Skip to content
    ← Podcast
    CREATIVITY
    Podcast 13 min read

    Why AI Can't Teach Taste w/ Brooke Hopper

    with Brooke Hopper

    Chris Do and Brooke Hopper examine why AI can accelerate design without replacing human taste.

    Chris Do

    Chris Do

    Founder, The Futur™ · September 9, 2026

    The machine can train the designer

    The danger is not simply that AI will imitate designers. It is that designers will start imitating AI. Brooke Hopper, a senior principal designer at Adobe whose work includes bringing Firefly into creative tools, sees that risk from inside the software being built to help them.

    Her concern is not an argument against the technology. It is an argument against surrendering judgment to it. A system that produces recognizable patterns can become a powerful influence on what its user learns to recognize as good.

    Speaking with Chris Do, Hopper puts the reversal plainly: “And if you are only following the patterns, it's training you as a human.” The tool supplies possibilities, but repeated acceptance can turn those possibilities into an unexamined standard.

    That distinction gives the conversation its central tension. Do sees AI as a way to work faster, explore more ideas, and move from production toward art direction. Hopper sees those opportunities too, but refuses to treat increased capability as evidence of increased discernment.

    Generating an option is not the same as judging it.

    For Hopper, that judgment depends on fundamentals, curiosity, and lived experience. Two people sitting in the same room do not bring the same history to an image or tell the same story through it. Their differences are not inefficiencies for software to smooth away.

    They are part of the work.

    Her perspective also comes from proximity to people who have reasons to be uneasy. She says she has spent ten years at Adobe, seven of them building drawing apps, and describes close ties to the illustration community. The concerns she hears about craft and its future are legitimate, not resistance to be dismissed.

    That matters because the easiest defense of creative AI is also the least demanding: call it a tool and move on. Hopper uses that description, but insists that building creative experiences requires holding competing needs together. Assistance must be useful without making the person using it incidental.

    The question is therefore more specific than whether AI can produce attractive work. It is whether a designer remains capable of saying why one result serves the idea, why another fails, and why the most immediately pleasing option deserves rejection.

    Hopper connects that capability to knowing the rules well enough to break them. Without that knowledge, a designer has fewer grounds for resisting the patterns the system offers. The output becomes not a proposition to examine, but an answer to accept.

    This is also where storytelling enters the argument. Hopper locates creative difference in the stories people choose to tell through their own experience, a concern reflected in The Futur's Storytelling Is A Superpower. Her emphasis is not on a more elaborate prompt, but on a more deliberate point of view.

    The promise of AI, on those terms, is not that every designer becomes equally capable of producing the same polished image. It is that more of the work surrounding an individual decision can become easier, while the decision itself remains consequential.

    A faster process still needs someone who knows what deserves to survive it.

    Close enough fails at the finish line

    Hopper is wearing a shirt that says “control freak.” It is an unusually efficient description of the problem she is trying to solve. Creative software must accommodate a technology that generates variations while serving people who care about exact placement, exact appearance, and exact relationships.

    Her account of Adobe's role centers on precision and control. The broad gesture is only part of designing. Tiny adjustments matter because they determine whether an image is merely convincing or actually the image the designer intended to make.

    Do recognizes both sides of that experience. Sometimes Firefly produces a result exciting enough that he calls his wife, also a designer, to see it. At other times, the software fails to perform the change he believes he has requested clearly.

    Hopper does not explain that frustration away as a prompting deficiency. She gets bad results too, including one during a presentation that day. The admission matters: disappointment with a generation is not automatically proof that the user lacks skill.

    She describes generative AI as a technology that makes something new rather than guaranteeing the exact thing requested. That novelty can open a direction. It can also become the obstacle when the direction has already been chosen.

    Exploration tolerates approximation. Delivery often does not.

    Her industrial design example makes the boundary concrete. A student or designer can use generation to imagine a product in different settings or explore how an idea could look. A company commissioning finished advertising, however, wants its actual product, not a persuasive relative of it.

    “Close enough is not good enough with a lot of those things,” Hopper says.

    That is a useful corrective to the assumption that a successful concept image is almost a finished asset. In her example, the criterion changes between those stages. During exploration, variation is useful evidence; during delivery, unwanted variation can invalidate the result.

    Hopper describes AI helping at opposite ends of her process, with precise creative work remaining important in between:

    • Early exploration can benefit from unexpected options. Generated visuals can help a designer investigate concepts alongside sketches.
    • Finishing requires deliberate control. Refinement still needs tools that support specific adjustments.
    • Production can benefit from pattern recognition. Repetition and analysis are useful places to look for assistance.

    These are not promises that every task inside those categories should be automated. They are distinctions between kinds of work. Hopper's argument depends on recognizing what the machine does well and what the designer contributes, rather than assigning the whole process to one or the other.

    She describes an industrial designer friend who wants to draw one or two views of an object and use generation to produce additional angles. The ambition is modest compared with replacing product design altogether. It is also more closely connected to an actual annoyance in a working process.

    That is where her standard for usefulness becomes sharpest. The valuable intervention is not necessarily a spectacular demonstration. It can be a small improvement at the point where a designer repeatedly loses time, provided the resulting work can still be controlled.

    The task is not to make generation responsible for everything. It is to place generation where its behavior helps rather than compromises the job.

    A classroom built around the wrong tool

    Hopper's teaching exercise begins with a restriction that sounds like technological evangelism. Students must complete an entire assignment using generative AI. No other tools are allowed.

    Many do not like it.

    That reaction is part of the exercise's value. A student prevented from reaching for familiar software encounters not just what generation makes easy, but also what it makes needlessly difficult. The assignment exposes the boundary between a tool's impressive capabilities and its suitability for a particular decision.

    Hopper does not present the restriction as a model for professional practice. Its purpose is diagnostic. By forcing exclusive use, she gives students a reason to notice where exclusivity breaks down.

    This is a more demanding form of technological literacy than simply learning to obtain an appealing result. The student has to identify where the tool belongs in a workflow and where an established method remains better suited to the task.

    The educational issue connects naturally with The State of Design Education. Here, however, the immediate problem is not which software a curriculum should include. It is how students learn to evaluate software without mistaking its availability for a reason to use it everywhere.

    Do offers a complementary example from a conversation with a graduate program director at SVA. Product design students sketch first, then use AI to explore iterations and materials. The original act of defining the object is not erased by the speed of subsequent variation.

    Hopper's assignment and the sketch-first example approach the same boundary from different directions. One deliberately removes alternatives so students can feel the limitations. The other gives AI a specific role within a process that already contains human decisions.

    Do then pushes the educational question further. If a system has encountered so much visual material, can it explain its choices? Can it generate a conservative version, describe the composition, and suggest alternatives in a way that helps a less experienced designer improve?

    Hopper's answer leaves room for assistance without equating assistance with authority. A system can check known requirements, especially when the criteria have been supplied. She describes this kind of help as “design spell check.”

    Her examples distinguish several uses:

    • Checking whether work follows a brand's stated guidelines.
    • Offering guidance on basic design decisions.
    • Helping a person notice something useful during the process.

    None of those functions eliminates the need to learn the underlying principles. A check can identify whether a rule has been followed. It does not settle whether following that rule produces the most appropriate design.

    Learning the standard is not the same as developing taste.

    Hopper is skeptical that a machine should become the source from which a designer learns taste. That skepticism is narrower, and more useful, than claiming AI cannot teach anything. She explicitly allows that it can guide people and help them learn along the way.

    The unresolved issue is dependency. A designer who accepts a recommendation without understanding it has received an answer, not necessarily an education. A designer who can evaluate the recommendation has gained assistance without handing over the responsibility for deciding.

    Do's proposed tutor would explain the work. Hopper's condition is that the person still learn how to disagree.

    The useful work of resistance

    When Do asks what she would want Firefly to do if anything were possible, Hopper initially answers in terms of the creative experience. She wants people to enjoy making things. More content produced at an ever-increasing pace is not the aspiration that brought her into design.

    Her examples are specific: time for an idea, a beautiful interface, or packaging she wants to make. The software earns its place by taking care of unwelcome work so attention can return to the part that feels worth doing.

    Do presses for a more concrete answer. Hopper then introduces an apparent contradiction: some friction should disappear, but some friction has value. Limitations can contribute to play, and effort can help a person feel ownership of the result.

    Remove busywork without removing creative participation.

    That distinction is easy to miss when speed becomes the only measure of improvement. A repetitive operation and an unresolved creative choice both take time. Hopper does not treat them as equivalent problems simply because neither has happened instantly.

    She has been thinking about intentional friction in AI experiences, but presents it as an open design question. She does not announce a policy of deliberately weakening software. When Do asks whether the tools should be handicapped, she returns to the harder problem of deciding where resistance belongs.

    Her typography exercise offers a small example of why participation matters. A person interested in typography assembled a group and asked them to generate type jokes in three different conditions:

    • First, the participants invented jokes themselves without searching online.
    • Next, they asked AI to produce type jokes.
    • Finally, they worked with AI to develop the jokes together.

    Hopper says the third round produced the better results. This is an anecdote about an exercise, not a controlled study establishing a universal rule. Its relevance is in the distinction between receiving material and responding to it.

    The participants brought their own ideas into contact with generated suggestions. A suggestion could prompt a revision or another direction rather than being treated as the final answer. The exchange gave human judgment something to push against.

    That places the example near the concerns named in Learning, Collaboration, Critique & Feedback: creative work develops through response, not just initial production. Hopper's contribution is to describe how generated material can enter that process without becoming its governing intelligence.

    She is nevertheless careful with the language of partnership. Calling AI a collaborator feels too strong to her in the present context of working with creative people. She recognizes the back-and-forth without wanting the label to obscure the person's role.

    Her point is also not that every generated result demands extensive editing. She concedes that a person can sometimes be satisfied with the direct output. But she expects many creative practitioners to want to revise it, make it more specific, and put something of themselves into it.

    Even a software bug can become material for invention, she observes. An unintended limitation can lead someone to an unexpected technique. That possibility complicates any simple promise that the best creative tool is the one with no obstacles.

    The distinction is between friction that obstructs intention and friction through which intention becomes clearer. Better tools need to understand both.

    A visual tool should speak visually

    Hopper's more concrete wish for future tools begins with an ordinary gesture: selecting something and changing it. Not describing it at length and hoping the system infers the right relationship. Pointing to the thing itself.

    She discusses mask tracking in After Effects and Premiere as a starting point. A selection can be followed through video. Her speculative next step is to make manipulating that selected object feel more like handling an object than issuing instructions to a separate production system.

    What if an element could move from one video into another through a direct gesture? What if video editing began to resemble image compositing? Those are possibilities she raises, not a list of promised product releases.

    The direction is more important than the imagined feature. Hopper wants creative interfaces to feel closer to the physical experience of moving, turning, and arranging things. Her reference point is a child picking up a crayon, not a user mastering a specialized vocabulary before being allowed to make a mark.

    Her objection to prompting follows from that concern. She describes herself as poor at it and questions whether verbal instruction is a natural primary interface for visual people. The ability to express an idea in words is not identical to the ability to see what needs changing.

    A designer can recognize the desired adjustment through direct interaction. Moving an object, testing an angle, or changing a relationship gives the person feedback inside the medium. Hopper wants technology that makes those actions more available, rather than forcing every intention through a text box.

    Do translates the ambition into a filmmaker's problem. A shot would work better from another camera angle. An action needs changing. Reframing the image after capture would solve a practical production difficulty, provided the result does not introduce unwanted anatomy or other visual errors.

    Hopper points toward the potential combination of 3D tools, generative technology, and traditional editing tools. She does not provide a release schedule or demonstrate a finished workflow in the conversation. The exchange is an exploration of direction, and treating it as a shipping announcement would miss that distinction.

    That same care matters when the discussion turns to agentic AI. Hopper describes a helper that can carry out tasks, gather information, and, in some arrangements, delegate work to other agents. Her example is finding visual references during brainstorming.

    Do connects that to building a mood board. Instead of manually finding every reference, the designer asks for material matching certain criteria. The system handles part of the retrieval, while the person evaluates what belongs.

    Hopper immediately adds a condition: the result becomes more interesting when the designer brings personal material too. Retrieval is not a substitute for every source of reference. The designer's own contribution changes the collection rather than merely approving what came back.

    This is another version of the type-joke exercise, applied to visual research. Independent human input prevents the process from beginning and ending inside the system's suggestions.

    The best interface in this account is not simply one that asks less of the designer. It asks less unnecessary translation of the designer, while keeping the decisions that give the work direction close at hand.

    The boundary belongs in the workflow

    Hopper's position is neither blanket approval nor blanket refusal. She describes her responsibility as embracing technology with healthy skepticism. The phrase is less a compromise between opposing camps than a working requirement: new capabilities deserve investigation, and their limits deserve attention at the same time.

    That stance becomes personal when the conversation turns to training material and compensation. Hopper says she values Adobe's approach to licensing and paying artists whose work contributes to its models. These are her stated reasons for feeling good about her contribution, not a detailed audit of training practices.

    The distinction matters. The discussion does not establish the full composition of competing datasets or offer evidence that would support a comprehensive ranking of their ethics or capabilities. Its substantive point is that Hopper considers the treatment of contributing artists part of her own professional responsibility.

    She also acknowledges using other generative tools in her work, including for editing rather than generating an entire piece. Her practice does not fit a simple division between those who use AI and those who reject it. The decisions are about application, boundaries, and what she can stand behind.

    Do raises the possibility that Adobe's approach constrains what its system can learn compared with competitors. Hopper does not turn the exchange into a campaign against every alternative. She describes her stance as a matter of choice and preference.

    That returns the conversation to the designer's agency. Choosing a tool involves more than assessing the attractiveness of its output. For Hopper, it also includes whether its role in the process and its relationship to creative contributors align with the work she wants to do.

    There is no named, numbered methodology here. There are, however, clear working boundaries in the examples she gives. Together they form a practical test for whether assistance is helping the designer or quietly directing the designer.

    • Use generated options to explore a concept, then assess them against the intended story.
    • Keep exact product representation distinct from an exploratory approximation.
    • Learn the fundamentals that make automated checks understandable and contestable.
    • Bring original sketches, references, and ideas into the exchange rather than only accepting suggestions.
    • Look for repetitive annoyances to remove without assuming every difficult creative decision is waste.

    The test is not whether a task can be handed to AI. It is whether handing it over preserves what makes the task valuable. In Hopper's examples, generating another view, collecting references, and checking a guideline each serve a different purpose and demand a different standard.

    For a designer moving toward art direction, that distinction becomes more important, not less. More options do not arrive with their own priorities. Someone still has to decide what expresses the concept, what satisfies the brief, and what should not proceed.

    Hopper's skepticism is therefore not a defense of unnecessary labor. She wants less tedious work and more direct interaction. She also wants the designer to retain the understanding that makes a decision meaningful rather than merely available.

    The practical challenge is to inspect the next impressive result before building a process around it. Identify what it solved, what it only approximated, and which decisions still require deliberate control.

    Let the machine offer a direction. Do not confuse the offer with a reason to take it.

    TRANSCRIPT

    The nature of AI is to recognize and repeat patterns. By its nature, it's giving us patterns. And if you are only following the patterns, it's training you as a human. And I think as designers, our challenge is to do things like breaking the rules. Hey, my name is Brooke Hopper, and you're listening to The Future. I have a rare opportunity to speak to somebody very high up in Adobe to find out what is going on. And Brooke, welcome to the show. Thank you. It's a pleasure to be here. Tell me a little bit about you for people who don't know who you are. What is your role at Adobe? Sure. I'm a senior principal designer at Adobe. First one ever, actually, on the Adobe design team. But I lead a lot of thinking around what is the future of creative tools? And specifically, my contributions have been sort of bringing Firefly into our creative tools, thinking about what the role is, and honestly, just making sure that it's useful in creative workflows rather than no one wants to generate just something. We all want it to be useful. And so that's really how I think about that. And where do we go in the future? So we are to give you credit or the blame you at Firefly is what we love? You can do both. Okay. I'm very, I'm actually, one of my big passions is really around helping people do the things that they really want to do, whether that's using AI as a tool, whether it's just doing it on your own, it doesn't matter. Really, what I view my role is I'm here to help people do what they love and also make it better. So one thing I think about that's really remarkable about Adobe is how intentional you are. AI as a tool, whether it's just doing it on your own, it doesn't matter. Really, what I view my role is I'm here to help people do what they love and also make it better. So one thing I think about that's really remarkable about Adobe is how intentional you are and thoughtful in creating AI tools because it's a tightrope. I think you're walking in that you're integrating AI technologies, but you're for creators. And so some creators are very emotionally not cool with AI. What is the philosophy behind what you guys are doing and how you decide what you're going to make or not make? Yeah, I mean, you'll hear people say this a lot as we view AI as a tool, but it's actually more than that. I've been in Adobe for 10 years. I spent seven of those years building drawing apps. So I'm very close to the illustration community. A lot of that community has real concerns and I think has legitimate concerns around what is going to happen to their craft, their art, what comes next. And that's really a lot of the things that I think about. And I think it's important as a designer, as someone who's building creative experiences to hold that intention and to understand both of those sides. And so when we bring this into our tools, it's not, again, you're just generating a thing. Adobe has been around for 40 plus years. We've been building creative tools that entire time. And so Adobe knows precision and control better than any other company. And we know that when you're trying to make something, pixels matter. The tiny micro adjustments are actually the thing that matters. Anyone can do big, sweeping, gorgeous designs, but when it gets down to it, I'm actually wearing a shirt right now that says control freak. Designers are control freaks. We have to have everything in a way that really makes sense. And so it's all about the craft. And that's really what we're leaning into as we're thinking. just designs. But when it gets down to it, I'm actually wearing a shirt right now that says control freak. Designers are control freaks. We have to have everything in a way that really makes sense. And so it's all about the craft. And that's really what we're leaning into as we're thinking about what is the future of design tools? What is the future of illustration tools? What is the future of video? How does this stuff all play together to make your life better? Creative tools should feel creative. I like the integration. I like the love of control that as an early user of Firefly, I thought it was really neat. You're building a tool to do something versus just a wide open sandbox. I've seen it go through different iterations. I have to admit, sometimes I have a love and hate relationship with Firefly. Sometimes it does exactly what I want. I rush over, tell my wife who's a designer, like, babe, look at what I just did. And sometimes I'm like, God, I asked you to do this and it's not working. Do I take blame responsibility that I just don't know how to tell it what to do? You're not the only one. I get a lot of bad results. In fact, I got one up on stage today. You get bad results? No way. I think that's the thing is like, that is the nature of this technology. It is not meant to do exactly the thing that you want to do, which kind of goes back to the point where I said Adobe knows control better than anyone. It's going to give you something random and that might spark a direction, but you have control over that direction. You get to decide whether you take it or not. And you can decide if you want to refine it or go deeper. And that's what we're here for. And totally agree with you. I mean, it's a love hate relationship with Jenny. Sometimes it's magical and amazing and allows you to do crazy things. And other times you're just like, you know what? I'm just going to pause for today. Now I have to say this because I describe myself as an AI optimist. If you use the tools to help support what you do, you can work faster and you can ideate. I mean, maybe we transition into this idea of like, maybe once I was in production, but I'm Now, I have to say this because I describe myself as an AI optimist. If you use the tools to help support what you do, you can work faster and you can ideate. I mean, maybe we transition into this idea of like, maybe once I was in production, but I'm moving up towards art direction, possibly creative direction. What do you see the role of AI and what you are making specifically to help me achieve those goals? Yeah. So I actually see AI helping me on the opposite ends, right? Like in the middle, I still, again, you're going to hear me say this a lot, but like I need precision control and I need finishing tools. But on either end, when I'm brainstorming and like thinking about what this concept is going to be, I might be sketching, I might be doing things. Maybe if I'm a graphic designer, maybe I use Gen AI to help visualize a couple of concepts. And then on the production end, AI is really good at recognizing patterns and repeating patterns and looking at data and analyzing and help you find those trends. And so being able to use it for coming up with random ideas, something that I might not have ever thought of. And I think that's where it's really important to understand what does the machine do well and what do humans do well? And understanding what your value is and what you bring as a designer, as a creative person to this process actually makes the whole thing so much more clear. One exercise I actually love to do with students, I get asked to into a lot of universities to help introduce students to Gen AI and what it means to their workflow, is I give them a project and I say, we're going to do this entire assignment using only Gen AI. You cannot use any other tools. And a lot of them don't like it. But what it does is it actually helps them understand where does this tool make sense and where does it absolutely not make sense? And I think that's some of those things, you know, we can get into curriculum and education. And a lot of them don't like it. But what it does is it actually helps them understand where does this tool make sense and where does it absolutely not make sense. And I think that's some of those things, you know, we can get into curriculum and education and how that plays into where things go in the future, which I think it plays a very important role. But a lot of it is embracing it and helping the students understand where does it matter and where does it make sense just to like do it the way that you've been doing it for a while. Because there's a place for both. I talked to the director of a graduate program at SVA and he talked about how it's very important for him, for his students in the product design department to sketch, then use AI to iterate, render different materials and explore really quickly. I love it for its brainstorming capabilities. Are we getting close to that part where, OK, I love the idea of it. Now I want to be very specific and ask for certain things that happen that Firefly, Photoshop, can help me get there. I think we're getting there. But again, just based on what Gen AI is and does is it's not a technology of perfection. The whole idea around it is it's taking something and it's making a new thing every time. And so when you're working for, say, you're an ID student, industrial design student, and you're working at a product company, you're not going to be generating that product into a bunch of ads. Yes, maybe in the ideation phase or when you're brainstorming and trying to think of how it might look, you know, sort of in situ situations. But at the end of the day, that company wants their product, not a generation or an approximation. Close enough is not good enough with a lot of those things. And so I think that goes back to use it where it works and where it makes sense. So, for example, I have a good friend who's an industrial designer. a generation or an approximation. Close enough is not good enough with a lot of those things. And so I think that goes back to use it where it works and where it makes sense. So for example, I have a good friend who's an industrial designer. And a lot of times he's asked to provide sketches to clients. And then not necessarily like a full turnaround, but they want different angles of what say a camera might look like. And his dream is I want to do one or two sketches and then Gen AI can generate the rest of those things for me. And we're at a place where that can happen. And so there are little tweaks in the workflow of just really addressing the things that are annoying or driving you nuts. And it's finding those little places where there's a catch or a hiccup or something just slightly annoying in the workflow that can make all the difference in the world. If you had all the power in the world and you can make anything happen, what's what would you wish for Firefly to be able to do now? I mean, this is very broad, but like what I would have it do is just like enable everyone to feel fun being creative. I think that this is something that I see a lot is we talk a lot about like people need content. We're creating more content. I got into design because I'm really passionate about it and I love it. And creating more and more and more content at a faster and faster pace isn't, I'm definitely not passionate about that. I can tell you that much. Like, like I want to have fun creating. And so if all it can do is just like take care of the things that I don't want to do so I can spend time coming up with fun ideas or creating a really gorgeous interface, or I'm a graphic designer at heart. I call myself a recovering graphic designer, you know, make some gorgeous packaging, like the world would be better. Okay, I'm not gonna let you off the hook. You gave me an answer. really gorgeous interface, or I'm a graphic designer at heart, I call myself a recovering graphic designer, you know, make some gorgeous packaging, like the world would be better. Okay, I'm not gonna let you off the hook. You gave me an answer, but that was not the question or the answer to the question I asked. I would really love to hear. I mean, people ask me weird questions like this all the time. I'm like, this is what I would do. Yeah, you do. Yeah. Is it okay if I push a little bit here? Yeah, yeah, you can. I mean, I can, I can share a couple of the things. One killer idea. The thing that I would love is, when we're using tools, just the nature of tools, there are limitations. And limitations can be good. Friction is good. That's a, it's an intentional part of play. And I've been thinking a lot about introducing intentional friction in the use of AI, because it also allows you to feel ownership of it. But then there's also areas where there's friction and it's not intentional. And so one of the things that I was thinking about not too long ago was we've introduced this idea of mask tracking in After Effects and Premiere, something where you can like select a piece of the video, and then we'll automatically propagate it out and follow that section. Well, what if I could just click it and change it? How crazy can we go with some of this stuff? And crazy is probably not the word. But like, what if I have two videos, and I can just click something from one video and drag it into another? What if we could treat video editing kind of like image compositing? Or how can we push the boundaries on some of this stuff to make it feel more natural? We live in a 3D world. And I think just the nature of how we work, on some of this stuff to make it feel more natural. We live in a 3D world. And I think just the nature of how we work, being in a 2D surface feels less natural. And so I think there could be a really cool opportunity to somehow allow our creation experiences to be a little more immersive. And I don't even want to say intuitive, but physical in a way. Like the first thing you do when you're a kid is you draw, you pick up a crayon. That's the most natural form of creativity. And so how can we make something like video editing feel as natural and intuitive as picking up a crayon when you're a two -year -old? I don't know. I don't have the answers, but like that would be pretty amazing. Imagine being able to like take an image and just turn it and rotate it and interact with it in ways. And just, I think there's a natural, intuitive way that you can take objects and move them and adjust them. And I'm speaking purely more video and image compositing at this point, not necessarily UX design, but the UX enables you to interact with things in a way that feels more how I could pick up this water glass and take a drink of it. Since I'm a pop culture guy, is it like what I think Blade Runner is or was? Where, you know, they do this like zoom in and enhance, but it's like, how are you seeing around somebody's shoulder? If I had the ability to take an image, any image moving or static and say, you know what? I'd love to reframe this. I want to change the camera angle. I actually want them putting down the cup instead of picking it up. And that would save my butt when I'm like shooting. That technology exists today. It does. And is this something that we're going to see more of and more refined? And this is where it gets real tricky, right? I have done those things where I've fed it. I'm like changes and it just goes bananas. That would save my butt when I'm like shooting. That technology exists today. It does. And is this something that we're going to see more of and more refined? And this is where it gets real tricky, right? I have done those things where I've fed it. I'm like, change this and it just goes bananas. Yeah. There's extra body parts. There's all kinds of weird stuff happening. But we're marching towards that. We are. And, you know, again, as Adobe, we know how to control those types of things. And I think a lot of times when you think of Adobe tools, you think of Illustrator, Photoshop, Premiere, After Effects, kind of all the standards. And a lot of people forget that Adobe also has a huge suite of 3D products. And those 3D products can do a lot of things. And when you combine that with generative technology and then combine that with some of our traditional tools, you can imagine how this starts to play together and come to life. Are we talking about extracting a 3D model from a frame and then being able to rebuild it, retexture it, and put it back in? Absolutely. Okay. Is there more you want to say about that? No, I just I think that your question was around if I could do anything with Gen AI, what would it be? And that's kind of what I'm getting at is like making things as intuitive as working with something in the real world. And working with 3D is not intuitive. I can tell you that as someone who has tried and failed to do 3D printing. But when we're working with 2D content in a 3D world, it's difficult. And so if you can take this water glass and just rotate the angle of it and it propagates throughout the video in a way and harmonizes itself, that's much easier. I am not a words person. I'm really bad at prompting. I don't think prompting is a natural way of creating for people who are visual. I want to be able to interact with things and move them around. that's much easier i am not a words person i'm really bad at prompting i don't think prompting is a natural way of creating for people who are visual i want to be able to interact with things and move them around and we are in a world where that can be possible where technology is allowing us to do things in ways that feel intuitive to us as humans who live and interact in a 3d world do you think people then become less decisive like i think one of the things is we get experience we're like this looks good change that and we kind of know it's not like we literally have to move it every single place and rotate 100 times to figure that out so are we marching towards that i love this question because the answer to that is the nature of ai is to recognize and repeat patterns by its nature it's giving us patterns and if you are only following the patterns it's training you as a human and i think as designers our challenge is to do things like breaking the rules and deciding not to do it and really understanding what are the fundamentals of design and your craft and you always hear the phrase you have to know the rules to break the rules that's very true and so people who don't know the rules of course they're gonna allow the ai to guide them through these patterns into a more possibly homogenous world that's where the beauty of creativity and humanity comes in is you have a perspective i have a perspective we're sitting here in the same room and we have very different experiences and you can tell a story through your lens i can tell a story through my lens and so it's making those decisions based on our experiences the stories we want to tell deciding that you're just going to go completely in the op is you have very different experiences. And you can tell a story through your lens. I can tell a story through my lens. And so it's making those decisions based on our experiences, the stories we want to tell, deciding that you're just going to go completely in the opposite direction. That's where the real creativity lies. And that's something that requires curiosity. You have to, again, understand and know the fundamentals of storytelling, design, et cetera. And creativity is definitely not dead or dying. I think it's very much alive. And I think the people who have that curiosity are really going to thrive. Weird question I'll throw at you, but I think given your background as a recovering designer, as an educator, and someone who's pretty high up in the Adobe food chain here, I'm trying to fix the world by creating educational content that teach people to be better without going through school and getting into debt. Will Firefly teach me? Can it teach me? Is that part of the agenda? Because it's seen a lot more than a person can see. Any person can see, right? And it knows what's supposed to be good and like theory and harmony and those kinds of things, rules of composition. I think it already knows those things. Maybe. I think it has to be taught. Let me frame it this way. There are some basics that this is good, this is bad. But really what makes design design is when you choose whether or not to accept those things, kind of back to those following decisions. And so, for example, if you're working with a brand, they definitely have brand guidelines that are good or bad. And so, if it knows that, it can check those. It can give you sort of general, I don't know if you were in the keynote this morning, but they demoed like, hey, I call it design spell check. Check some of these fundamentals for me. Maybe you want help making. good or bad. And so if it knows that, it can check those. It can give you sort of general, I don't know if you were in the keynote this morning, but they demoed like, hey, I call it design spell check. Check some of these fundamentals for me. Maybe you want help making those decisions. And I think that's okay. But I think that as a human, you still have to learn those things. And yes, it may teach you some things along the way. I think that's great. But I think ultimately, at the end of the day, you need to learn your taste and what taste is. I'm a little skeptical that you should be learning taste from a machine. I think it can guide you. I mean, I feel like that's my role as a creative person at Adobe is to embrace technology with a healthy skepticism. I would love with its vast data set for not only to generate what we want, but to educate us in the process. Like here's the first iteration. And this is why I did this. This is the most conservative version. And the composition is X, Y, and Z because you've asked for that. May I make other suggestions? And may we talk about these other things that we could do? And like it was talking to a very junior level designer so that the profession can rise much faster perhaps, or is that too crazy? No, no, no. I think, I mean, I think that's how a lot of people look at AI is, and I hear a lot of people calling it their intern. I'm a little shy of calling AI a collaborator that feels, I think, just for the time and place where we are today and the position where I'm in, where I'm working with creative people, collaborator feels a little too much for my personal taste. But that's really kind of what it's doing. I did an exercise recently with someone who, she's very much into typography and she got a group of people together and she said, okay, I want you no looking online. But that's really kind of what it's doing. I did an exercise recently with someone who she's very much into typography and she got a group of people together and she said, okay, I want you no looking online, come up with as many type jokes as you can just from your mind. And then we did that for five, 10 minutes. And then we used AI to come up with a bunch of funny, I put that in quotes, type jokes. And then the third exercise was, okay, now do this again and collaborate with the AI. And they're by far better because you're bouncing off of this thing and it's giving you ideas. And really that's what the creative practice is. That's what you're doing with Firefly is, yes, sure. You could take the thing straight out that it gives you, but is that really going to be the thing that you're happy with? Maybe sometimes, yes. But I would argue as a creative person, probably not. You want to take it and you want to edit it and you want to make it yours and you want to work a little bit for it. You want that friction, whether you admit it or not. Like that's part of the joy of being a creative person is like having to work a little bit for it. So do you artificially handicap the software to make it, to add that friction into the process? Honestly, I think that's the question. You know, there's a lot of times where there's a bug in a software that causes a point of friction and it ends up being like someone uses that to make something incredible and amazing and innovative. It's a feature, not a bug. I'm sure you know that phrase. And I think that's what some of this is. And the questions that I have is, if so, if we do introduce points of friction, where are those? Or should those just sort of like occur naturally somehow in the process? Final two questions. I think everyone just... of this is? And the questions I have is, if so, if we do introduce points of friction, where are those? Or should those just sort of like occur naturally somehow in the process? Final two questions. I think everyone's just catching their breath. Literally every day on Instagram, social media, oh, this new tool does this. And I'm like, oh my God, I cannot keep up with this. And there's this term being thrown around, agentic AI. I don't even know what that means. I think I know what it means. What is it? And should we be concerned or happy? Yeah. I mean, really, it's just the concept that we talked about was using AI as a collaborative tool to sort of come up with more ideas and help you do things. It's the intern concept. Really, that's, I mean, I think in the most basic sense, that's really what it is, you could argue. But really, it's the helper is kind of what it is. How's it different than gen AI now? There's a whole new term for this now. So it must be doing something different. It is slightly different. I mean, it's giving you context and it's able to produce answers. One agent is able to take things and deploy what they call sub agents to go off and do other tasks. You know, it's mini worker bees all working together. And it's most basic form, you can think of it as a chatbot. And then that chatbot can go and do other things behind the scenes while you're interacting with it or doing it. And so it's just bringing more information to you in a faster way. One really great example is say I'm doing some brainstorming. I can ask my chat, hey, go out and find a bunch of references that are like X, Y, and Z. And it can go out, search the internet in two minutes and bring all of that stuff, two minutes or less, honestly, and bring all of that stuff to me. Whereas if I were to go and do that on my own, I wouldn't number one. a bunch of references that are like X, Y, and Z, and it can go out, search the internet in two minutes and bring all of that stuff, two minutes or less, honestly, and bring all of that stuff to me. Whereas if I were to go and do that on my own, I wouldn't, number one, I wouldn't find all of that stuff. Number two, it would take me forever. It could take me hours. It could take me days. And so that's a really great example of an agent or a chat helping you work quicker and faster in a way that makes sense. The practical application for this would be like, I want to create a mood board or something like that. Like I kind of like this and find me a bunch of things and I'll pull it off the internet, build the mood board for me. That's an example there. Okay. But you could be pulling in your own. And I think, I guess that's where, that's where, again, I always insert my sort of point of view from Adobe is like, it can do all of that for you. But from a creative perspective, it's way more interesting if you're also bringing in your own stuff, kind of like the font joke. You can do it on your own. AI can do it on your own. But if you do it together, it's actually way more interesting. So suppose that we're living in infinite multidimensional reality with infinite timelines. The five -year version of you in the future looks back at this moment in time. What do you see happening? Is this like an inflection point for humanity, creativity, that this was that fork in the road and we hit it? What has happened? So your five years in the future, looking back, what have you seen that's either made you happy or really sad? I think for me, I say this quite often with the industry I'm in, being a creative person, working in generative AI. I mentioned before, I like to have a healthy skepticism. I think that's really important. I think I'll be really happy with the way that I am helping and my contributions to how Adobe is approaching this. I care a lot. and generative AI. I mentioned before, I like to have a healthy skepticism. I think that's really important. I think I'll be really happy with the way that I am helping in my contributions to how Adobe is approaching this. I care a lot about the creative community. I care a lot about education. And the fact that we're doing this, we're training models by compensating the artists for their work. We're licensing everything that we train on. We're not training on IP. That makes me feel really good. And that makes me feel really excited about what the future can be. And I'm happy to have contributed to it. I think that's amazing that Adobe's taken that line. I also feel like sometimes it's hurt Firefly in terms or the AI engine to learn because I don't think the other companies care that much and they look at everything. And so they have a much richer data set of images. Do you feel like that is accurate or? Yeah, of course. Okay. But you're going to fight that battle on ethical sword or whatever to make sure that line isn't crossed, right? I don't think it's a battle to fight. And even for myself, I will use other generative tools in my work, but I'm not generating the entire thing. I'm using it for editing. And for me, being on the side of helping ensure that artists are paid and compensated and I can feel good about the work that I'm doing and that I'm helping the people who are contributing to this. I mean, that's really where it comes down to for me. To me, it's not a battle. It's a matter of choice. It's a matter of preference. And I understand that there's people who don't. It's not a concern for them. And I think that's totally okay. But for me and my personal viewpoint and stance, that's what I feel the best about. Great. I think that's a great way to end this. Yeah. Thanks for being my guest. I understand that there's people who don't, it's not a concern for them. And I think that's totally okay. But for me and my personal viewpoint and stance, that's what I feel the best about. Great. I think that's a great way to end this. Yeah. Thanks for being my guest on the pod. Yeah. Thank you, Chris.
    Download transcript

    Enjoyed this? There’s more where it came from.

    Get insights on creativity, business, and design from The Futur.

    You can unsubscribe anytime. By submitting, you agree to receive communications and to our Privacy Policy.

    “You're building a tool to do something versus just a wide open sandbox.”

    — Chris Do

    Key Takeaways

    Details

    Topic
    CREATIVITY

    Share