This is the Nielsen Norman Group UX Podcast.
I'm Tim Neusesser, and I'm not Therese Fassenden, who's usually the host of this podcast and the
person you might have expected. Therese is on maternity leave right now, and during this time,
Samita Tenkela and I, Tim Neusesser, will be your hosts of the NNG UX Podcast. Therese,
if you're listening, we send our warmest regards and eagerly await your return.
In today's episode, we dive into a subject that's capturing the attention of many,
artificial intelligence. We'll explore how AI is reshaping our personal and professional lives,
the evolving impact it may have in the future, and how we can stay ahead of these changes.
Additionally, we'll discuss the usability of AI tools and the journey towards making them
universally accessible. To explore this topic, I spoke to two exceptional guests.
First, we have Henry Mottesad, the head of design at Perplexity AI. Henry brings a wealth of knowledge
in AI, and he'll share valuable insights into the inner workings, the challenges, and aspirations
of Perplexity AI. Our second guest is Kate Moran, the Vice President of Research and Content here at
Nielsen Norman Group. She will offer a unique UX-centered perspective on AI. So, without further ado,
here are Henry and Kate.
I think it's going to be really interesting to have two different perspectives. Kate, she's
here at NNG, right? We're more the researchers, maybe have a little bit broader perspective and
try to kind of educate the whole UX industry about UI. And then you, Henry, are really working on an AI
tool and really designing an AI tool. So, way more hands-on. And I think it's going to be really
interesting to have these two perspectives meeting here today and discussing it. So,
maybe, Henry, you could kick us off and tell us a little bit about Perplexity AI,
the tool that you're working on, and maybe what's your core mission there?
Sure. Yeah. So, Perplexity is... We've kind of tried to describe what it is a lot of times,
and it keeps changing. But, like, the core purpose was to make a product that answered any question
that you might have instantly. And so, when we got this first version of Perplexity working,
it worked every time. And I was like, oh, this is what... This is this thing I've always wanted to have.
And so, we've been building just, like, functionality that we've wanted, that we wanted to use,
that we thought might be useful, and kind of built a mission around that, if that makes sense. So,
you know, by all accounts, we did everything wrong. But Perplexity is supposed to be
just the fastest way to get information. And what it's become is, I think, something even bigger than
that, which is, if you have all the information that's available in the world, and you have
yourself, and you want to learn something, well, Perplexity has become this bridge between
you and that information. And not just, like, you know, there's a basic version of that, which is,
you know, it finds summary, it finds moments where, like, where things, like, need to be kind of put
together or pulled apart. Like, maybe there's two viewpoints, and it will represent both of those
things. It does all that, but it also can, like, act as a, as, like, a translator, sometimes literally
from one language to another, or sometimes from, like, hey, I have a PhD in math. Tell me about
nuclear fission. Like, I understand a lot. Or I'm in middle school. Tell me about nuclear fission.
I don't know anything. And it becomes this, like, magical bridge between, like, you know,
all information world that's been compressed, and reorganized for you. So, sorry, it's a very
long winded answer. But it really, we really kind of did start by just building something that seemed
useful. And then we shared it, and then people started using it. And then we built a company.
That's awesome. You know, tools like Perplexity kind of remind me of the early days of
Google. So I was a 90s kid and a nerdy kid, kind of a techie kid. And I remember when Google first
came out, it was so much better than, like, Ask Jeeves. And I remember this feeling of, like,
oh, God, like, I can learn anything. Like, anything I'm curious about, I can find it. But of course,
with search, there's a lot of work in manual work involved with that information seeking process,
formulating your queries, evaluating the sources, scanning for the information. And it's really
exciting. Again, like, I kind of get that feeling over again with tools like Perplexity that are
so shortcutting that information seeking process. It's really exciting.
Okay, kind of, let's say, an updated or upgraded version of Google, right? It's a
better tool to find information. Henry, like you said, not only finding the information,
but sometimes also like processing it and translating it into something that it's maybe
more accessible for us, right? What do you think makes Perplexity different? And maybe sets you up
to be the Google of the future that everybody's using? Because there's so many AI tools out there,
right? How do you try to distinguish yourself from the other ones?
Yeah, absolutely. Well, there's like a, I could write like an essay on the answer to that,
I guess. But I think we didn't start the company to compete with Google or any search engine. And I
think if we had, that would have probably made us fail, right? We started by just trying to build
things that would be useful for us. And then we shared it and then learned what would be useful
for more people. And then kind of like kept walking down that road until people were like,
oh, actually, I can use this instead of other tools. And that's kind of like, I think the only
way. But the cool thing is that what I have conviction about is that what is the best thing
is to just get information as fast as possible. And that's been like a flag that we planted
really early on in terms of the product design and the engineering principles, right? Like it's just,
that is the most important thing is speed, literally, and how the technology works, but also
in the user experience. And that kind of has just being obsessed with that and being obsessed with
information being delivered to you, like in a way that would even be cognitively fast to process.
Um, that just has differentiated us naturally. And also like, there's a lot of ways in which people
do use Google where it just immediately feels like a lot of work, like you said earlier. Um, and so
the product kind of has become naturally that because of our, it feels funny to say like our niche,
even though it's like the, one of the most human needs. Uh, so it's like a very universal niche of just
wanting information and, but that's like, Oh yeah. Go ahead. I was just going to say, but that,
that is a niche though. If you compare it to something like chat GPT, which is like, what is
chat GPT a tool for kind of whatever you can think of. Yeah. And like, we know that when tools are,
you know, broader use like that, when there's kind of more ambiguity, it's often harder to design a
satisfying experience. Absolutely. So Henry, like what role do you think having, even if it's big niche,
having even a slight, slightly more narrowed niche has, has kind of played in perplexity success?
Well, it's, it's, it makes the design a lot easier. Um, but it also, the hardest part was
coming to that opinion and having the conviction to not like, I mean, it's really, it's a trap to fall
into when you're building a startup and going zero to one to like, look at a competitor and then like
make minor tweaks to it. It just never works. Uh, or if maybe it can, if your business is
completely different, right? Your operations work different or something like that. But when you,
when design, when product is going to be the way in which your business is differentiated,
you kind of have to just like ignore your competitors, I think. Um, but the, it's all the
details that change and, and yeah, you said it right. Like open AI, um, they're building an amazing
thing and, but they're building a different thing and they're building a platform that's going to
have to work for all kinds of use cases that we're not thinking about. And, and I'm, I don't know,
platforms are like one of the hardest, most complex design problems because you have to build systems
that are meant to work in ways you don't even know. And you don't even know how people are going
to use that. And therefore it's really hard for, um, any individual experience, uh, any,
any individual like vertical use case to be great. So then for us, it's like the, just being very
confident in that people want information as fast as possible and want to trust that information.
Well, then it's like actually quite a simple design problem. And that's why you, the product kind of,
in some sense feels quite simple. There's the sources and then there's an answer and it's not a
conversation. It just told you what you asked for and that's it. And delivering on that and,
and providing multi, uh, multi-modalities of information, right. When you will, we'll show
you texts, show you videos, we'll show you images, we'll show you maps, uh, just trying to like give
you instantly what you wanted and, and, and represent that information in a variety of forms. And then
it should feel kind of, you know, like it just works.
And it seems like you were saying you really were focusing with a design on making this as easy as
possible, right? Someone is seeking information and you want to make it super easy and that they
can access that information. And it seems like you've done a good job, right? I know you just got
your series B funding with, uh, 73 million. We have Nvidia involved, Jeff Bezos, right? So it seems like
it's, it's recognized then you're moving somewhere. Um, what do you think is maybe the next step with
this funding? What does it set you up for? Where do you see perplexity AI going with this? Um, how can
you make it maybe even easier to find that information? Yeah. Yeah. I mean, so the, my strategy for making it
easy at the beginning was, okay, this, what's happening here is actually really complicated. Um,
there's a search engine, there's all the AI stuff, multiple uses of it at different stages.
And, but when you just try it, it, it works. And if you, if you can just think of one thing and ask
it, you can see how the product is unique. And so my goal was to just put very familiar looking UI
on this new magical thing. Um, and also just like, I've, I've worked on consumer products my whole
career and I, and I kind of know how hard it is to get people to kind of do anything. And so I think
like the most successful UI design is it has sort of like a gravitational pull. You look at it, you squint
kind of like hit the brightest thing on the screen. And just if hopefully that aligns with what the
product's supposed to do, and then people can just kind of like tumble through it. I think that's the
best you can do for consumer design. Like people just have no, especially a free product. You've got
like 300 milliseconds of someone's attention before they don't decide that they don't get it or that
it sucks or whatever. Uh, so anyways, that, that has been like the most important, I don't know,
you like mindset, I think, and being humble, just being humble about that. And, um, but yeah,
we've, we've found a lot of traction, which is, um, obviously very amazing. And, and I think it's
been cool because I, we just had all this conviction about what the product, how the product should be,
and then using it ourselves, finding, finding it useful. And then it's just really like a,
how do we tell more people about it type of problem, which is a fun place to be in because
that's great. Then we just keep adding depth to the experience. And yeah, whenever I'm talking to
people about, um, the design of, of interfaces for these AI tools, um, I'm often pointing to
specific things that y'all have done at Perplexity as, as good examples. And I think that focus on
consumer applications and thinking about this being for like the general public has been a really big
advantage for you because, you know, if you look at these tools are, they're trying to get better.
I mean, we have like, you know, mid journey alpha coming out that looks like it'll have a,
maybe a slightly friendlier UI, but like these were clearly tools that were made for nerds
who can figure it out, not made for, you know, my grandma to understand.
Yeah, no, it's, uh, well, I mean, you have to remember that chat GPT was, uh, essentially like
a tech demo for the models, but it was never meant to be a consumer product. And what's amazing is
it became the most successful consumer product launch of all time. Right. And that just speaks
to how, how, you know, the novelty of, of this technology and, and, uh, of LLMs and these models.
Right. So yeah, nobody was really trying to make a consumer product like from the beginning, uh,
and the people, you know, thank, thank God for all these amazing engineers who are just like willing
to build something and, and share it and not worry about the UI because, you know, if you cared so much
about that, you probably wouldn't have built any, no one would have built any of this stuff. Right.
And I really believe in like being very scrappy with, especially with new stuff. Um, I love
beautiful interfaces and I love being thoughtful, um, but it can hold you back. Um, and we're in this
like really fun phase of experimentation and sharing and, um, and not everyone's even trying to build a
business, right. There's all this open source stuff. Uh, I think there's just an energy right now of
moving quickly and, and exploring and, and enjoying the, the discovery of new things.
Yeah. That's super interesting. You said you're kind of at that place in the journey where you have
to get more people on the platform and they start using it. Right. So now you kind of figured out the,
the technical aspect, of course, not fully, but a big part of it. And now you kind of have to make it
usable, right. Maybe some of the biggest challenges that you discovered there and that you see in the
future as well. Yeah. Really make it usable for like everybody, right. Just like Google, almost
everybody knows how to use it, but how do we get AI tools to a similar place where everybody all
around the world can just go in there and like, figure out how to get this information, how to
translate it, like you said earlier, or maybe how to generate something by themselves.
Why it's hard is we still have the early users, right? We still have people that deeply understand
AI and they maybe use chat GPT and us, or maybe they're like, even there's some even deeper tools,
LLM playgrounds. And they're like, man, I really wish I could, you know, up the temperature on,
on the model or whatever, or even switch the model or, or maybe show two outputs simultaneously.
And it's really hard to be like, sorry, we're not really building that. Um, but then at the same
time, I I've done user research or talk to users and, and, uh, there are buttons on the screen.
There's not that many, and there's buttons on the screen that they don't even see. Right. So it's like,
this is the, the, the major challenge of balancing at a basic level, like new user power user spectrum.
Um, we have that, which is like, I mean, a blessing to have the spectrum available to you,
but it makes it hard. And so, yeah. And what, what's so cool, I guess is so the dream, the dream
product, the dream interface system that I've always wanted, right. That anyone probably wants as a
designer is to be able to show everyone the right interface for them, right? Like if you're a new
user, you kind of like get easy mode. And then when you're a power user, you somehow get power user
mode, whatever that is, however that manifests. And, uh, the closest I've ever seen to this is maybe
in video games, but what's cool is I feel like we're probably going to get some of that pretty soon
with how, uh, even when our product, we have some generative interface moments, um, where it like
all I, as the designer, all I defined was like, there will be, you have like dear AI, you have the
choice between text input, radio button, things like that. And you pick the right interface to show
like we're giving it the choice to then present the right interface. I can totally imagine that
is just like the beginning of a reality where like we can show people, we can kind of like let the
interface unravel itself and the capabilities unravel itself as people use it. And it can be a truly
personalized experience. That's like, it sounds a little bit, it sounds like, yeah, on one hand,
exciting, kind of terrifying to like design for. Um, but I, I have to, I really believe that's going to be
the best thing. The coolest thing. Okay. On a casa counterpoint, I was in the middle of nowhere
in Canada, like six hours from a major city wearing a perplexity sweater and some, uh, a group of four,
uh, people that are like my parents age came up to me and they said, are you, Hey, are you affiliated
with perplexity? And I was like, what? Uh, and they're like, we, we, uh, and I say, well, yeah,
I work there. And they're like, they were shocked to see me in this small ski town. And I was,
and I was shocked that they're talking to me about it, but they're like, we use it.
We like that. It's fast and simple. And I was like, okay, well, I'm done. I did it.
That's the, that's a great feeling. Yeah. I can't believe it. So anyways,
somebody's figuring it out. I want to talk a little bit more about this idea of generative,
um, interfaces because I've been kind of waiting to, to see this appearing more in, in our field.
And, um, and it's really exciting to hear that you guys are already kind of working on that. Is
that happening in the copilot feature? Yes. Yeah. Okay. So for, for listeners who aren't as
familiar with perplexity, one of the things I really like about perplexity that I think they're
doing well is they have this copilot feature, which you can enable that if you submit an
information request and it has clarification questions for you, it will ask those questions
before it runs that prompt. And I think that's a great way to help people because we're seeing
from the user research, you know, general side, we're seeing a lot of people really struggling to
know how to use these things, how to write their questions. Um, so you get a little window that pops up
that says, you know, uh, you know, you asked about, um, building a DIY, building a sauna and are you
talking about an infrared sauna or, uh, you know, and so you can, you can narrow your question that
way. So can you tell us a little bit more about that? So like how, how that generative UI is working
there? Yeah. I mean, this is, uh, one of our, I think most interesting features and it has,
it's a very nascent concept, but yeah, the idea is that you, you ask a question,
and then it, what it, it tries to figure out whether if maybe it would benefit, you would benefit
from more precision and it'll ask you questions back. But what's cool is it like you can ask
anything and then it's going to ask, it's like, it feels like it learns the subject of what you
just asked and becomes an expert in like, you know, seconds and then produce, you know, like you
said, you're like, Hey, I'm going to build a sauna. And it's like, okay, well suddenly I'm a sauna
expert. Do you want an infrared sauna or whatever? Um, and I, I'm always kind of surprised when I ask
like something very niche, um, that it does that. But, but the other part of it is, so like I said,
I mean, the core product principle is low cognitive load, uh, and typing sucks, especially on your phone.
Like why, why should people have to type so much? And it's also not like the best way, uh,
it's not, it's not the, it's not necessary. Uh, and it's also just not the best way to capture
information all the time. It's not the most efficient interface. So like, for example, let's say
I was trying to plan a trip or buy a plane ticket. And it's like trying to figure out where my,
what, when a date picker is a great interface, why would I type something? Um, you know what I
mean? So it's just, there, there are plenty of things that are solved that we don't need to throw
away. And so what Copilot does is it, when it tries to figure out what you want and maybe help you get
drilled down a little bit more and it will use, it has the ability to decide to use different
interfaces, uh, to more efficiently, uh, extract information from you. So sometimes it'll radio
inputs. So, yeah, that is so cool. And it's just like a small, a small step in that direction,
but like, I I'm really excited to see where as, as you know, a UX and design field where we can take
that and like where, where that ends up. Like it does, as you said, opens up the ability to on the fly
design interfaces that are better suited to the individual. And I'm thinking about something you
mentioned earlier, Henry, that, uh, perplexity can tailor the response based on your experience and
your knowledge, which is not something I had, I had thought of or tested or ever realized about
perplexity. So that's awesome. And that's another thing that makes me really excited for the future
of information seeking because that's always been a challenge we've seen in research where,
you know, we have somebody who has a PhD in oceanography, he's searching for something.
He's seeing a lot of stuff that's meant for middle schoolers. Um, and I, I'm also thinking
about like, these are some of the major challenges that UX and design has faced for as long as I have
been in this field, which is coming up on 15 years. And it's, it's exciting to think about
a future where we can actually tailor the UI to individuals.
Yeah. And I love what you said earlier there, Henry, you're kind of taking small pieces from
everywhere. Right. And Henry, you said, we see something in a video game. We take something
from there. We see, we know radio buttons, so we know other UI elements and you kind of just combine
it and mix it together in a novel way on a new playing field. Right. Yeah. But I think that's
probably the best way to make new tools usable because then people are already familiar with it
and already figured out and have used them hundreds of times. So they already know how to use it. And
then it's just the kind of the mental model of your, your tool to understand that, but they don't have
to understand the UI like a hundred percent or they, they already know most of it. So they don't
have to fully relearn it. Right. Absolutely. But with that, how do you think you can further help
people to actually understand like how to even get better in using these AI tools? Because I've always
seen these like long lists of how to write a prompt, how to write the best for, I don't know how to
build a sauna, right. Or, or something else. And you said the first step is maybe having your AI
platform already asking, like, can you give me some more information, but what's maybe the next step
after that? Yeah. Well, I think I kind of, I kind of treat it as a normal consumer cold start type of
problem. Um, I mean with, with prompting, I think prompting is like the worst software experience
ever. Uh, I mean, if you've ever tried to generate images, it's, it's awful as an experience,
but the payoff's amazing. Right. So people do it, but we're, it's just, we're in a small
blip of software experience where this is even going to be a thing. I just, there's no way it
survives. Like when we actually, we have image generation on our product, but it's just a button.
It's like, you want it to look like an illustration and you click that button. And then it, it does,
there's a lot of prompting. I spent hours of my life picking, picking the prompts, but I don't
want users to have to do that. Um, so yeah, I think to, to make it intuitive, like I said,
I kind of take a consumer product approach, which is you want to, you, people need to feel
somehow bought in to first, you need to understand what this place is. Again, you've got like
milliseconds here. Um, and then ideally you can look at some example content right away
to, to, to build that mental model and start to be modeled, like the behavior of the platform.
Uh, so, you know, when you boot up TikTok for the first time, you've never seen it, heard of it.
It's pretty easy to figure out what it is. And you don't learn how to make a TikTok by jumping
right into the creation flow. You learn how to make tech TikTok by watching 20 TikToks. And
maybe like when you get in there, there's a little bit of help with, I don't know, I've never done it,
but I'm sure there's like some little pop-up maybe, but I don't think any of that actually works
because when you watch people use software, they just click on stuff really fast. Nobody's reading
anything, but just trying stuff, um, which I think is totally fine. Um, and so for us, like
I mean, if you go to our homepage, when you're, when you're logged out or making a new account,
there's just like a couple of things you could click on. Um, and you can type in and it feels
familiar. Hopefully you can just type something, maybe how you would use Google or chat GPT.
Um, but then there's a couple of things you can click on that are examples. Like, and it's like
this dynamic thing and just questions, maybe one of those, you're like, okay, I guess I'll see what
that answer to that one is. And then hopefully you get it because there's a lot of things we could
have put on the homepage that we didn't. There's a lot, it's all in the end, I think distracting.
I think my goal is just to get you to try it. And I'm going to just keep thinking of ways to do that.
And I think another consumer strategy is around like getting, making it feel like you've got some
skin in the game or some kind of connection to it. Um, so maybe when you're signing up, you,
you tell us about yourself or something, or you, you feel like kind of some kind of profile that,
that does work as well. Uh, so you feel like, all right, I've, I'm invested in this a little bit.
I'm going to give it a couple extra minutes in my evaluation time because people balance if they
don't get it really quickly. Right. Um, that's just the nature of consumer.
I love it. I love what you said. Sorry, Kate. I have it right behind me here, right? Like you said
I have so many things I could add, but we really want to keep it to its core because it should be easy
for people to learn. Um, Katie, your, your turn. So, um, I love this conversation because we're
talking about mental models here. We've used this term a couple of times. I just wanted to
define it for anybody listening. Who's not familiar with this concept. So no, it's fine. Yeah. I mean,
our field is just chock full of jargon, which makes it so hard to get in here. Um, but so when we're
talking about mental models, we're thinking about the, the way that the user understands how this tool
functions and what it can do for them and how they're supposed to use it. So it's very much
connected to like, let a lot of, a lot of times what people are doing when they're forming new
mental models is they're leveraging prior experience with similar things. So, you know,
Henry, you've already talked about how you're trying to leverage things from consumer apps that
people have experiences with. So they kind of have a sense for how they're supposed to use it,
which is best practice for UX. We are really interested and we're doing a lot of research on
how people, especially people who are not us, we're not like techie, um, how they're experiencing
these tools for the first time and how they're figuring out what they are. So I think a big
advantage that perplexity has over something like chat GPT kind of related to the narrowness
is that perplexity can look like a search field because it does provide, you know,
this is an understatement, but it does provide roughly a search function. So we've seen in like
the diary studies we've done when, when people are new to chat GPT, they'll try to type in like
keywords because that's what they're used to. Yeah. Yeah. Yeah. Uh, that does not work.
Right. It has no idea what you're asking for, but with perplexity, it can probably,
it can probably guess. And then with that copilot feature, there's even that hand holding to say,
like, there'll be a little more about what you mean. Yeah. Yeah. We, we make this big assumption
that you're trying to get information and, and that just like, like pulls people over the hump of
like not knowing what it is to understanding what it is. Um, and then, you know, people, but people try
to use it like chat GPT and it kind of fails. It's like, it's not really there to chat with you.
Um, and it's not about outputs. It's about information. How do you think these AI tools,
they are changing everybody's life, right? What do you think is maybe a big difference in the future,
how we live and how we work through these, your AI tools? Yeah. Well, I kind of just think
like everybody is going to be empowered, um, by whatever they do there, there'll be something
that like accelerates them. Um, whether it's, if they're doing research, I mean, for us, it's really
all about learning and research. And I'd like to think that that process is accelerated. And also,
maybe there's new things, new behaviors that might open up that like, you didn't even know you could do.
Um, like the, the one that we've built that is still kind of, I'm sort of trying to wrap my head
around is you can with on the, on our app, you can take a photo and that, and have that be your
question. So I can take a photo of like, I don't know, like there's a, I've seen people use it a
bunch of ways, but like, you know, take a photo of, of my dog or something. And I'll be like, um,
what, what's a good name for him? And like something like that. And it like figures out that it's a dog
and it figures out that he's a brown schnauzer and it's like, what about Coco? You know? And it's
like, Oh my God, you know, there's stuff that like, no, I didn't. I named him Rupert, but, uh,
also a great name. Yeah. Uh, but you know, it's like the, and I've seen people take, um, a photo of
the whiteboard and interface, and then they'll take a photo and they'll be like, what's the, write some code
that makes this. Um, so there's a lot of things that like, I don't even, we will definitely, we are definitely
doing a lot of, let's just build it and see what happens kind of mindset, uh, less about like some
vision for the future of anything. It's just more like, I don't know, this, this could be useful.
Maybe, I don't know. There's a combination of stuff where I'm like, this is definitely useful.
And then another, another set where I'm like, I have no idea what to do with this, but somebody
will come up with some ways. And that's the cool, coolest part about building a tool. You don't have
to think through every use case. Uh, which is like the polar opposite of how UX used to work.
Right. Yeah. Um, yeah, that is, that is really exciting. And I, you know, I'm thinking back to
something you said at the beginning of this chat, Henry, where you were like, well, we started with
this purpose and then that kind of became something else. And we just keep, you know, redefining ourselves
and, uh, you know, talking about how you're using generative interfaces. Um, I think this,
this technology, the pace of it, this development is just so beyond anything we've seen before
that as a company, as an AI company, you probably are going to have to keep redefining yourself
as, you know, things change. And that's such an exciting place to work and like kind of context
to be in. Yeah. I mean, I I've been through this already multiple times in, I mean, to be clear,
we've only really been a product for a year. Um, and we built a lot. And, and if we have a strength
as a team, it's our agility and, you know, kind of our open-mindedness to, okay, well, let's try this
and build it fast. And, you know, I, I care a lot about craft and so I'm going to, I'm going to execute
on it to the best of my abilities. Um, but yeah, it's, there's like the, the image thing,
you know, we, we find out that it's a capability and we're like, what should we do with this? Um,
how is it going to work in our product? Um, could it, does it make sense in our product?
Cause there's some stuff that doesn't, and we won't, we have to have some kind of,
you know, idea of who we are and what we're, what we want to be great at. But, um, yeah, I mean,
I, I expect the market there there's the technology, there's the market, they're all nuts.
Um, you know, there's a massive, massive competition happening with millions of dollars.
And we're just kind of like in the room, um, trying to make a useful product, uh, which is,
um, I don't know. It's a nice kind of clarity about who we are. Um, and you know, on our, on our product,
you can, if you pay for our pro version, you can, you can use open the eyes models.
You can use, um, anthropics. Um, so we're, we are, we are really more focused on just building
something for that's like a product and a brand and something that people want to use and building
just like the way you would build a normal consumer product company. Um, so I think if we were, if I,
if we're trying to compete head on with, you know, the, these massive AI research teams, like,
I don't, it would be much more stressful.
I think that's your competitive advantage in a lot of ways, because like we saw after
chat GPT exploded in a way that nobody predicted, we saw all these companies suddenly rushing like,
well, we've got to get
AI and there's a lot of like copying going on between the competitors. And so people aren't
really coming up with anything that's innovative or different. And I think that's part of why
Plexity stands out. Yeah. It's, it's actually wild. The development we've seen in the last,
I don't know, 12, maybe 15 months, right? Yeah. Okay. What, what do you think is maybe our role
as NNG in there, right? Because we've never seen a change so fast in these tools and now everybody's
using and the world is changing really fast for all these designers, but also for all the consumers
as well. So what's our role in this? To be honest, it's been kind of chaotic, you know,
usually like NNG, we kind of pride ourselves on being sort of like a pseudo, like a bridge between
kind of academic style research and practical style research. And we try to be a voice of reason
as much as we can. And kind of anytime there's new hype around something, we try to be measured with
our approach and, and, you know, not make kind of like sensationalist claims. And I think we're still,
we're still trying to play that role and we, we are playing that role and we're trying to help people.
Um, you know, something that, that Caleb Sponheim, one of our, our coworkers at NNG,
um, who's actually creating, uh, an, our first AI course right now, something that he, he said to me
is, you know, there's a lot of noise about what AI is going to mean for UX and there's not a lot of signal.
There's, there's all these prompt guides. There's all these like sweeping statements,
like, you know, just like that, oh, this is going to change everything. And I'm guilty of,
you know, getting overexcited at times about that kind of stuff, but we're really trying to
move as quickly as we can, but also make sure that we're giving people practical advice that's
actually based on reason and what we have in front of us.
I think that the thing I feel confident in is that not everything should be a chat.
Actually, most products, most products should not be a chat. And most products,
I think this is another take is that most products don't benefit from having like an anthropomorphize
AI concept. Like this technology, there's a lot of, there's actually this tech AI, right?
There's a lot of things happening. There's the generative images, videos, sound, there's that,
there's the LMs. It's just going to be available every in everything and in everywhere. It's it,
it'll be kind of just like a way to enable some core product experience.
It's automation. Yeah.
Yeah. And, and, and it'll make some new software that that's amazing. And it'll make
some old software and just like accelerate it. And yeah.
I agree with that. Yeah. Definitely not everything needs to be a chat. We're seeing
like a lot of, you know, these poor executives at these companies are being told like, you have to
put AI into this product. So they're like a chat. Cause that's, that's kind of like the example that's
held up before them. Yeah.
I definitely agree, Henry, that like the anthropomorphizing is not necessary in every
context. I actually have an example, kind of a LinkedIn debate that I've been in this week is
around the concept of synthetic users. There's actually a company with that name,
but there's this debate, you know, do we need to have user research if we can have AI pretending
to be people giving us feedback as if they're human? And what I'm kind of stuck on is first of all, no,
I don't, I don't think that's true. And second of all, why do they have to be synthetic users?
Why can't it be, you know, why can't you train a model on, on how experts evaluate systems based
on heuristics and like proven best practices and then present it that way? Why do you have
to pretend like it's a mom in Georgia giving feedback?
Yeah. Cause there's an uncanny valley no matter what.
Right. Yeah. And the danger there I think is like, and I also think like this is kind of
causing some collective, like not delusion, but there's, there's some moral panic happening
in society as a whole. And there's a lot of fear that like these things are going to replace us in
various ways. And I think the, the fact that we're putting human faces on a lot of these things is
definitely contributing to that. Yeah. Maybe to wrap it up here, let me build on that. What
you just said, Kate, there's a lot of hype around it. Right. And a lot of people are afraid.
And I think part of it is what you said earlier, Henry, we make people, or especially you and all
the AI tools make people a lot faster. There's a lot of automation, right? So the world is changing
in itself. And the way we work is also changing. What is maybe something that people out there can
do to not be left behind. Right. Because I think that's a big part of this fear. Yeah.
Like, wow, everything is changing. Like what's going to happen with me in a few years, in five
years or in 10 years. All right. For Kate and Henry, what, what is something that people out there can
maybe do to like work against this fear? That's an interesting question. I, cause I, I guess I kind of feel
like the, I kind of feel like, cause I'm on the other side, right? I feel maybe more like if we
don't make software that's useful, all this technology goes to waste. And cause I feel like
the market will vote on us on anybody. Um, and so it's not, I, I kind of don't see it as like
people need to learn stuff. I'd see it more as like software creators need to make some stuff that's
useful and fit it into people's lives. It's like on us to present it and to grow it. And
like people, I don't think people are going to change that much. Um, there's always, there's
always people that are willing to try new stuff. There's always like an appetite. There's always
a long tail of people that eventually catch up. Um, so, but I mean, obviously like everyone should,
I mean, this is like maybe general advice, just be open to trying stuff, but I don't know,
like as someone who makes software, you've got, we've got the whole world that we need to talk,
tell about our thing that we've made. And, uh, it, we, we're never going to be successful if we
ask anybody to change who they are and how they feel about software. Um, so it's more like,
how do we tell the story about how useful it is?
I love that Henry. Um, and I agree with that, that, you know, just, just try it. Don't,
don't freak out. Don't be afraid of it. Just try it. I think that's really important because we see
that over time, individuals who are afraid of, of new products and don't at least try them and kind
of form these mental models around them. Those are the people that get left behind and they, they don't,
they're afraid of using technology. They don't understand it. So for example, my, my grandmother,
um, she worked, um, she retired kind of later in life and she worked in advertising where over time
she had to learn how to design things on a computer and use the internet. And eventually she started
coding HTML because she was just curious about it. And now she runs her garden club website,
which is amazing. Yeah. She's so, um, I think that's a good example of like, you don't have to
keep up with the pace at which things are changing unless it's your job or it's something that you're
interested in, in which case it's really fun to follow, I find, but you, you should use this,
you know, to keep up kind of, but also just because they're good tools that make your life easier.
For sure.
So I have a great example of this. I was just visiting a friend in San Diego who is a biologist.
She's a researcher and she's studying measles. And she was telling me one of my, the things I have
to do as part of my job is keep up to date on, on recent research on measles, but it's really hard to
find the time to find these research papers and read them. And I was like, yeah, why aren't you
using perplexity? I liked it. I was like, summarize the recent research on measles and like showed it
to her. And I was like, this could be like saving you so much time, making you better at your job,
making your life less stressful.
Well, Henry, Kate, thank you so much for being on the show. I really enjoyed the conversation.
I think I'm even more excited for the future of AI, more excited for the intersection of AI
and user experience, and hopefully talk soon.
Thanks, Henry. Thanks, Jim.
That was Henry Modisat and Kate Milwright. Check the show notes for links to anything
we've referenced so far. But also remember that we have thousands of articles, videos,
and reports on our website about UX design, research, strategy, and even UX careers.
That website is www.nngroup.com. That is n-n-g-r-o-u-p.com.
And if you enjoyed this episode, please follow or subscribe on your podcast platform of choice
and share your thoughts with us on social media. This episode was hosted and produced by me,
Tim Noiseser. All video editing and post-production is by the Larry Moore Production Company.
That's it for today's show. Until next time. And remember, keep it simple.