AI Is Scaffolding
AI Is Scaffolding
Does the real promise of generative AI lie in the chatbot? Paul and Rich don’t think so. Building off a post by tech entrepreneur Dustin Moskovitz on “scaffolded AI,” they discuss extreme visions of our AI future and position themselves in the center—where tools work with the user, rather than attempt to replace them.
Read the transcript
- Threads officially draws a distinction between an individual “post” and a “thread.” So here’s Dustin Moskovitz’s original POST that arguably turned into a THREAD.
- Paul’s most recent WIRED piece: “Generative AI Is Totally Shameless. I Want to Be It.”
Paul Ford: Hello. I’m Paul Ford, the co-founder of Aboard.
Rich Ziade: And I’m Rich Ziade, the other co-founder of Aboard. Good day, Paul.
Paul: Good day, Rich. You and I are actually headed to an AI conference as we are recording this.
Rich: That’s right. No humans, just robots, rolling around.
Paul: Happy little robots. So, Richard, what is Aboard?
Rich: Yeah, Aboard is a software platform that helps you manage data. It sounds very dry, but most people do it every day, at work, at home, whether you’re keeping lists of stuff for your business, or keeping track of things you need to get done. And Aboard is a really lightweight tool that’s really smart about managing your own information, collaborating with others. Check it out at aboard.com.
Paul: And one thing about Aboard, which we’re going to talk about today, you might have heard this if you’ve listened to the podcast or seen the website, is that it is AI-powered, which some people hear, and they go, “Oh my God, throw it out the window.”
Rich: [robot voice] Another robot!
Paul: Other people hear and they go, “Oh, thank God, we’re in the future.” The way we use AI is really specific, which is we use it to help get you started. We use it to help set up the data. You give it a prompt, and it says, “Hey, you could organize your world like this.” And we don’t expect to get it right, actually, we expect to get it kind of right. And then we count on something called the unbelievable power of mild human annoyance, where people go, “No, that shouldn’t be called claims, that should be called requests.” And then they can easily go in and rename it, and they can organize stuff.
So we let the AI be a slightly annoying little brother robot to get you going so you stop talking and get working. And I’m proud of that for us. I think we did a good job with that. I like that part. That’s it. Let’s do it. This is the Aboard Podcast.
[intro music]
Paul: All right. So, Rich, you were on threads.net. I think you were one of the three or four users of threads.net. And…
Rich: No, I think it’s a growing social media platform that’s extremely PG-13. I’m enjoying it.
Paul: It is. I’m a Bluesky person and a Mastodon person. I’m on all of them, and I’m on them less and less. And regardless of that, let’s talk a little bit, you found an interesting Threads post, and it’s about how AI is not the main event, but actually scaffolding that helps you get to the main event. We should talk about what the main event really is there is.
Rich: Entrepreneur named Dustin Moskovitz. Do you know who that is, Paul?
Paul: He was very, very early at Facebook, so he’s one of our—
Rich: That’s a good place to be. Right place, right time, right there, my friend.
Paul: He’s one of our younger billionaires.
Rich: Aw. [laughing]
Paul: That’s right. I think also—yeah, he also did Asana. He created Asana for your to-do list.
Rich: Asana’s, like, collaborative work software with to-do lists and whatnot.
Paul: He’s a big charity guy.
Rich: Yes. I think he’s part of, like, what is, he’s gonna give away all his money. Giving Pledge?
Paul: He’s Giving Pledge, yeah.
Rich: Which, I gotta say, it takes a lot when you have $5 billion to say, I’m only gonna keep $200 million.
Paul: Yeah, I know. That’s an incredible act of bravery. No, good for him. Good for him. That’s right. He’s also a big Democrat guy.
Rich: Okay.
Paul: Okay?
Rich: All right. Well, he put out, he put out a—I don’t know what you call a post on threads. A thread? [laughing]
Paul: I think it’s a thread. Could be a post. The thing is, is nobody cares.
Rich: Nobody cares.
Paul: Nobody cares.
Rich: All right. Let me read it out loud.
Paul: Should people brace themselves for, like, a four-minute Threads post? Because those can go on.
Rich: No. It’s two short paragraphs. It won’t be too painful.
Paul: Okay, go ahead, read it out.
Rich:”One of the big issues with the chatbot paradigm is they invite the public to think of AI as “working” or not in a binary way. You do a prompt, you get a result, and you decide if you like the result. I’m not surprised many are having bad experiences.
“In contrast, scaffolded AI like Github Copilot or Midjourney is engaged in a collaboration with the end user, doing pieces of the work (sometimes large pieces) but by no means expected to intuit all the intent/give you something finished one shot.”
I think the dream—you know, our noted philosopher Stanley Kubrick, the thinking, prescient, sort of all-knowing robot being is something we’ll always reach for. It feels like science fiction is evidence of that. We’re always going to kind of seek it out.
Paul: We come up with new names for it every now and then. So we’ve got, we had the singular—first, we just had artificial intelligence, but that didn’t happen. That started in the fifties. Then around the nineties, the idea of, like, we’re going to get so much computing power that we’ll have, it’ll all become super conscious in a minute. That’s called the singularity.
Rich: Mmm hmm.
Paul: And now we’re really big on, like, we’ve commoditized the singularity, and we call it AGI, for artificial general intelligence, meaning, you know, a computer that’s kind of indistinguishable from a human. This fantasy goes back, I mean, you know, to literally Mechanical Turk by Amazon is named after a chess-playing robot that actually had a guy inside, wasn’t really a robot. So we’re always playing with this idea that the machine, the golem, could come to life.
Rich: Right.
Paul: It’s very exciting. It’s very like, it really, I had a friend, a good friend of mine yesterday was like, “Hey, I want you to come on my podcast and maybe explain to me why, how scary artificial general intelligence is.” And I’m like, that’s, like, number 50 on my list of things that I, as an incredibly anxious person, freak out about.
Rich: [laughing] There’s a lot of scary things in the world.
Paul: There’s really scary stuff in the world. I’d put that one way, way down no matter what, Sam Altman’s got those big eyes, he goes in front of Congress. But, like, that is not… [laughter] But, you know, you make a thing, and it’s just like, you’re like, “Hey, are you an adorable baby that could grow up to murder everyone?” And it’s like, “I will never murder everyone. Everybody relax.” The fact that it says that really freaks everybody out. Like, it just, this whole thing is an artifact of us being weird chimpanzees that like to project our chimp nature onto absolutely everything, including rocks. Did you ever see the movie Everything Everywhere All At Once?
Rich: I did.
Paul: Critical scene in the film is two rocks talking. And as a movie watcher, you’re—one of the rocks, has googly eyes on it. And as a movie watcher, you have no problem with this. You’re like, oh, cool, the rocks are chatting.
Rich: Well, this is Academy Award material right here. [laughing]
Paul: And I actually don’t think that what’s happening with AGI is much different than, like, watching the movie where the rocks talk. Like, it’s just, the rock is, you know, it’s sort of like the robot is repeating things. The term that I like is stochastic parrot. [laughter] Like, it’s, it’s using a statistical model to kind of say stuff. Here’s what’s real: It’s a very loose model of sort of the domain of all human knowledge. And you, turns out you can do things with that, just like you can do things—you can do amazing things with just the names of Wikipedia pages and the links between them. You don’t even need the content, right?
Rich: Yeah, yeah.
Paul: You can take the data from Wikipedia, and you can make a model of human knowledge, and you can see how things are connected. And we’ve been doing stuff on the internet like that for 20 years now, 30 years now. So when everybody’s like, “Oh my God, it’s going to become intelligent and it’s going to, you know, it won’t just schedule your dentist appointment, it will actually murder everyone in the world.” I’m like, okay, cool, good. Go think that.
Rich: Yeah. Let me ask you this. I don’t mind it, because it creates such excitement.
Paul: It creates fear, though. I don’t, like, I feel that we live in an era in which we just, we should stop creating fear all—I wake up in the morning and I get my dose of incredible fear.
Rich: Yeah, well, maybe you shouldn’t grab your phone.
Paul: Okay, regardless. Maybe I’ll grab my phone. Maybe I—but I might look at it on the train on the way to work.
Rich: I guess what, what I mean is it’s, it’s less about the fear side of it, it’s more about the, like, enthusiasm and excitement side of it that comes out of the Valley, right? Like, which is like, “Here we go! It’s the next wave!” And they can’t help themselves, right? They can’t help but trumpet that this is, this is going to change humanity for the better. I got to say, Altman isn’t, to me, I don’t know anything about his, like, credentials. He’s obviously a very thoughtful business leader because he’s found success after success. But he strikes me as an absolutely world-class spokesperson, because he’s, he’s optimistic. He’s got those doe-y eyes. He just kind of looks at you and says, “Well, we’re going to just make the world better.”
Paul: No, you know what he is? He is the, and this is a very specific thing. He’s not the next Jobs as, like, an entrepreneur, business leader. Although then again, he did get thrown out of his own company, and he would manage to accelerate the return—
Rich: Eerily, yeah. [laughing]
Paul: But he doesn’t have—Jobs actually did have a deep technical knowledge, like, down to, like, soldering. I think Altman does, too. Like, you know…
Rich: Yeah.
Paul: But he is like the, he’s like that next generation of, like, pure value creation. He was Y Combinator. He evaluated zillions of startups. And onward he goes. It’s tricky because Steve Jobs is now a saint, like a capitalist saint, so, but he, he is the next Steve Jobs in the kind of, like, “I will look at you earnestly and tell you I absolutely believe that a revolution is coming and that it’s absolutely in your best interest to buy our products, and I will believe it top to bottom.” And—
Rich: [overlapping] You believe him.
Paul: The ability of—well, to do that at a scale, at a, like a trillion-dollar scale, and to tell the world, like, if I went out and I said, I need to raise $7 trillion for Aboard, the future of data, people would not take me seriously.
Rich: I don’t think they would. Paul. You’re great. You’re a great speaker, a great writer, but they wouldn’t take you seriously.
Paul: Even if I said $7 trillion for the data industry, they wouldn’t take me seriously.
Rich: Yeah.
Paul: This guy is able to float stuff where a portion of the community goes, “Yeah, I mean, look, he’s out there, but he’s got a point.”
Rich: Yeah, yeah, yeah. And look, world leaders, when you start meeting, like, world leaders so that their sovereign funds can invest, you kind of know you’re on your way at that point.
Paul: Yeah, that’s true. When, like, Qatar is, like, stopping by.
Rich: [laughing] Like, six lime green Lamborghinis pull up to…
Paul: Yeah, now, you know what it is? It’s the, it’s the plane. It’s the 7, like, 747 with four seats on it. Yeah.
Rich: It’s, like, refitted.
Paul: Yeah.
Rich: Okay, so let’s point one. Excitement and enthusiasm around new tech is a thing we’ve seen over and over again. We’re in the middle of it. Now, let’s contrast that with what Moskovitz calls scaffolded AI. I use the word scaffold a lot in the office, actually. How do you contrast scaffold AI—scaffolded AI. Which, let me tell you, if you really want to, like, not excite people, use the word scaffolding.
Paul: Well, this is our world, right? Our world is the boring stuff. You and I like the boring stuff. We like the infrastructure. Scaffolding is also a very New York City thing, because it actually is, about 80% of the architecture of New York City is expressed as just pure scaffolding that’s left up for for 5-10 years.
Rich: Non-New Yorkers: New York City is permanently adorned with scaffolding. It’s just everywhere. It’s everywhere all the time. Years, you start to kind of see a building with its scaffolding as the permanent state of the building. Like, you just assume it’s going to be there forever. It’s not clear why it’s up there for years, but it’s up there for years.
Paul: Well, and the ones that are put up for construction, those go away because they need to sell the building like they need to sell the units.
Rich: [laughing] Right. Yeah.
Paul: So they’re like, tidy this up. But after a building is like 75 years, and they’re like, oh, we got to repoint the brick. They will drop that guy up. And there is at that zero pressure for it to come down, and it’s expensive to bring it down.
Rich: So what does that mean to you? I know what it means to me, but, you know, he’s trying to make a point that, hey, stop trying to look for, like, some magic-bullet answer every time you talk to AI. Scaffolded AI, which is, frankly, collaborating with the user or giving the user a head start, is kind of where a lot of innovation is happening right now, but nobody’s talking about it because it’s frankly less exciting. It’s less anthropomorphized.
Paul: Well, look, here’s the, here’s what’s been sold. The thing that has been sold is that you no longer have to do anything. You can say, “Write me a college term paper about this complicated subject,” and it will do a really good job for you. And the fact is, it doesn’t really do a good job, but if you don’t know what it’s doing, it doesn’t matter. You just like, you turn it in anyway.
Rich: It’s extremely confident. [laughing]
Paul: That’s right.
Rich: It speaks with authority.
Paul: I wrote about this with, in my last piece I did for WIRED that just ran. It is, it’s shameless. And I actually kind of admire AI for just being baldly shameless. It just gets out there and it says whatever the hell it wants. It’s the world’s greatest salesperson. And in a world like ours, there’s actually a place for that. The problem is that the place, the ultimate place for these kind of chatbot-driven things is like the direct email that you get telling you that there are a lot of people who would like to write code for you in another country, and it just keeps sending you email after email pretending to be your friend.
Rich: Yeah. Yeah. Yeah.
Paul: Right? So I think it’s like, it’s text generation and it’s image generation and it’s sort of like you’re going to personalize around you. That is a very specific domain. And frankly, it’s one that all of—nobody really wants to consume this stuff, with a few exceptions. I think we will play video games where the non-player characters are capable of having sort of richer conversations on guardrails because that’s fun, right? And I think there—
Rich: Or you won’t know, you won’t know where the plot goes, potentially.
Paul: That’s right.
Rich: Which is interesting.
Paul: I think that future curricula will include people doing first drafts with AI and then editing them, you know, with the professor, and like—
Rich: Ah, sounds like sca—okay, that’s interesting. AI is a first draft is a really good way to look at it.
Paul: That’s right. And I think, so let me get to scaffolding, right? Like, scaffolding goes up. I think there’s a specific kind of scaffolding, which is, it’s the kind that you put up when you are building something inside of it, not the kind that, when you’re, like, fixing the building. Right?
Rich: Yes.
Paul: And so that’s what I think AI is. AI is, I’m going to build this sort of very light protective shell that gives me a sense of the shape of the thing inside. What was our thing? Our thing with Aboard is where you would open up a Google sheet to kind of create some spackle in between a couple different systems and organize a little bit of data, and like, Sally has to send that spreadsheet every Tuesday to Mike, and then Mike is going to put into the system, if you go to companies, there’s a million spreadsheets like that floating around. We’re like, well, what if we could make that into good workable software? And it turns out that AI is a little bit of magic trick in that it will create a lot of that scaffolding for you to solve those kinds of problems. It will write the first draft. It will give you a list of things to research and learn.
Rich: Yeah.
Paul: You have to, like, if your vision of humans is that we’re going to sit in pods and consume, then that isn’t very interesting to you. If your vision of humans is, or if you believe that all things that are done inside of large corporations or inside of organizations are actually wasted and we should just live on farms and look at the sky and touch a fox.
Rich: Also not for you.
Paul: That’s not for you. But if you’re in a place in your world where your big not-for-profit or for-profit company has to organize a list of people who need to get a mailing once a month about progress, setting up that list is actually can be like weeks of work, and it sounds bananas. It’s like, it’s a list of people. No. You know, because this person needs them to do this and this person needs to do that. And so we all become, in our modern sort of data-management world, we all become knowledge workers and data modelers. And I feel that AI is just really, really good at, like, some of that stuff.
Rich: Yeah. I think that relationship with AI and that relationship, frankly, with technology generally, the idea of not just purely consuming, but using it in a way makes it additive. I think there are two, two big benefits here to thinking about AI differently. One is, yes, it can save you time. It can, instead of you researching for three hours, it’ll skip a few steps and provide some convenience. But there’s another, I think, interesting benefit, which is it’s by default a little creative.
Paul: Yes.
Rich: Like, when you tell it to try again, it tries again, and it tries again differently. And I think a lot of discovery for humans, a lot of forward motion is, like, lateral thinking is kind of baked into AI. Like, in a weird way.
Paul: It adds interesting friction, right? And that is actually what motivates humans to do things. If you give a human a task, it’s very hard for them to do that task. I learned this as a manager. You say, “Hey, can you do this?” And people simply are like, “I can’t,” right? [laughter] They don’t say that to you, but they simply cannot do it because you haven’t given them an on-ramp.
Rich: Yeah.
Paul: You haven’t given them a process. If you just—and so you throw that over the wall and they come back and they’re like, and they just haven’t made any progress, and then you sit, you, you get empathetic and you sit down, you’re like, “Well, here’s how we usually would start one of these.” And I actually think that, like, what AI does is it’s going to show you the form of the thing. It’s going to be like, here we go. And you’re going to go, oh, boy, that feels wrong. And then you have to figure out why it feels wrong, because AI can’t tell you.
Rich: It’s a dialogue, in a lot of ways.
Paul: Exactly. That’s the dialogue.
Rich: It’s also relying on you as a thinking partner.
Paul: So I think we’re getting, we’re getting to the point here, and I want to throw this back to you. We’re getting to the point where, like, all right, maybe don’t worry about it so much. Don’t, it’s not coming for your job right away. It’s not, blah, blah, blah. Okay, so just, like, give it a minute. You’re gonna, you’re gonna be, I think you’ll actually have more time to get your bearings than Silicon Valley would have you believe. I think it’s gonna be years before this stuff truly shakes out.
Rich: Yeah.
Paul: Like, I don’t think we’re gonna have AI pop songs in the top 20 between now and, you know, 2027. I think we got a minute.
Rich: Yeah.
Paul: So you’re saying scaffolding. This is good scaffolding. We have our product. Good for us. But, you know, you’re listening to this podcast, and you gotta tilt your career in a certain direction. You gotta tilt your life in a certain direction. You just said scaffolding. What do you do now? What do you do with this exciting information that Rich and Dustin Moskovitz have shared, that this is a way to organize things, and get things started?
Rich: I’ll tell you what you do. I’ll tell you what you do.
Paul: Oh, there we go. Little Bay Ridge came out in that.
Rich: I’ll tell you what you do.
Paul: I’ll tell you what you do.
Rich: Jimmy. Listen to me.
Paul: Ey! Jimmy!
Rich: Listen to me. By default, if you think a piece of tech has shown up so you don’t have to use your brain and you can instead take a walk and get bubble tea, just hang it up right now. I have little kids. Everything is, they talk to all sorts of boxes around my house, and the boxes respond with nice British voices, and everyone’s losing their minds.
Paul: We have a very good friend whose kid’s first word was essentially Google. Because it was, Hey, google—
Rich: [laughing] Hey, Google! yeah.
Paul: Hey, Google.
Rich: That’s right, that’s right.
Paul: So her first word was [baby saying “Google” sound].
Rich: That’s right. Look, and I’m not gonna, I’m not gonna—I love tech. I believe in it. I believe in what it can do for us. But if you see it as a wedge that can open you up a little bit and help you go down a certain path and you learn something, then, wow, what a cool tool. Like, what a magical tool. If you see it as, like, I’m going to hit a button, write my English essay, and then go have brunch with my friends, then God bless. It’ll do it for you. It’ll work all day long.
When I think about scaffolding, I think about something that is temporary, that is a means to an end, and is a stepping stone to something much better. Right? Because nobody says, oh, what do you do? I build. I designed some of the best scaffolding you’ll see in town. Nobody says that. It’s temporary. It is something you build on top of. And the thing that you’re building on top of is yourself, is you. And I’m sounding a little corny here, because I think that’s the mistake people often make with this stuff, is they just assume you can skip the day.
Paul: Here’s where we’re. We’re trapped between two narratives. Okay? Narrative one is, all of this is evil, and it’s, it’s going to ruin people’s lives, and the tech industry is very evil, and it’s purely exploitative and extractive. And there are many ways that you can sit down and make that case. And on the other side, it is a bold new utopia is emerging, and it will, we will—in fact, it’s going to be so powerful that we’re going to have to just give everybody money because the computers will do all the work. And here we go.
And those are both extremely disorienting narratives, especially as the real world of humans keeps kind of asserting itself. And I think there is—look, there’s a case to be made that, like, AI stuff plus political advertising is an absolute disaster, right? Because you can flood the zone in ways that you never could. You can scale stuff up. There’s a case to be made that we can optimize a lot of stuff and make a lot of jobs a lot better and make things more interesting for humans. Somewhere in the middle is where we are. Somewhere in the middle. And it’s kind of up to us, which is always awkward. I think we’d love to have a little customer service here and be like, [laughter] this is evil, this is good, and here we go. And it’s kind of my job and your job to do that. I got to tell you, we’re just in the middle. We are not at the evil crisis point of the technology. You know, a good example of evil crisis point would be, like, when social media really started to melt down. We’re nowhere near there with this stuff.
Rich: No.
Paul: We’re just getting our—we’re putting our shoes on.
Rich: That’s right.
Paul: Scaffolding, though, I agree with you, I don’t think the future is chatty little chatbots. I mean, I think they’ll be here forever. I think you’ll always have one if you want one.
Rich: Yeah.
Paul: But I think the future is, let me draw, let me sketch that for you to show how the—or let me show you how you might draw that house if you were doing it with a pencil. I think humans still need and want to do things, or otherwise we get real squirrely.
Rich: Well, we create problems.
Paul: Yeah.
Rich: We don’t just get squirrely. We literally look for problem. We create issues. We create friction.
Paul: [overlapping] We do. We’re sure doing it right now. And so I think, like, AI is not the way out of that pickle of human behavior. I think it either is a helpful little friend or you’re making a mistake. So I think, like, if you’re going for scaffolding here, figure out how to make this thing actually get in with it. Let it help you with a few processes, but don’t think you throw stuff over the wall. Figure out how it can help you do stuff throughout your day. That’s its job.
Rich: That’s a negotiated stance. That’s not, like, stay away. You can’t use it to write a paragraph. It’s like, no, use it to write. But read the paragraph. Rewrite the paragraph. You build the building. It’s your building. AI is good for scaffolding to help you build your building.
Paul: That’s what you are, the one. You’re doing the communicating, not the robot.
Rich: Exactly. Exactly. And also writing—honestly, I can say, I’m an attorney. I’ve read some of the most convoluted texts on this earth, going to school and reading contracts. Writing is one of the hardest things in the world. I really believe that. Like, I think it’s harder than programming. Writing well is hard. It’s so hard.
Paul: Yeah, it is. It’s harder than programming. I gotta tell you. [laughing]
Rich: It’s harder. Yeah. I mean, Paul Ford is saying it’s harder than programming. It’s hard. It’s hard. Writing well is hard. But, boy, what, what a beautiful thing to be able to go on that journey and become a good writer. Like, don’t let that be taken away from you and, like, you can use it to help you get there. I want my kids to use AI, this isn’t a, like, stay away. Don’t let it write anything. Don’t ask it que—I installed ChatGPT for my, for my son. Like, I want him to learn things.
Paul: Mmm hmm. Mmm hmm.
Rich: I don’t have a problem with it, but I want him to see it as something that enhances where he can go more than anything else.
Paul: It’s just software. [laughter] You know, people, when word processors came out, everybody was like, oh, you’re going to get in the way of the important work that people do with pen and paper. Here we are. All right, so, Richard, if people want to get in touch, what do they do?
Rich: They reach out at hello@aboard.com. This is a part of our mission, Aboard is a tool that enhances how you work. Check it out at aboard.com. We love questions. If you have podcast topic ideas, if you have questions about the product, we’d love to talk. Give us five stars. Do people still do that, Paul? They still still give podcasts five stars?
[outro music]
Paul: No, they get chatbots to do it for them. Anyway. Hello@aboard.com, we’ll talk to you soon. Everybody have a great week.
Rich: Have a wonderful week.
Paul: Bye!