contracts, which to me is very concerning, especially when Homeland Security is grabbing people off the street. So I'm curious if you know more of the details of that, and then just what your guys' take is.

The way that these contracts tend to start is: there should be an AI assistant inside the Pentagon, and you could ask it, “What missile do I put on this jet?” That's real. That's sort of what they are. That's the foot in the door.

But to your point, yes, this goes to so many dark places so quickly. Facial-recognition technology is already basically perfect. You combine that with AI systems, and things get dark in a hurry. So I don't know what to tell you other than, yes, I'm super worried about it.

And I actually think that a role of journalism in this moment is just to point out what tech companies are doing with the military and writing about what I imagine are inevitably going to be some dark things.

Yeah, I don't love it either. I have some real qualms about it, and it's interesting: some of the people who were building this technology from the beginning were so concerned about the use of this technology for military applications that they actually started requiring clauses in their contracts and acquisition deals. DeepMind had a sort of famous clause in its Google acquisition deal that they couldn't use their technology for the military.

Now, they've since adapted that. But I think the tension here is that there's so much money to be made. The Defense Department has a huge budget. They're willing to spend lots of money to sort of catch up in AI. And, yeah, it really—I don't love it, and I'm glad that there are people inside these companies who are sort of resisting that pull.

All right, I see somebody right here.

Thank you so much. A longtime listener, and I just have to say, I hope you guys do this again here in San Francisco.

Should we do it again?

Yeah. And I definitely feel like I'm listening to you right in my headphones. So whatever you're doing on stage is definitely translating well.

But you can't pause, and you can't—

No, I can't.

—make us faster, which is great.

I was really hoping Sam Altman would be here tonight. I'm so glad you guys got him. One of the questions I have for you guys, and maybe in your vast experience of covering AI and everything that's come about, one of the important things to me is regulation. Is that going to be someone you feel comes forward from industry who kind of makes a bundle and then sort of sees the greater good, and then really sort of goes from poacher to gamekeeper?

I mean, that's going to be a real quantum leap, to my mind, to get Congress and people really educated into how things need to move forward for our safety. And whether it's the military or kids or anything else, it impacts so many different people in so many different ways.

I just thought, do you guys have someone in your mind who would be top of mind that you would nominate for that role today? Or do you see someone like Sam doing that in 10 years, 20 years?

I think we cannot look to the industry to lead this, right? I mean, Kevin, you asked about this during the chat, but where a lot of these folks start is, “Oh, no, the people building AI aren't safe enough. I'm going to do my own thing, and I'm going to do it differently and safer, and we're going to pass a lot of regulations,” and then they get really successful, and they're like, “I'm going to amend my comments about what I said,” right?

And so I think we just cannot sort of trust them to do that. There are a number of nonprofits who work in this space, like the Future of Life Institute, that have really mobilized around this potential 10-year moratorium that we're seeing. But I will just say: 10 years to have no regulation at the state level is an extremely long time.

If you believe anybody who's come on our show, in 10 years that will be sort of the end of the ballgame. Very powerful intelligence is here. So what I hope is that as people learn more about this stuff, they push more, and I think we're going to need to see some sort of bottom-up public participation here because it's not going to come from the industry.

All right, next. Yes.

Love the show. It's been said over and over. And several people you've interviewed on this show are technology leaders, and there's a lot of techno-optimism. I think a lot of it is well warranted. But one area where I'm not buying it is the idea that everybody will reap the benefits of the wealth and the monetary benefits of this.

I understand how the people who create the AI reap the benefits of the wealth, but I have never heard anyone talk about how that trickles down or gets redistributed, and I'm wondering what you've heard. If you've heard anything else.

This is such a good question. This drives me insane. When people just come on the show and they say, “This is all going to trickle down. We're all going to have radical abundance.” And it's like, that is a deliberate policy choice that many people do not want to make. That does not happen automatically.

So I think if I were running these companies—which, thank God, I'm not—but you actually do have to be out there advocating for the kind of world that you want to see. It's not just enough to build the technology. It drives me insane. I'm just going to get this off my chest. I'm popping off tonight.

Preach it.

People are always comparing this to the Industrial Revolution. During the Industrial Revolution, there was a 50-year period called Engel's pause, where productivity and profits went up but wages did not. So workers literally for 50 years did not see the benefits of the Industrial Revolution.

What worries me when people say, “This is just going to be like the Industrial Revolution,” is that they have not actually read their history. It sucked for a lot of people for a long time, and I want them to know that.

Yeah. Well said. All right. I haven't really called on anyone over here. I see a person at the farthest extreme of the theater, and I would like to—

We just want to make sure our mic runners are getting their steps in.

Yeah, the steps are in.

It’s very important.

The steps are in.

Vamp, Casey.

Oh, we’re right there. Okay, great. Right here. Yeah, over here.

Thank you so much. Love the show. I’ll say it again. Longtime listener. So, Casey, in particular, I think you had mentioned a couple of weeks ago that you had a really clear picture of your definition of AGI.

Yeah.

And I wanted to hear more about that. And Kevin, I’m interested in yours as well, but Casey, it seemed like you had a picture of it, so I want to hear it.

Yeah, and I want to say that I’m not necessarily advocating for this as a perfect outcome, and I’m not rooting for this to happen as quickly as possible. But I work with an assistant who is amazing. She helps me schedule things. She helps me sell advertising. She helps me help you with customer service if you’re a platform or subscriber and you need to change your email address.

So when people say AGI to me, I think, “Oh, it will do that.” Do you know what I mean? And I think everyone in this room has their version of that. There is a subset of things that you do that are mostly routine, that feel a little bit like drudgery to you, that are not the sort of creative part of your job that’s exercising all of your human muscles. And if 0789 can just kind of do that layer of things, that would be AGI to me.

Now, I love my assistant. I want to keep working with her. So one question would be, well, is there some new set of tasks that she could work on? That’s to be explored. But when everybody is like, “AGI, it’s just a marketing concept, it’s so fuzzy,” it’s like, no, it would just do what my assistant did. Does that sound crazy?

A little bit.

Okay.

You did talk about having an assistant a lot. Let’s do one more. Let’s do one more, and let’s do it right up there. Yes.

Okay, I got it. So, gentlemen, last week, standing in my kitchen, chopping vegetables, listening to you earnestly try to give PR advice to Mechanize, I almost cut my finger off because I was laughing so hard.

Lose a lot of fingers that way.

So—

So I want to give you the opportunity to give PR advice to 4 or 5 other companies that you think might need it right now.

Give PR advice?

Yeah.

Let’s see. Who needs it right now? Meta could use some PR advice, and here’s why. They think that when you go around saying, “We’re offering people $100 million,” people are going to be like, “Wow, that’s so cool,” when what it really means is, “You would have to pay me $100 million to work at Meta.” All right. I think that’s everyone. From the bottom of our hearts, thank you. Thank you! That was Hard Fork Live. Thank you. Thank you. Thank you, everybody. You’re the best. We did it. Hard Fork is produced by Rachel Cohn and Whitney Jones. We’re edited by Jen Poyan. We’re fact-checked by Caitlin Love. Today’s show was engineered by Chris Wood. Original music by Mariel Lozano and Dan Powell. Video production by Sawyer Roquet, Pat Gunther, and Chris Schott. You can watch this whole episode on YouTube, and you should, at youtube.com/hardfork. Special thanks to The New York Times Live Event Team, Hillary Kuehn, Beth Weinstein, Caitlin Roper, Kate Carrington, Chantal Renier, Melissa Tripoli, Natalie Green, Angela Austin, Kirsten Birmingham, Marissa Farinia, Jennifer Feeney, and Morgan Singer. Thanks to everybody at SFJAZZ, and thanks to The Brass Animals, our live band. Also special thanks to Matt Collette, Paula Schumann, Pui Wing Tam, Dalia Haddad, and Jeffrey Miranda. You can email us at hardfork@nytimes.com, but you should know that we’re on vacation right now.