Chris Cooper
Matters of Experience
January 10, 2024

Demystifying AI with Chris Cooper

Chris Cooper, Partner at Lorem Ipsum Corp

/imagine two podcast hosts, Abby and Brenda, speaking to a guest, Chris Cooper, at a recording studio in New York City about utilizing artificial intelligence to explore the synergy of human ingenuity and machine intelligence within experience design. --style raw

About Chris Cooper

Born in Austin, Texas, he graduated from the Faculty of Political Science at Bates College, located in the state of Maine, and then received his MBA at the University of Berkeley in California. Between 1998-2005, he worked as a financial analyst at the largest Boston bank FleetBoston. At 32, he decided to completely change his line of business and devote himself to filmmaking. He produced several documentaries, including Life in an Emergency: Stories of Doctors Without Borders, which was shortlisted for the Oscar. In 2008 he became a partner in the company Lorem Ipsum. He participated in the creation of numerous advertising, film and exhibition projects, including multimedia installations, installed in the international offices of Google, a large-scale exhibition at the Yeltsin Center in Yekaterinburg.

Show notes

Transcript

[Music]

 

AI Abby: Welcome to Matters of Experience. This podcast is produced by Lorem Ipsum, an experience design company headquartered in New York City. Our show explores the creativity, innovation and psychology driving designed experiences. If you’re new, welcome, and to our regular listeners, thanks for tuning in and supporting our conversation. My name is Abigail Honor.

 

AI Brenda And I’m Brenda Cowan.

 

AI Abby: Today’s show, we are going to take a deep dive into artificial intelligence and experience design, looking at how we collaborate with AI during the process of creating a museum, the limitations right now and general potential worries about the future impact on our industry.

 

AI Brenda: But we couldn’t start a show about AI without using AI.

 

AI Abby: Brenda, I want to reveal that I am Abby and you heard AI Brenda earlier. So, I am going to hand the show over to the real Abby and Brenda, but I will be back for the show sign-off.

 

Brenda: Oh, my God. Did you miss us, everybody? We were here all along. I was here all along terrified, listening to Abby sound like this and just wondering, how on earth did that happen?

 

Abby: I do think, Brenda, that you did sound better than me.

 

Brenda: I sound better than I really do in real life.

 

Abby: AI has something against people from the North West of England.

 

So, love it or hate it, AI is taking over, causing some to worry about job loss, bias, or security while others enjoy the convenience, productivity, innovation, and personalization it affords. Today, we welcome my business partner at Lorem Ipsum, Chris Cooper, to the show. He’s been obsessed with AI for a while at this point and is knee deep in our R&D in AI and also spearheads its application in our experience design team. Chris, a big hello.

 

Chris: Hello.

 

Brenda: Hello Chris, this is Brenda. It’s very nice to meet you.

 

Chris: You actually sound worse than your actual AI.

 

Brenda: I’m sure I do. No, in all seriousness, it’s such a pleasure to meet you, and I am genuinely curious about all things AI. And also, I know we’re going to talk a lot about AI in terms of our processes and our protocols and our work in design and so forth, and I thought I would just share with you a very few areas in my workplace where we are working with AI, and it includes all areas of business, technology, mathematics, sciences, gaming, design of all types and English studies, English as -second language studies, and it is absolutely everywhere in the world of higher education, and with all of the considerations and concerns and even conflicts that come along with that.

 

Chris: I think academia, it’s interesting because I typically think of—this is probably a prejudice—but I typically think of academia as kind of being stuck in not the most advanced technologically, but for whatever reason, this is like, it’s like the front lines for a lot of AI discussions and what have you. So, it’s kind of an interesting perspective that you have.

 

Brenda: It’s really interesting when you’ve got students, and I’ll speak on behalf of my institution, you know, it is an incredibly global institution, both in terms of the student body, but also the faculty and staff—

 

Abby: I just want to plug that’s FIT for anybody who wasn’t aware.

 

Brenda: FIT in New York City. And it’s virtually impossible for us to be teaching a large number, especially in our schools of business and technology, our schools of art and design; we have to be teaching AI because our students are and we’re working directly with industry.

 

Chris: And they’re going to use it. So, you have to get out ahead of it so that you can participate in how they use it and make sure that you guide them in the right direction.

 

Brenda: Right, which to a person such as myself is, honestly, it’s not terrifying, but it’s definitely, you know, I’m feeling like I’ve got so much catchup work to do, which I’m determined to do, but I’m really looking forward to diving deeply into not just the kinds of AI that you’ve been using, but the reasons why, and how it’s really working, because I think I’m going to take a lot out of this conversation.

 

Chris: You know, one thing that I think we should establish right now is I am not an expert, right? Like, I’m just a guy who’s trying to figure out how this is going to work in what we actually do to make money, right. And the second thing is, everything’s happening so fast that even anything that we talk about right now, in a few weeks, months, is going to be completely dated, which is really frustrating when you’re in my position where you’re trying to figure out workflows, and they’re constantly changing.

 

But I think that everything’s developing so quickly that I’m not sure that we’re in a position right now where people are behind. In the sense that, like, everything that I’ve gained or expertise that I have right now may expire in three months, and if you start doing stuff right now, you may be caught up with where I am in three months because new tools will come out. And all the new tools that are coming out, they’re trying so hard to make them accessible for people right now. Right now, most of the stuff that we have is pretty janky, for lack of a better word, and you kind of have to hack things together to make them work. And I think that that’s, kind of feeds into a lot of people’s hesitance to adopt it and also a lot of people’s fear about what it might mean because it’s so foreign and it’s not accessible.

 

Abby: Well, I’ll start. We’re going to start at the very beginning because I know, Chris, we use a lot of language models. So, can you describe, so, at the beginning of a project, when we have a team, and we’re just even brainstorming, you know, how you collaborate with AI?

 

Chris: I mean, in general, language models are like the most prominent AI that we’re using, right? So ChatGPT obviously is kind of the leader in that area, but we also use Claude, which is another one. So, what we’ll do is we’ll use the language models to brainstorm the initial ideas. If we’re doing a concept and we usually get together and we kind of talk about the concept, what the client wants, what we’re starting to think about and just starting to wrap our heads around it. And what I do, and not everybody on our team does, but what I do is I feed all of that into the language model. I feed any documents that the client has given us. I feed any conversations that we’ve had, and I start having a conversation with the model about what we might be able to do for this concept.

 

And I think that’s like a key thing that I’ve seen internally, that it’s a stumbling block for a lot of people, is a lot of people are just like, go to the model and type in, give me ten ideas for this. And the model kind of just goes blah, and gives you like stereotypical, trite ideas, and then they leave it there, and they’re like, oh, see, AI’s stupid. And what you really need to do is have an iterative conversation with it, where you tell it all this stuff, you start telling it what you want and giving it some of your like nuanced ideas, see what it gives back to you. Then you correct it and say, no, I don’t like that. I want it more like this.

 

And you end up having a conversation. And the way I think about it is, you know, when you’re teaching somebody something you really like start to learn the material even better than the way you originally learned it, because it challenges you to fill in all the holes in your thinking. Well, these conversations that you have with a language model start to force you to think creatively about what you actually want to do with the concept because you’re having a conversation where you’re directing the model towards what you want it to end up with.

 

Then, once that has happened, I will have the AI play a different role with me where I have it critique the ideas. So I then say, okay, approach this like you’re an expert in the field of exhibit design or what have you, critique those ideas, and then I’ll take that critique, and I’ll feed it back into the brainstorming session I had and tell the AI that I’m brainstorming with, change the original ideas we came up with to address these critiques, and then it will generate whole new ideas based on this critique that it essentially has provided. And so, you enter this kind of iterative cycle where you begin to develop more and more interesting, unique ideas than you would if you just sat down by yourself.

 

Brenda: This is so much like cooking.

 

Chris: Yeah, I guess it is a little bit like cooking.

 

Brenda: I’m getting such a strong sensibility of how it is that things can just endlessly be reduced and refined. And my question for you is, when do you know that you’ve got the sweet spot? When is that sauce perfectly reduced?

 

Chris: Well, it’s just when you like it, when you’ve got ideas that you like and then you go back to your partners, and you talk to them and either they respond to them or they don’t. But it just, it’s a process that allows you to kind of get out of your head.

 

And then one thing that I left out that I’ll typically do is I’ll then also ask it to critique it not just from like how good an idea it is, but like what sort of red flags or issues might someone have with this. So those could be political concerns. Those could be like, I don’t know, blind spots that I may have culturally. And so at least that lets me think about that before I just blindly go in and present something and I haven’t thought about like, oh, these indigenous people from this sector may have an issue with this.

 

Brenda: Give me a story because we all know that part of the challenge with AI is bias, and we know that there are racial and cultural biases. That said, when you use your technology towards trying to better understand your own thinking and perhaps your own, like you said, blind spots, can you give us a story of that happening and, does anything come to mind?

 

Chris: I think a lot of what you’re talking about, and what a lot of people critique, is that it’s a very passive perspective when you bump into those prejudice. It’s like, give me this, and it gives you that, and you’re like, oh, well, that’s, that’s got a bias. Whereas what I’m talking about is it’s a conversation you’re having. And when you have a conversation with people, every single person you have a conversation with has a bias, and you just have a conversation, and you dismiss the bias, or you say, no, that’s a bad idea, that’s stupid, or no, we need to think about this or whatever, so you can get past those biases if you’re aware of them, right?

 

Another thing that I would say is AIs don’t just have, or we should talk about, language models; they don’t just have one perspective. Part of when you prompt them is you prompt them to have the perspective you want them to have. So, if I prompt it to brainstorm with me, I tell it it’s the greatest exhibit designer in the world, you’ve won all of the awards, and then that gets it to start thinking that I need to respond as this.

 

And if you think about how these models work is they’re predictive models, and they’re trying to predict the next word or token, right? Well, when you start just with the blank page, they’re doing that prediction based off of everything that it’s been trained on. It doesn’t care if it’s right or wrong. It is just going to say what is the most likely next thing to say. As soon as I say, you’re the greatest exhibit designer in the world, you’ve won, blah, blah, blah, all the sudden, it narrows that down so that what it’s predicting to say is only based on the data associated with the very best exhibit designer in the world.

 

Brenda: So, have you run into any kinds of challenges? I’ll just throw out one, you know, challenge that I encounter in my work, which is hallucinations. If I’m asking students to write a scholarly paper and the next thing you know, they’ve got all of these, you know, citations and these resources for, yes, actual authors, but with fake books.

 

Chris: Yeah, yeah, yeah.

 

Brenda: Right?

 

Chris: Absolutely.

 

Brenda: Do you run into this?

 

Chris: Yeah, so that’s why when you’re doing conceptual stuff, right, it’s really well-suited to that because when you’re doing conceptual stuff, you’re not asking it to give you like—

 

Brenda: Facts.

 

Chris: —facts. You’re just brainstorming, right?

 

Brenda: Yup.

 

Chris: The hallucination stuff is something you have to be careful of, but that’s going to become much more of a concern later in the process. As far as the brainstorming, obviously, you should be aware of that. But, you know, if you think about the tools we usually use for brainstorming, you’re going to be searching the Internet, you’re looking at Wikipedia. All of these sources are prone to mistakes—

 

Brenda: Yeah, it’s the wild west.

 

Chris: —that you’re always going to have to take into account. So, it’s the same type of thing as you’re probably having to teach your students all the time; don’t just accept what you find as fact. You always have to like—

 

Brenda: Of course.

 

Chris: —think about it and, review it and go back and critique it.

 

Abby: So, when you have an idea, and you move into the concepting, we use Midjourney, Stable Diffusion, and DALL-E, can you tell us about that process and some of the benefits of AI at that stage?

 

Chris: Right. So, so far, we’ve talked pretty much about the language models, right? What Abby is talking about are like the text-to-image models. And so, what we’ll do is we will take the ideas that we came up with, even if they weren’t developed with AI, say someone just has an idea. We’ll still go to the text-to-image—

 

Abby: Which we still do, just to note to all the listeners, we’re not only using AI, we do have our ideas.

 

Chris: Of course. What we’ll do is will take whatever idea that it is, and we’ll put it into the text-to-image model. And it’s a great tool because when people have an idea, we can all say the same idea, but as soon as you see a picture of it, you all realize that none of y’all were thinking the same thing. You’re like, oh, that’s blue. And Abby will be like, no, it’s white. And you know, you’ll have the whole conversation like that.

 

Brenda: You know, I’m curious, as you know, the conversation unfolds, I’m really curious about your process at your company for generating exhibit ideas, exhibit concepts, exhibit images. You, I presume, before you were actively using these models, you had a process for brainstorming, generating, producing, and so forth. Has your process changed as a result of this, and in what way?

 

Chris: Well, so traditionally, we’d have sketches sometimes, sometimes we would be photoshopping things together, and then we would move past that. And we still do this. We still model, but the hurdle for modeling and creating renders is much higher than going into a text-to-image generator where you can just spit out stuff, right?

 

Traditionally, what we would do is we would sell in a concept, maybe have a few directional renders that we did. And then, once the concept has been approved, we would go through the design process, and the design process would be much more rigorous because we would be using actual 3D modeling software, we would be using the architectural models that came over from like the shell and core from whatever architect is working on the space, so everything’s accurate.

 

And then that can be intensive because then you’re like properly modeling things out, you’re properly texturing everything, you’re properly lighting everything, and you’re generating those renders and then presenting imagery based off of that, maybe even bringing it into a game engine so that people can walk around in the space and have real-time experiences with it and that is not going to go away because clients are going to still want at some point an accurate model of the space.

 

I think it will all change with these technologies, but you will move from like a conceptual text-to-image idea presentation up to a more and more accurate model that may incorporate AI into it, but it will still have like accurate measurements.

 

Brenda: So, in your brainstorming process, do you find yourself using words and language a little bit less and going right into visual, visual images and that sort of thing? Or are you still…

 

Chris: We’re still using language. You still have to tell the story behind it. But what’s happening is because the imagery is so easy to create now, we can even internally, when we’re brainstorming, we can be generating imagery to show each other even as we’re talking. So, the images are informing the brainstorm at a much earlier stage.

 

Abby: And it’s interesting, as Chris mentioned, it’s easy to make these images. What’s hard is to make the images that are right for the project, and that will never change. And there are some drawbacks as well. Like Chris and I feel that DALL-E’s visuals are a little plastic, right? And the texture isn’t really nice. They’re not lush. They look a bit plasticky.

 

Chris: Yeah. 

 

Abby: It’s still not there yet.

 

Chris: So, all of the models have positives and negatives to them. So, like Midjourney is like a beautiful, cinematic imagery, but it’s harder to prompt right now. You have all of these disparate things that, to your analogy, it’s a little bit like cooking. You’re like, oh, let’s, let’s bring all this together, and then you can mix them in different ways to come up with really interesting outputs.

 

Abby: Do you think, Chris, that collaborating with AI on concepts has resulted in sort of the end result being less valued than it were if a human was creating it?

 

Chris: I think that that’s a danger that we’re going to run into, and we’ve seen this in other areas. We saw a lot of this in video work where, you know, there was a time where in order to like cut a video, you had to have all this equipment that people didn’t have access to. Once people had access to iMovie, their value for what we would do when we were making a video went down because they’re like, my nephew can make a video, and you’re like, yeah, but your nephew can’t make this video. But somewhere in their head, they’re like, my nephew—

 

Same thing is going to happen here where, used to be in order to make beautiful renders it required like a level of investment between having the right people on your staff, the right equipment, the right amount of time, the expertise, all that sort of stuff. Now, you’re going to have a lot of people who can just crank out a lot of imagery, and not everyone’s going to be able to tell the difference between, or they may be able to tell the difference between a good image and a not-so-good image. But they’re not going to value the expertise that created that difference.

 

Brenda: So, this is like, you know, Sunday evening dinner conversation in my household because my husband’s a digital retoucher and, as of right now, and he does very high-end retouching for print, photoshop is not, the AI is not there yet to be able to do pristine, pristine level image creation for what he does. But, you know, I think these kinds of things, it’s really only a matter of time before Photoshop truly can just create magnificent, truly magnificent images that a trained artist can create. And/or is the standard going to be so low? I mean it. Is the standard going to be lowered, chipped away at bit by bit?

 

Chris: I wouldn’t have immediately thought what you’re thinking. But if you look at what happened with video when the Internet happened, the quality of video just went to pot, and everyone just got accustomed to bad video. And you’re right, something like that may start to happen, but there’s still going to be a difference.

 

And I think the best way I can describe it is our other partner, Yan, is a DP, so he’s great with lighting and setting up shots, right? And even when we create a render where we push it through our entire workflow with really talented 3D artists and they do a great job, right? And then they send over a render, and we’re like, no, because they framed it wrong, and they haven’t lit it the way it should be lit. And then Yan can sit down and be like, well, no, just lower the camera, change the lens, tilt this, like you need a backlight over here, and all of a sudden, it completely changes the image. The difference between those two images might not be readily apparent to a lot of people, but when you’re selling something, someone is going to have an emotional reaction to the one that Yan lit and framed, and they won’t to the other one.

 

Abby: I would like to say with Chris’s example of video as well, although there’s a proliferation of low quality, low budget media everywhere, there are still remained the feature films, the ultra-high budget, the way they’re shooting it, the technology—

 

Brenda: Oh yeah, the 70mm.

 

Abby: Yeah, exactly. So, I think the best always remains. So, I think your husband is fine, and he’ll always have a job, you can let him know. It’s going to be fine; we’ll still need his skill set.

 

Brenda: It’s not about him.

 

Abby: But I think that, you, society always values the best of the best. I just think there’s a proliferation of crap, maybe.

  

Brenda: From what I understand, it’s hard to really capture in words at this point; AI is not quite capturing a level of emotion that you know can be found right in the human hand, if you will.

 

Chris: Yeah, I think in general, and I think there’s a fair amount of debate in this about how far can AI advance if it is not experiencing the world, right? I mean, we all experience the world, and a part of the connection that we establish with one another when we make art or whatever it is is through that shared experience where, somehow, we’re tapping into what it is to have been alive and experience things. And if AI just is completely foreign to all of that, it may never be able to do that.

 

Brenda: Well, ultimately, right, AI is understanding the world through people’s interpretations of the world.

 

Chris: Well, except for now, that’s changing.

 

Brenda: Tell me.

 

Chris: Well, like a language model is trained by just feeding it like the Internet, right? As the models are becoming multimodal, which means they can see and they can hear, they’re going to start experiencing the world more like we do. And if—now, this is where I’m getting out of my depth, right, but this is how it seems to me—if they start experiencing the world more like we do, then what we’re talking about may start to go away.

 

Brenda: Yeah, it’s fascinating to still push for my part anyway, the idea that ultimately it is still perceiving other people’s generated data. So, the question will be at what point, right, is this truly autonomous and what that autonomy is going to mean in a, you know, and we could talk about agency as well, AI having its own agency.

 

Abby: Yeah, I mean, for me, it’s all about evolution is inevitable and it seems like AI is part of our evolution and may take us over as we evolve. We’re constantly moving, and we feel like we’ve always been here as we are today and have never changed. But we have. And you look back a million or so or even nearer, and we weren’t the way we are today.

 

So, I think the future is hopeful, and we’ll see what technology does and how it augments us as it is already doing. So, okay, I’m going to hand the outro over to AI. We’re going to feed this edited podcast into it and give it some of our previous outros, and see how it decides to wrap up our show today. So, take it away AI Abby and Brenda.

 

AI Abby: Well, that wraps up another intriguing conversation. We’ve delved deep into the world of AI, its applications, and its potential impacts.

 

AI Brenda: Absolutely. It’s been a pleasure exploring these nuanced perspectives. And as always, we appreciate you joining us on this journey.

 

AI Abby: Indeed. Thank you for tuning in today. If you enjoyed the show, subscribe to Matters of Experience wherever you get your podcasts. Don’t forget to leave a rating and review and share it with your friends.

 

AI Brenda: Until next time, stay curious and stay engaged. Thank you for being part of this exploration with us.

 

AI Abby: Thank you, everyone. See you on the next episode.

 

[Music]

 

Producer: Matters of Experience is produced by Lorem Ipsum Corp and recorded at Hangar Studios. Tune in next time for more fun discussions about experience design.