John Maeda in Conversation with Dean Anijo Mathew

John Maeda’s conversation with Anijo Mathew was one of more than 10 events presented by Chicago Humanities and the Illinois Institute of Technology on May 18, 2024. Following is a transcript of their talk. 

Phillip Bahar:

I can think of no better place to have a conversation about the future of technology and AI than at that great institution that was so central to The New Bauhaus and the founding of innovation and technology here in Chicago and much more broadly.

Philip Bahar:

Before we begin, I want to mention our partners. Our presenting partners are Builders Vision, Friedman Properties, and the Robert R. McCormick Foundation, and our spring season sponsors, the Richard H. Driehaus Foundation and Nielsen IQ, who also underwrote a whole series of programs on artificial intelligence throughout this year, thinking about AI through the lens of humans and the humanities. And I want to thank all of you who are members and donors, you really do help bring these programs to life all across the neighborhoods of Chicago. So if you’re not a member and you love what we do, please consider joining us. Philanthropic support covers about 60 percent of our annual budget.

So a few words about our speakers.

John Maeda is the VP of Artificial Intelligence and Design at Microsoft. Such a privilege to have him here at this moment in time. He’s a celebrated technologist. He previously held leadership positions at Publicis Sapient, Automatic, and was a tenured research professor at the MIT Media Lab, one of the great centers of innovation here in America, known for its cutting-edge, multidisciplinary approach to design. And he has led pioneering work in computational design, coding for artists and human-centered AI. Esquire named him one of the 75 most influential people in the world, and across his lifetime he’s championed the vital role of artists and designers in making a better world.

His books, like How to Speak Machine, he’s considered how to leverage AI’s potential, while keeping human value centered. Today he’s joined by Dean Anijo Mathew here from Illinois Tech at the Institute of Design, who is focused on entrepreneurship and urban technology. His research evaluates new models of innovation, enabled by technology and media convergence, and he works with global organizations to adapt and implement change in a rapidly changing environment.

We’re going to start the presentation tonight with a few comments from John, and then Dean Mathew will be joining him. So thank you so much for joining.

John Maeda:

Hello everybody. There is a clicker somewhere. Here it is. Wow, it’s so nice outside. It’s so good outside and you came here, but you get that window seat. It’s expensive in the airplane. Okay. All right, let’s see here. How are y’all doing here? Doing good. Looking in the faces here, everyone’s good. It’s good. Okay. You always have your phone in case I get boring. It’s all right. Good.

Right. Well, I’m very honored to be here, specifically because I’ve had the opportunity to sit with Dean Anijo up there. This looks like amazing. Who knows Dean? Who knows a Dean Anijo? There we go. There we go. Oh, fans here, right? Little applause please for a moment.

It’s very interesting, carbon life form, very filled with brimming with stuff, I’ve learned so much. I feel like I won the lottery in terms of being re-tuned for this design, business, social activism, Chicago flavor that I really, I haven’t been able to taste before, so I’m pretty excited. I live in Seattle, Washington, and anyone knows Seattle? Seattle? Very rainy. Rainy. Yep. And I moved there because my parents are in their late eighties, and any caregivers out there? Caregivers? Yeah, hard. Oh, caregiver’s hard. So I had to move out there because there was a situation. Anyone have siblings you can’t trust? Situation. So you know what I’m talking about. So I had to find a job, and they live in Redmond right near Microsoft. So I had to find a job, preferably at Microsoft, and it was October of 2022, and I joined Microsoft because the person there who was saying, “Oh yeah, maybe we have a job here for you.”

I was playing with OpenAI Beta and he said, “Oh, if you come here, I might show you something.” I said, “Oh, what is that thing? I don’t know what it is.” So it was October 2022, and I went in to do AI stuff, and all my friends were making fun of me like, “Oh, John. Oh, that’s really embarrassing. You’re working on AI at Microsoft.” And then in November of 2022, this ChatGPT thing came out, and suddenly all my friends were like, “Oh, John, good idea.” You never know your friends. They kind of flip or whatever, so anyways. But I didn’t know that was going to happen either.

So anyways, I have been lucky to be really at the center of a lot of learning, because I worked on AI in the 1980s. I was at the MIT AI lab. And it was right near the end of the AI boom, the so-called AI nuclear winter. All that is saying it’s a very dramatic way to say nobody wanted to put money into AI. That AI could do something called expert systems. It’s like if this, then this, if then this, and this pile up, because it ends up if and thens. Idea didn’t work very well, it turned out. A lot of people invested in it. Didn’t do that amazingly. So eventually it dried up.

But this new kind of AI is very different. This new kind of AI, has anyone used this kind of stuff recently, touched this kind of thing? It’s different. And I want to go over what I’ve learned about it because it may help you navigate what is happening right now.

John Maeda:

So I used this phrase, to frame a report I give every year, called the South by Southwest Design and Tech Report. It’s kind of a double design against AI, designers against AI, or is it am I designing against AI trying to compete? It’s got a duality to it. It’s also got the word, little letters AI in there. I love little word puzzles like that.

And I have different versions of this, but I want to have the conversation with Anijo, which I think is exciting. So I’m going to do roughly 13 minutes of this. Okay? Here we go. Rapid. Okay, here we go. As if you know how younger people have realized you should go up to the upper right and set 1.5x? I don’t have to talk at regular speed. All right.

Okay. So first off, someone, my colleague said to me, this new kind of AI behaves unpredictably. That is not true. It is predictable in how it works. Because it turns out we human, remember the Donald Rumsfeld matrix, unknown unknowns? Well, there’s an unknown unknowns, known unknowns, unknown knowns. Those three quadrants, we humans don’t do very well with. Ambiguity, lack of information. Oh, humans, no good. Where do we do well? We do well on the known knowns. What else does well there? Large language model AI likes known knowns. So it’s predictably, it’ll fail in these other three quadrants the same way we humans do.

And I like how there’s a boom now where people are saying, “I’m going to get AI to take my prompt and make an entire website from it.” Ooh. On the other hand, you have a customer who says, “This website’s too complex. AI, make it into one button.” Do you see this weird thing? We over hydrate the idea, and then we desiccate it. Does that make sense? No. A little questionable.

Okay, Conversational Design. Chat, not new, this book by Erika Hall, she’s coming out with a second edition, very timely. Erica points out that language is the oldest interface. If you’ve ever taken care of a baby, they want something, they want milk. What are they using? They’re using a voice UI. Milk. So it’s the oldest interface to get something done.

And then this is a book in 1967 by my former boss, Nicholas Negroponte. It says here in purple, you could build a predictive model of your conversational performance. Such a machine could then reinforce the dialogue by using a predictive model. 1967. That’s totally spot on.

But most importantly, this is a book by my AI professor. In the eighties I took an AI class, and I’m sure some of you know that when you’re younger, you’re kind of stupid. So I was like, I had AI, and then I had the instructor that, I liked the other cool one, and there’s this other guy, his name was Joseph Weizenbaum, and I don’t know, but I skipped class too, so very bad student. Years later, I realized he’s super famous, because he invented the first chatbot. In 1966, he invented a chatbot called ELIZA, which emulated what’s called a Rosarian psychotherapist. It’s very simple logic. 1966 ERA software. It did two things.

First thing is it was very good at saying back to you what you said, like, “I had a bad day.” “Oh, you had a bad day.” You’re like, “Oh my gosh, you hear me?” Right? The second thing it would do, it would look for keywords, like mother, girlfriend, boyfriend, whatever. So I was like, blah, blah, blah, blah, blah, mother. “Oh, tell me about your mom.” So just those two rules enabled this program to eerily feel like a human being to another human being. 1966. He was very concerned about this.

And in this book, which is available on the internet and PDF, if you read it, the first entry says, what I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people. 1966. He was in the New York Times warning people about AI, because he as a youth fled Nazi, Germany, and scientists of that era were very concerned about things that can be used to weaponize bad things. So he didn’t go out into Zuckerberg it. He spent all his time in the humanities to tell people. He didn’t go computer science, he went total liberal arts, and so a very unique person.

And actually, as we were talking over there, me and Peter and Anijo, I realized I want to give you a demonstration.

Do you see eyes? Do you see eyes on this thing? Eyes? Do you see eyes, eyes? Hi, how are you? You’re looking at me? Look over here. That’s weird, right? It’s because we’re wired to go to where the voice or character seems like it’s coming from. So when we have something talking at us and gets us and engage the conversation, we enter a delusion that there’s a human being on the other side. And maybe you can tell your friend, now this is all math, they say, “No, you got it all wrong. There’s a soul behind that.” That is the fundamental problem when you use this kind of AI for chat, you’ll fall into this delusion.

So a lot of what I do every day is I don’t chat to it, not to fall into the delusion. And when you cross that barrier, you discover that it’s a different way to write software, and it’s a different way to write software, and I mention a lot about something called function calling. Function calling is a different way to write software that anyone… Who’s done function calling? A couple say… I can talk to an entire room of engineers, maybe 1% have done it, but it’s going to change how we write software, and software is there to serve us and support us. That’s what I find very interesting, and also exciting.

Okay, so I want to get to the part of my mother. One second. My mother. Okay, I want to get together. This is like a 4x right now. Okay? We go 4x, 4x, 4x, 4x, 4x, 4x, 4x, 4x, 4x, 4x, 4x. Oh, slowing down. There we go.

By the way, I have a show online called the Cozy AI Kitchen, Mr. Maeda’s Cozy AI Kitchen, where I cook AI recipes. And so you can check that out. It’s free. So if you’re afraid of this stuff, you’ll be less afraid if you start cooking with it, understand the math, actually practice with it. And I did it because it’s a side hobby, not even a hustle. I have friends on campus who have this green screen and they were like, “Hey John, there’s extra hours,” and so I started like, “Oh, let me go over and do that.” So we started making this thing, it’s kind of like a lecture company, Fred Rogers and AI. And it wasn’t avowed by any marketing people. We pushed it out there and it’s got a little audience, so if you want to learn AI, check it out.

I’ve also released a risk management course with AI where I’m building things out of clay. That just released this week. It’s sort of a Cozy AI Craft Corner.

Okay. Now I want to note that my mother… Who does tech support for their parents that are not computer educated? Yes. Some of them are computer educated, so they’re okay. Some of them are like, “Whoa, what is this stuff?” So my mother loves her phone. She was a legal secretary so she can type like you wouldn’t know it. She’s like, whoa, you’re 88, you can type faster than me. Anyways, her iMessage does autocorrect, and it really bugged her. It was like, “Oh, this iMessage autocorrect, it’s terrible.” But recently she reversed her decree. So if you don’t mind hitting play, you’ll see my mother’s decree.

Thank you. Oh, can’t hear it. Oh, problem. Oh, audio problem. Ah, see that? I would tell you what she’s saying. No, it’s not as good. Anyways, we’re going to keep going. Okay, thank you very much. I’m going to keep moving.

But basically what she’s saying is that she asked to turn it on again, because she types things like komalanga, she types things she doesn’t want to type. And so she wants it back. And in the end she says, “I like AI. It’s a good one.” Okay, going quick, fast, fast, fast. This is all available in 1x, if you like. Step in there. This is all free information that created every year. Then slowing down. Here we go. No, whoa. Oh my gosh. Over like… Okay, there we go. Liberal arts again.

Ah, here we go. Okay. So this is something from an expert at Stanford that I redrew that I like a lot. It’s a Brin, super long last name. This is yellow. It’s a task that only humans can do. And then there are human tasks that AI could automate, stuff you don’t want to do. And then there’s this whole sea of tasks that humans can do only with the help of AI. It’s kind of like a, I don’t know, like a scuba diver suit or something. You can go to places you didn’t think you’d go. I like this framing. And with that, I bring up the dean.

And I’d like to officially give the dean his own AI Chef Apron.

Dean Anijo Mathew:

All right. That’s what I’m wearing to work tomorrow. So thanks for coming everybody. It’s a great pleasure to be here. When we came in, they had chairs on the table, I mean on the stage, and it looked very good. But I saw John perform at ID and I realized that putting chairs in front of John is a bad idea. John does really well when he’s standing. So I asked him to stand so we could talk through this.

John Maeda:

I can see people too.

Dean Anijo Mathew:

And see people too. Yes.

John Maeda:

Over here. I can see you. I see you back there.

Dean Anijo Mathew:

So my name is Anijo Mathew. I’m the dean of the Institute of Design. For those of you who don’t know the Institute of Design, we are an 87-year-old design school, we are based at Illinois Tech, and we were pioneers of systems design and human-centered design, which went on to become popularly known as design thinking.

So John, you talk about three types of design. This goes back to your 2017 report, right? Can you just do a quick primer on this because I want to talk a little bit more about it.

John Maeda:

So when I landed in Silicon Valley in 2013, I was working in venture capital. And the reason why is because design was creating or inflecting companies and people didn’t understand it. And there were these two guys who were busy grads. One was named Joe Gebbia and the other Brian Chesky, and I visited them, maybe within the second year I was president, I was in San Francisco. I said, “Hey, do we have any entrepreneurs?” And then yeah, we’ve got these two guys, and let’s go see them. Went to see them. They’re in this apartment with six people around a table. They had this ridiculous idea. They were going to enable people to rent out their bedroom or sofa and give them an air bed and they’d serve breakfast, air bed and breakfast, Airbnb. And I thought, what a dumb idea. And then every year it got bigger and bigger and bigger.

So anyways, the investors wanted to understand this thing, and I saw this problem. It’s a problem where people were like, “Oh my gosh, Joe Gebbia, Brian Chesky, design, the game changer.” And so they were all hiring designers, but designers who had no experience in technology, and it wasn’t working out very well for them.

John Maeda:

So I tried to lay out three kinds of design.

One kind of design is classical design, the design that we all know and love, the built environment. It’s like, oh my gosh, design, amazing.

Then there’s design thinking—design thinking, invented at IIT and invented at ID and invented by someone at Stanford. Everyone took the claim, but it was here first. And design thinking though is not design, design, whatever. And classical designers generally dislike design thinking, but I always tell them the design thinker makes a six-figure salary.

Then lastly was the kind of design that involves, in the tech world, that I called computational design. Computational design being anything involving the Moore’s Law effect. I discovered that everyone I told this to, nobody understood and made a funny face when I said the word computation. So I wrote the book, How to Speak Machine to try to change that. I failed. Here we go.

Dean Anijo Mathew:

So keep that in mind, the three types of design. And for those of you who have known John, he’s been writing about this since 2017, I think, right? Now I want you to do another explanation for us. How many of you know what function calling is? Okay, there are only a few people.

Can you explain to us in very human terms what function calling is?

John Maeda:

Yeah, so the way we write computer software is who’s used the basic language? Basic maybe, or logo? Or some of you are forced to learn these things by the school education system, but it’s like, oh my gosh, I’m going to make the computer do this, and then it’s going to do this, and it’s going to do this, and it might keep doing it, and then it might do some more and do it and it might get done. And that was a computer program, right? A human being wrote that computer program, they controlled the flow of logic.

They’re setting variables, they’re calling functions. This whole thing is laid up. Function calling is using large language models to do all the function calling. Meaning, you say, for instance, the example I gave was you have a function that can create a poster. You have a function that can email anybody. You have a function that can move, set up a shelf, set up a table somewhere. Basically three functions that do stuff. And if I had wrote a computer program to sequence how to get ready for this event, I would write it out, 1, 2, 3, 4, 5, 6, 7, go do it. Function calling means instructing the large language model to create the situation for a successful setting of a show like this. And what it’ll do is it’s able to describe what needs to get done and determine what kinds of tools it needs.

So you give it those three tools, like anyone, any person you met, it would look at the three tools to see what it do, and you tell it what to do, it determines the right order of things to do and call the functions without writing any computer program. That’s a function calling this. So it’s a revolution in what software you can write. I will tell you that just a couple of weeks ago, I spent three hour… For this risk management AI course. I show an example of this. I had to shoot it at 11:00 AM. I woke up at 7:00 AM. I wrote a program in three hours that should have taken me probably three weeks. And that was all done using function calling, as an example.

Dean Anijo Mathew:

So let’s take this idea of function calling, and imagine a future world where AI is able to call functions for anything. So John described, we have an exhibition at ID which celebrates our 85th year that has functions that are required to be done. We needed to make tables, we needed to buy the plexiglass, we needed to print posters, we needed to put all these things together, we needed to send out marketing emails, we needed to send out branding emails, all of this, right? Now, if you had AI that could actually call the functions, its large language models has all of these informations in the form of manuals, so it could go in and read. How would Marty Thaler have designed those tables? And you could have, hypothetically, a robot making the table. You could go in and say, how would Mitch Gordon have created the branding messaging? And it could do the function for you. And in none of this is the AI actually interacting with a human being.

John Maeda:

And that’s why there’s this interesting acronym, HIL. It stands for a human in the loop. A human in the loop, it happens to us all the time. How many of us have a dialogue box puppet, but their face it says yes, no, cancel. I think all of us. Yes, no, cancel? That’s human in the loop. Do you really want to do this? The question is, how much human in the loop do you want? So the thing about these models is you can dial them so there’s lots of human in the loop, micromanaging, or less human in the loop. If they’re successful, good. There’s going to be a process of, as developers learn this, to understand which patterns work better, which patterns work worse. But it’s very important to remember that… Remember the known known quadrant? If you leave the known known quadrant, there’s always going to be errors. So engineers will have to really consider this. They’re called guardrails.

Dean Anijo Mathew:

So going back to your first comment about the three types of designers, or design.

What do you think is the evolution of design as computers get more and more better, better and better, at function calling?

John Maeda:

Oh, great question. How many of you who do design, not just for making images, but it’s been useful for scenario planning? Have you tried that yet? Yeah. And you know that it’s not great sometimes, and sometimes it’s good. Like a friend you might call, right? Are they always good? And if anything, it helps you get the job done. So I believe that, you know this notion of the hallowed crit?

Dean Anijo Mathew:

No.

John Maeda:

There’s something called a crit. You never know what a crit is? A critique?

Dean Anijo Mathew:

Oh, you mean like a crit, a jury?

John Maeda:

Yeah, like a critic. I’m going to have a critic come and the critic’s going to show up and doesn’t really know everything, but you’re like, “Okay, go ahead, ChatGPT.” And they’re there. Okay. What’s different is that you may be able to get much more structured critique. You may be able to source a knowledge base that you’re really interested in. I think there is a IDEO person, there’s IDEO’s red book, and all this sort of IDEO knowledge. You could be critiqued in the manner of an IDEO expert in a certain category. I think that’s new, the ability to have a mirror that you’ve chosen versus a random mirror that bounces back information that is sometimes random.

Dean Anijo Mathew:

So that’s really interesting, that you are now predicting in some sense, or speculating in some sense, that the user experience of computational design is also going to change. So now I want you to go back to ELIZA. How many of you have seen ELIZA?

John Maeda:

ELIZA?

Dean Anijo Mathew:

So you should check it out. It’s a really cool chat interface where you ask the question, and ELIZA will ask, “How you doing today?” “Well, I’m feeling miserable.” “Oh, you’re feeling miserable,” and it’s just a text-based chat interface. Just day before yesterday, OpenAI released ChatGPT-4 Omni, 4o, so you saw the Kurzweil curve in work.

Can you first describe the Kurzweil curve? And second, talk a little bit about how will designers play a role in this future where speedcraft becomes much more important than handcraft?

John Maeda:

Yeah. Well, I really spent six years writing How to Speak Machine. It’s coming out in paperback. Just for instance, my MIT Press, I defected and made this book, and MIT Press was always upset at me, so they want to bring it back. So there’ll be like a, I’m going to redesign the cover because the cover was scary. I didn’t choose that. But anyways. I talk about the Kurzweil curve. I actually drew it, redrew it. It’s basically the accelerating power of computing. And in 2011, Kurzweil was on Time Magazine‘s cover. He’s the person who invented the first machine to read for the blind. By the way, he’s amazing scientist. But he predicted in 2011 that computing power would be equivalent to an insect in 2013, a rodent in 2017, a human in 2023, and all the humans on earth in 2034, or something like that.

And it was interesting that, sure enough, the insect happened, and the rodent happened. And you can argue on the intelligence side of a human being in 2023. And I think that he’s living in Moore’s Law time. Do you all? So I really didn’t understand Moore’s Law, even though I heard it so much, until I heard this old riddle. It goes like this.

So it’s a British riddle, and it’s like there’s this pond with lily pads on it. There’s a biologist who lives in the side of the pond. The biologist removes all the lily pads and plants a certain species of lily pads that doubles overnight. So day one, it’s one. Day two, it’s two. Day three, it’s four. Day four, it’s eight. So the riddle goes. So on day 30th, the pond is completely filled, on what day was it half full? So when I was asked this question, I thought, well, half of 30, 15. You know this, you’re smarter than me. Day 29. Day 29, how’s it possible? Because it doubled. So what’s occurring with all this technology is doubling, which in early innings you don’t notice, but in later innings it gets incredibly large. That’s what’s occurring.

Dean Anijo Mathew:

So in your report you talk a little bit about handcraft versus speedcraft.

John Maeda:

Yeah.

Dean Anijo Mathew:

And this also has something to do here, in the sense that you made the argument that we have to become more critical and intentional about the things we create while we operate at the speed that emerging technology is moving. So talk a little bit about that.

John Maeda:

Well, given what I’ve learned about ID in an intense master class from our dean, I think that I wish I came here when I wrote that book, actually, I was pretty ignorant actually, when I consider it. So I’m like, I’m pretty awfully ignorant, actually.

I had a concern that computing was moving this quickly, and it was outpacing culture. Yet in 2024, is that this year? 2024? I feel that these large language models are still slow, and so there’s an opportunity for culture to do its best to catch up. And I encourage people to understand it because it is highly understandable. There’s just two kinds of models. One is a model that can complete the sentence. It’s called the completion model. There’s another model that no one talks about. It’s called an embedding model. It measures similarity. It can find similar words, similar sentences. It can compare a sentence to a paragraph.

It can reduce that comparison to a number. Dog to fog, versus dog to cat. We humans can say zero to 1.5, whatever. A computer could never do that. A computer can now do that. Approximations. Why does chat sound so real now? It isn’t real when you first visit it. It knows nothing about you. When you start talking to it, like ELIZA, it has context. When the context grows, the similarity model is used to find similar things in what you’ve said, and it brings it in to the completion engine that said, when it said, “How are you doing?” “I’m pretty good.” “So where are you from?” “I’m from Seattle.” You’re like, “What’s your favorite food?” “I love tacos,” and blah, blah, blah. And it’s like, “So what’s for dinner tonight?” So you compare what’s for dinner tonight to everything you said, it pulls out, they said, “I like tacos,” so the completion is just, they said, “I like tacos.” “What’s for dinner tonight?” “Tacos.” You see?

Dean Anijo Mathew:

So John, can you spend a little bit more time on this diagram? I love this diagram in John’s report. And for those of you can look at this diagram and interpret it, basically it argues that there are a lot of things that humans can do because we are human beings. And there’s a small subsection of that that AI can do because it’s AI, and there’s the task-based interpretation of what humans can do. But then we step into this large sea, and you have a sequence here. You argue, if I’m assuming correctly, that the task-based things come first, and then we figure out the humans and how humans and AI can work together. Is that correct?

John Maeda:

Well, when I look at this now, and this is by Erik Brynjolfsson, that guy at Stanford, he was at MIT, he’s a genius, so check out his stuff. The diagram didn’t look too good, so I redrew it, made it cooler-looking. Oh, lost my train of thought. What was the question?

Dean Anijo Mathew:

So could you explain this first, the sequencing, and then secondly, what are some the ideas, what are some of the things that are task-based that AI is automating right now? And when we get into this large sea, what are some of the things that you see getting?

John Maeda:

Ah, I remember what I was going to say. Thank you.

Dean Anijo Mathew:

Yes.

John Maeda:

When I saw this, everyone knew, never know, maybe some from my generation. There was this group called Shakespeare Sisters. Shakespearean Sisters. Okay, thank you. Thank you. Shakespearean Sisters. I had to look this up, but they had this song that was a hit, that was a beautiful song, that the lyrics have this moment where they say, “La, la, la, life is a strange thing. Just when you think you know how to use it, it’s gone.” And young people saying this, but it really stuck with me for some reason.

So I always think it gave me a hint, because I also got a lesson from someone else, a man named Mitz Kataoka, in my forties, early forties. Mitz was a very strange person, but he was a very interesting person. But I met him in my twenties. He was the professor of someone named Igarashi Takenobu, who was the Tama Art president.

Anyway, so when Igarashi introduced me to Mitz, he was 60, I was 25. I was living in Tokyo. Mitz would call me in the middle of the night. This is back when there was no cell phones. So he called someone’s house, middle of the night, awkward. But he was very different. And then Mitz, he would call me, wants to talk to me, okay, I’ll go this, whatever. Anyways, I went to a Chinese restaurant in LA. You know those strip malls have amazing restaurants sometimes, you know what I’m talking about?

And we were having dinner, and he said, “You know why I’m like this John?” And I said, “Why?” And he said, “Well, I used to be like you. I was in my twenties. I was up and coming. Good things were happening for me. Was married. My wife was in medical school. We bought a house. We had our whole life planned out. And then one day she came home and didn’t feel right. So a medical student got first in line to get seen with the doctor. Doctor said she was fine. And then the second one said she was fine. But because she was a medical student, she had extra, extra stuff and she found out she had terminal cancer, and she was pregnant.” And he said that after she gave birth, a month later she died. And he said from that moment forward, he never counted on tomorrow coming. So that’s why he was unusual.

Okay. Anyways, I’m 40-something, and I get the RISD president, presidency, and then Mitz calls me up. He said, “Whoa, John, I think that job’s little early for you. You’re like 50-something.” Like, “What’s going on?” He says, “I don’t know,” he says, “I don’t know, there’s this Obama era, 2008,” he says, “Yes we can.” So I was like, “I could do it maybe.” And so then he said, “Well John, don’t forget, life is lived in four quarters. First quarter, 0 to 25. Second quarter, 25 to 50 years. Third quarter, 50 to 75. And fourth quarter, 75 to 100.” He said, “Don’t forget, John, most people don’t make it to the fourth quarter.”

And I was like, “Whoa. I thought I had four quarters.” This was like poof. And he said, “You finished your first quarter,” and that was also poof. And I was ending my second quarter and he said, “In the third quarter, your body starts to fall apart, so make the most of your second quarter.” So I’m in my third quarter now and it is falling apart, and so I am excited by using this technology to really that Shakespearean Sister, I kind of know, and so I’m going to, that’s what excites me.

Dean Anijo Mathew:

So you told me that one of the things that you’re passionate about is speeding things up, because you’re, in some sense, impatient, and you’re saying that if I can do things faster-

John Maeda:

I can do it. Yeah, I want to do it, before I’m gone. So I like that I’m getting… It’s almost as if I feel like I’m getting extra time. I don’t know if I deserve it, but I want to try to use that time differently. So anything I can automate, I’m automating. And it’s great because it’s forcing me to think about how I think. Another thing, again, some of you are healthier than me. So if you’re in third quarter or fourth quarter, no shade. But I know in third quarter, I got like, whoa, this thing doesn’t spin up as fast as it did before, just least me. So by using large language models, I’m more aware of processes done in my brain, but in that architecture, that I can use to make myself think better, or think the way I could think before.

Dean Anijo Mathew:

Great. I have one more question, but before I get to that question, I think it’s important to reach out to you and see if you have questions for John.

Phillip Bahar:

So what are examples of things that you’re automating? Anything you can automate, you can?

John Maeda:

Yeah. One thing I’m automating, I’m embarrassed, is how I use time. Time is the most valuable resource, you come to realize that when you really get older. When you’re younger you’re like, “Oh my gosh,” you waste so much time, it’s amazing. When you’re older you’re like, “Whoa, I got to use my time differently.” So I wrote software to analyze my calendar, and it helped me construct a better next week. Because I’ve given it information about things that are interesting to me, and it can help. Basically it’s a calculator, is how I see it, so it’s able to help me calculate with time constructs and put up what’s important to me.

Another thing that has helped me is I gave the LLM everything I’ve written since whatever, 19, whatever. And it’s interesting because I can basically find things I said before, and I was like, “Whoa, that’s so wrong.” Like, “Oh, I could use that now.” So it’s like injecting some thinking that I had before, so I find that very useful.

Yeah?

Dean Anijo Mathew:

The question?

John Maeda:

Thanks.

Justin McElderry:

Hey, John. Justin McElderry here.

John Maeda:

Oh my gosh, how are you doing?

Justin McElder:

It’s good to see you again.

John Maeda:

Oh my goodness. A Harvard degree.

Dean Anijo Mathew:

He works at ID now.

John Maeda:

Oh, boy. You really are Yankees here.

Justin McElderry:

Anijo, it’s good to see you on the weekend.

John Maeda:

I shouldn’t say Yankees here, I think. But go ahead.

Justin McElderry:

One of the questions that always comes up to me about AI is the idea of the canon changing. So in architecture is there’s canonical buildings, buildings that are ‘more important’ than other buildings. Same thing holds true in literature and other fields as well. So as I think about the advent of AI and historically unjust copyright laws, I’m interested to know your thoughts on the future of copyright and perhaps the way it must evolve, might evolve, given the fact that people are interested in doing more creative things now with these tools that are somewhat limiting, given the fact that you can’t access certain data sources without outdated copyright laws, I’ll put it that way.

John Maeda:

Justin, question. A couple of things. There are new, what’s called, have you heard these large language models? And large language models. There’s also what’s called small language models. There’s a model called Phi, and Phi came out, Phi-1, Phi-2, Phi-3. Phi-3 is called a textbook model, which means it’s trained on things that the copyright is controlled and it’s high quality information, and also it’s open source. People can use it, change it, retune it. I believe that as we discover how to use these kinds of models, I think a lot more interesting things are going to come out of them for people who are primarily driven in the creative space.

In the architecture space, however, is it okay if I’m a little jaded here? You’re the optimist, or you’re Jedi. I left the Jedi. So I do know that if you look at the history of creative people, those who succeeded have tended to come from privilege, because they knew the person to help them get published, or they knew the person who could help them build a building. I think most of the Pritzker Prize winners, you can check them out, there’s a lot of stuff there. So access to opportunities is built into every built structure that we have. I want to strongly say unfortunately, and also want to say reality.

And so I look at, I’ve been so inspired by you, Anijo. I didn’t know why I felt that way. I’m always curious how to change things going forward. I’m always interested in that. And so I’m excited about more people asking, “Okay, so how do we make this different?” Ideally different better. Different worse? It can only happen if more people understand this stuff. If they’re stuck in the delusion that there’s a cup talking to them, and they tell all their friends, and their friends tell all their friends, the people who can access it will not access it because it’s bad. And if anything, has anyone been to the Exploratorium? Exploratorium in San Francisco? Most people don’t know that Exploratorium is probably the most important museum on the planet, because it was founded by Frank Oppenheimer, who is the brother of the Oppenheimer movie Oppenheimer, but he was also a nuclear physicist.

And he, I think, was kicked out of University of Chicago, he was an elite academic. He was kicked out, it was the McCarthy era, communist scare era. So he wasn’t allowed to teach in any of the elite universities. So he ended up somewhere in Nevada or some really out in the boonies, and he became a science teacher, high school science teacher, and he would make these apparatus things to teach kids science. And so a decade later, Caltech was noticing all these kids from this unknown school just killing it in science. And that’s Frank Oppenheimer.

So if you go to the museum, it is an eye-opening experience, because Frank was attempting to get more people to understand science so they would be curious about it, and creating it, because he saw what can happen if you let science be something that others cannot participate in.

Yeah?

Stephen Hunsicker:

Hey, Stephen Hunsicker here. I have so many questions for you, but I’m mostly curious what someone like an ID grad, where someone would fit within an organization that is helping to push AI forward. I’m seeing two types of companies right now. There are the types of companies that are scrambling to figure out, how do we fit AI into our strategy? How do we apply this to our existing product? And then companies that are, while doing that, while pushing AI forward. So the Microsofts, the OpenAIs, the Googles. I’m seeing maybe some of the same kind of job titles you might expect for designers and hybrid designers. But I’m wondering, are there other ones, not just a UX designer, UX researcher, but there are… I’m seeing AI strategists pop up at places like AWS. What is that? How are companies and departments that are pushing AI forward, how are they structured and what should we look for if we’re trying to participate in that?

John Maeda:

Well, this is not a paid commercial. I would say that in my analysis right now, ID grads are very well situated for the following reason. I see why you used the three kind of designs. You got the classical design, you got the design thinking, and you got the computational design. Design thinking is about application and fit to enterprise, enterprises, and also culture in general. But it is not stuck in classical design dogma, it is much more accessible and commercially viable.

And anything in this category that leans computational, not necessarily UX, all that stuff. Being a strategist who is AI enabled and also understands the thing I described just now of the embeddings model and the completion model, and really not just… There’s so many experts out there that just talk about it because they just used the ChatGPT the other day and they’re experts. Just saying. If you actually sit in the code, you run it, you touch it, you change it, and you’re like, “Oh, I can do this. Oh, I can’t do this.” That is the exciting time right now. There is an opportunity for more human experts to have an opinion for how this really can be used well, and I think ID is good stuff.

Dean Anijo Mathew:

We have one question here. Can we get there? Oh, behind there.

John Maeda:

Yeah.

Dean Anijo Mathew:

Sorry.

John Maeda:

In full disclosure, I am like a computer programmer now, much less a designer. So I spend all my time coding.

New Speaker 1:

Hi.

John Maeda:

Hello.

New Speaker 1:

I’m curious about the faster these large language models advance, and we’re getting closer to what people call AGI, right? So how designers can start to be part of the creation of AGI and actually make it safe and make it properly, because it’s like designing a human being kind of parallel behavior, so it’s like designing a human. So at what point designers are going to be part of the creation of this AGI, and what is the pathway for designers to get into the industry or getting into that step of what it means creating AGI?

John Maeda:

Yeah, I’m not a good person to answer that because I really don’t eat all that stuff. I think there are some people who have a really strong opinion about things like AGI, et cetera. I’m not one of them. I’m more of a pragmatist. I grew up in a family with no education. They made tofu for a living. I made tofu with my hands all the time as a kid. I somehow was able to leave Chinatown because of education, and I’ve just been curious about the technology and what it can do for more people. These questions about what it does to change society, and I think a lot of the science fiction movies create this kind of feeling. I’m not in that world, so I regret that I can’t respond to that kind of question. I’m less in there. I’m more really in, what do we do with it? I’m curious how to use it, and how to use it really as a tool.

Dean Anijo Mathew:

I’m going to close out the session.

John Maeda:

Can I get this person in purple here? He had that moment that I had recently where I lost a microphone. You sure? If you don’t mind. I was there. I was going to ask a question.

New Speaker 2:

So we’re seeing all of this rise since the large language models came about. When we were all sitting at home during COVID, service delivery was everything and everybody was too busy hiring more UX designers. And today we see the new darling of the industry is anything AI. But as a designer, I can’t help but wonder how natural the whole paradigm of interacting with machine via prompt is, and what does it mean to go away from this retrofitting and go to AI-first design world?

John Maeda:

What is really cool for those people who are defining their career right now is that there are a lot of unknown unknowns, and there are unknowns in the category that you’re describing too. A lot of unknown unknowns by sitting. And I really having the privilege of to sit with you for a few hours, I don’t have to pay for that. That was great. I’ve been so struck by how making is what design does. And you can either make good things, or make bad things, but making is the proof. What is that phrase? A prototype is worth a thousand meetings? And to your point, the designers are really good at making that prototype, and that’s the only way it’s going to be resolved. But I ask to try to stay away from the eddy pool of the delusion of talking to something because it’s quite powerful. 1966, it’s been warned.

Dean Anijo Mathew:

Great. And thank you, John. Thanks everybody. I’m going to close out the session, and I’m going to quote something that John references in his presentation, in his book, you cite Josh Clark, who argues that, just like Mies van der Rohe’s generation, saw steel and glass as the material of design. You are arguing, and Josh Clark is arguing, that AI may be the material of design right now. And I think you are just describing that if you don’t have the craft of molding AI to you and to do the stuff that you want to do, you may be left behind.

John Maeda:

Well, you’re hallucinating.

Dean Anijo Mathew:

You’re hallucinating, to use AI term. So thank you so much everybody. Thanks to John.

John Maeda:

Thank you, thank you.

Dean Anijo Mathew:

All right. I do want to thank a few people before we step out. I want to thank Lauren Pacheco and Phillip Bahar for bringing Chicago Humanities to Illinois Tech. I want to thank all the deans at ID, Illinois Tech, who supported this. And I encourage all of you who have not been to the Institute of Design to go and check out the Institute of Design. It’s just a hop over, you can’t miss it. It’s a white building with an inflatable second floor. So go in and check out the Institute of Design. We have an 85-year exhibition there. So thank you everybody. Have a great evening. And we look forward to seeing you back again.

Thank you.