hi everybody today I'm excited to talk with Chloe Barrow Chloe has been living at the intersection of Design Technology and AI for some time and sharing a lot of her work and learnings publicly in this fun and Visually interactive chat we're going to talk about how AI has changed and will change how we create visual content how it applies to user interfaces and how a lot of our creative roles will need to adapt this can be a fun one so let's get started Chloe it's absolutely great to chat with you and we're going to be having this in a pretty virtual environment and doing all kinds of interactive things on screen so it's a whole new format so I'm excited to try this out and to learn

more about you and your interests and how you got started and all the fun things that you do so before we go on can you please introduce yourselves

yeah of course um so I'm Chloe I kind of have a hard time categorizing myself into any one thing I'd say I'm a designer engineer also a product person I'm just very interdisciplinary by Nature I'm currently wrapping up my masters at Stanford I'm very interested in the intersection of computer science and art

really cool so right now I'm currently looking at what looks like a portfolio of various things you worked on you know can you walk me through it

yeah I figured it would be helpful to share some of the tools that I've used over time um for on the design side which is something that I've been doing since 2018 UI ux design um it started off with Sketch figma and vision and now my workflow incorporates things like mid-journey and so I include some screenshots as well I think in the last year I've been Amazed by the number of tools that can use a form of Engineering in the art practice whether that's mid-journey using a pen plotter or doing things with more Graphics like in blender

so before we get even there let's go all the way back what got you interested in computers and in UI and engineering and all these fun topics

yeah ah making something from scratch you also get that with making art right like you take your crayon and paper and as soon as you make a mark you see visualize on a screen what you've done and I feel similarly with web development and just computer science you're really making something from the ground up and I think that's what's always attracted me but even more so Technology's impact on people because a lot of people now have cell phones you can make something that becomes accessible to a lot of people and that's really exciting to me as well

what's the first thing you ever remember creating

poof um creating in general well I mean I started with little Clay figurines to be honest in terms of Art in terms of uh like technical stuff I made a website to sell my jewelry um and that was using PayPal um and really janky janky website Basics that's that's actually quite pretty cool we're using a platform like Shopify or Etsy or we're building it entirely on your own and it's going to be PayPal at the end yeah it was just me and my sister doing it like Bare Bones and then something called store Envy existed at the time um where I then hosted my products basically the equivalent of Shopify or uh Etsy

so I'm looking at the screen and I see what looks like a plotting device like a mechanical plotting device um okay tell me more about what's going on here

one is called the axi draw and it's a machine that draws for you so in the world of automation instead of you having to go and use your bare hands to grab a pastel or a pen to make them work it will actually take your instructions and do that for you

that's really cool so your singers currently studying HCI or human computer interaction correct correct when I took you know I did not major in HCI major in computer science but when I took a few classes in that area because my hobby and interest were always around the intersection between development engineering and UI development it was very different back then you know we used Java we use swing you know this is like 2006 so it's like a quite a while ago at MIT the media lab and the thing we were always focusing on was like you know the basics of all the various techniques on how to make sure the UI is discoverable how to make this work you know in a nice way and over the years

though the entire discipline seems to have gotten much more formalized there's a lot more resources a lot more just but general knowledge on what to do here but AI has also changed a whole lot of it as well because nowadays I look at it I go I used to have to build a web page by running HTML CSS JavaScript and a framework now I just jump into framer or webflow and all of these have like really nice either tools to generate content for me I think canva does a lot of good things in this area as well so I'm curious to know like you know you're you're studying at write this intersection where a year ago half these things that exist and right now a lot of

things exist and by the time you graduate uh more things will exist as well how have you been able to like you know navigate through this massive amount of change especially in the field that you're currently studying in yeah

um well I mean I think you're right that prototyping has been become much easier and so I think when that happens you start thinking about higher level questions um of how like people interact with screens and systems which is cool um and you're also right that there's been a lot of change um I mean every month or so there's something new that comes out um and I think my Approach has been try to use it whenever I can so when I'm designing something when I'm implementing something I go to chat gbta your mid-journey to see how it can help me um and kind of build off of its first draft um and it's been surprisingly effective I mean for mid-journey I wouldn't say it produces like a polished output but it's

been really helpful in helping me generate ideas and brainstorm along the way um yeah I think whenever I can I try to see how far it can get me and then I fill in the rest that AI cat

and the part you're filling in I think is like the magic secret sauce that a human being can provide that an AI might not be able to do you feel that at some point those gaps that humans need to be intervening will also be replaced where AI ends up you know solving it in a nice way

I mean we're trying to do that right like we're trying to have multiple AI agents to give feedback to each other and kind of help each other fill in but I don't think it's ever perfect especially since AI has difficulty with context um and I think that can be really helpful for a human to be in the loop and to create I guess what people call alignment or just more feedback um so I don't think it's like 100 replaceable but I do think there's a lot of really cool work being done to minimize how much we actually need to weigh in as humans

yeah no that's that's a good point and because I always look back at the loads of abstractions that have been added even before AI existed you know the way the tools we use back then at least I never had to use it but I see examples of like latex being a Syntax for marking up documents for various activities now we just use other Microsoft Word or Google Docs and we just write naturally the way we think we should write and behind the scenes I'm sure the representation is horrifically complex but we don't worry about it anymore I sometimes do think that AI might fill in that same kind of a bucket where we'll be working on more higher order things and not worrying about the

basic areas of creation but the thing I also worry about is I'm a product manager I see what chat GPT can create for common questions and situations based on cost knowledge because for the most part none of us are like trying to build Rockets to go on outer space we're just solving existing problems in a new contextual way which unfortunately isn't that AI is fantastic at and so have you ever run into those kind of situations where you're like you know will my skill set even be valuable you know a year two or three year from now

I've I'm not as scared as maybe most I mean this makes me like a tech Optimist I don't know but um when I see the output from something like chat cheap but you're even my journey like you can tell I mean I can recognize it for Mid Journey for sure for chatty but it's a little harder and I think it's because it takes so much of like past data that it becomes just very stereotypical um and I think where we as individuals can be helpful so we each have our unique perspectives a lot of subjective Aesthetics and opinions on how text and tone should exist um and so that specificity really helps and I think that's where I can continue to um find myself valuable in the process

um I'll kind of stray it into more of a niche than it's currently doing

so you mentioned mid-journey should I go to the mid Journey section in the figma

file sure yeah um and actually we can go to the AI tools sections foreign yeah I don't know if I should assume that people know what mid-journey is but so I'll just briefly describe um it's similar to Dolly if people are familiar with open ai's Dolly um it allows you to take a piece of text and translate it into an image using Ai and mid Journey specifically for visuals they don't do text to text at least for now and a lot of artists and creative people have been interfacing with mid-journey which exists on Discord because of just how much it can help with the creative process and so I have an example of that here yeah

yep so here you are describing in Fairly good detail a soda can and then describing all the various levels of output you want the system to create and it created essentially four variations

yeah and I think this is where you kind of see the progression of trying to understand and Wrangle these AI tools at first I started off with just like a simple sentence that I thought was descriptive and then I realized that people are putting in so much like detail and Nuance in these prompts to try and bring more life to the image and so then I started doing the same and you can see it's like almost a paragraph of descriptions that aren't exactly coherent they're really just words strung together um to try and enhance the prompt and I think what's interesting for me to see now is that sometimes that's not sufficient like looking at these images they're okay knowing that they're made by AI but they're not something I would

actually put out in the world um and what has helped is using reference image reference images which if you go to the right um I provided one and that to me at least looks so much better so what I'm doing here is providing a image that I think I as a human think is what I'm going for and then adding a really like simple sentence to give it a little more guidance and it's just so much better than having spent a time writing an entire paragraph So stuff like that where you just like learn how to get what you want with the AI tool over time just by trial and

error it's almost as if you replace traditional programming with English language but prompt engineering and like designing the input that go into it is going to become in its own way a very you know like lucrative and like you know educational environment for people who learn programming

yeah yeah I mean I don't know that it's like exactly the same as programming but um it's definitely a new a new skill and I think a lot of it is experimentation and observational rather than pure logic which I find super interesting because I like working in those gray areas

it absolutely I've seen some like a lot of proposals for new schemas for making the prompts more formal formalized in terms of like what the output can be and it looks like a combination of like XML and Json but it was very interesting to see like the specs being created at least for I think I've seen it for like chat GPT where there's like a very formal structure for defined input and that to me look you know pretty comprehensive and also a little bit more of like okay I learned one more thing here which is exciting actually I love learning new things here so I've been filling with that and this is this even more natural though I'd rather describe in English just like as a human being

you know if I was conversing with that person how would I describe it this just feels like more appropriate yeah it does

feel more appropriate I think an interesting HCI question though is also like expectations versus reality right like when you start getting people to use just natural language they're gonna start talking to the machine as if it actually is a person and there are a lot of things that the machine won't actually understand that you think it does as a result of having a huge language human language interface rather than something a little bit more constrained

yeah there's always this expectation of the real human being that I might have said 80 words of which only 50 might actually be the real thing I'm actually looking for that dirty might be you know just context but not really important the machine at least as of right now may not know that so I might say like casually mentioned like I need to have like a reflective background and you're like no that he's kidding there's no way it was a reflected background but the system would be like oh reflect the background check let me go ahead and create a reflective background so it's going to be a little bit more of like this behavioral change in how we start expressing our ideas

for sure yeah and I'm excited for that to evolve I mean right now mid journey is All Tech space and I think the ways that we express our ideas especially for visual images is visual it's not in text and so if we can get to a point where we're like sketching ideas and using like basic sketches to guide uh the output that would really be cool and I mean that people are working on it mid journey is going to work on it um but yeah I think that's the next phase

the exciting part also is that you know I I grew up in a world where there wasn't Facebook or Twitter just yet you know just as I was like you know entering into like high school I saw like you know like Twitter was just starting to become more popular and then a lot of studies that I've been reading over the years I can't pinpoint any exact one but the the often said that because of the influence of social media Facebook Twitter and all these things people's behaviors have changed a certain extent as well like the things that they do and say in their typical day there's more things that are doing that's different than what they've done in a world where these things not exist

and I often wonder because let's say that we all become you know really good at defining prompts in a way that computers can really understand and so you start thinking about problems in terms of like okay I've had to give this to a computer how would I translate it into all these you know finite steps would that end up changing how you and I would end up conversing at some point because in my mind I'm like so trained to automatically jump to prompts as a way of like expressing what I'm trying to do and then when we all start changing our talkings style the words we used to be different will it tone be different spacing that's what I'm actually really curious about when this

becomes really popular how much will our own behavior have changed hmm

yeah I mean I think social media has changed our behavior and using a lot of AI tools certainly will as well um but it's hard to extrapolate because I think the AI tools are changing so quickly I don't actually know what format will stick if it ends up being very similar to chat GPT then maybe our communication will get more concise and literal maybe that will be an outcome um but I yeah I mean that's the whole purpose of HCI right the belief that humans interactions with technology will change Vice Versa let me take a second to remove my email tab so that it's not making sounds in the background oh yeah

I did not hear any of it so kudos to the echo constellation noise cancellation at Riverside might be having uh I'm sure it's AI powered in some sense as well which you know I mean things like everything is called AI these days but the reality is that it's just you know simple pattern matching which has always existed for a very very long time so it's a buzzword yeah uh I

mean that's the other thing right like AI is kind of making guesses of what to add and what to remove um and I think it's a helpful skill for humans to know when AI might do that right like so for audio it might we might assume okay it's going to remove some background noise and that's an understanding but when technology moves so fast sometimes it's removing or adding things that humans aren't generally cognizant of and I think

that's where it gets scary absolutely I was reading this article on AI tools that historically retouch photos from like 100 years ago 200 years ago and so on and a lot of the criticism is that no one really exactly knows what the exact colors were used and things like that and what the AI is ultimately creating is almost recreating what history might have been because it gives you eight variations you have to pick and choose is this the right coming like well I mean I want to choose what makes the most sense to me looking at the context of 2023 this seems like why we might appropriately be it and of course we have some general idea of like you know brick must probably be in the color

of a brick and the edges must probably be like you know brick cement break and all these things but it then it loses the authenticity in some cases as well yeah exactly I mean I think it's interesting to think about when is it okay for something to look broken right and for something to visually admit that it's not there where there isn't confidence in what to display

um I think right now we're operating in okay can we productize all these images can we like make all of these Technologies practical and so it'll we err on just guessing the color of the image um filling it in but maybe it's okay for there to be gaps and to kind of visually show how the system doesn't know something yeah which is always funny because when

you study history oftentimes they talk about like how our perceptions of History are often due to the lens of our present way of interpreting various events and things like historiography or something like that I could be wrong about the word but you know that you know when someone looks at a situation in the 1800s of like what happened in let's say the 1600 their interpretation of it will be very different than us looking at it right now and so some of the common feedback you get is you know these AI photos that are coming in it's natural because our interpretation will always be changing and this is the it might actually be accurate because for our current moment in time this is our history and our understanding of the

world has always been it's always a beautiful lens of what we just learned and how our perception is so it's always funny seeing all these things

yeah yeah that's true um I don't know about you but I don't know how many people are actually aware of like what a mid Journey well what journey is and what a majority image looks like so there's also I think who sees the AI art or tools that you create is important because the average person may look at a warpy mid-journey image and have a very different takeaway from someone like myself or yourself who's a bit familiar with what's going on

yeah I think we're in The Cutting Edge here you know we're basically right at almost the ground floor trying out all of these various things and learning what's going to happen I think a lot of these services will ultimately be made more readily available in more common things that people use on a day-to-day basis you know like we're seeing how like both the Google search engine and the Bing search engine are going to be adding AI prompts and generating visuals naturally as a part of their data activity that's going to be a lot more people because you know most people like you mentioned you right they probably don't know about mid-journey or chat GPT or stable diffusion or the difference in all these things are but they're

probably to go to Google and type in a question about like find me the best pizza in this location and they may end up benefiting from the results of all the things we're doing right now being returned for them and it could even be a death question of like you know how do I get from point A to point B there may be sequences of the journey where it's generated by AI because actual photos don't exist but we have enough context that's available and you'll never know you know what is AI and what isn't yeah or maybe we'll know what's AI what isn't um if HCI doesn't stop well but yeah it is exciting for all this to become more just commonly used and because it's

so Cutting Edge we're trying to like reimagine what these interfaces look like it's almost like CGI in movies you know it was one of those things where back long time ago the techniques for doing CGI were very primitive you know I think that's like like Wizard of Oz or something where it talked about how like the I forget what are the characters it's you know where the the show it was like you know blue and like the face morphed it was actually because they used a different lens on the camera that optimized for blue in one scene and red in the other so it made it look like the face is morphing and that was the technology back then whereas today of course it's all done on the computer you

know you watch any modern film it's often hard to tell what is real and what is green screen and what is actually done you know by clever visual tricks

and I do worry sometimes that we make games at world where it looks real sounds real and it may also be just you know to us our perception is like yeah it is real

yeah yeah it could just be the accepted reality but no the thing that always excites me about this this kind of the new technology comes in there's a generation of people who are like you know a little worried they always say like oh no this is going to take what I'm doing and the term is like creative destruction the idea is that as technology you know moves forward as things happen people will find more higher level Pursuits to follow as opposed to doing what they've done before we saw that with like the rise of Photoshop on illustrators you know someone I spoke recently used to design Billboards for a living in the 80s and once the digital you know tools became really popular he transitioned to

learning digital tools and became very successful doing like film intros and so on so you know I do think there's an exception I think a lot of people end up you know doing something very different and I'd be curious to see like how all of this changes essentially what you know you will be doing you know in the future in your career and how what I will be doing as I'm currently doing right now how that was evolved as well for sure we'll have a check-in in a couple years from now exactly exactly so I'm looking at some of the other parts you know Parts in the in the fake my document anywhere else I should go to continue this conversation

um oh there are a bunch of parts um zoom out of it foreign cool um oh okay yeah we can talk about um like what it means for these tools to become just like common common knowledge commonly accessed um in the case of art so up there actually yeah where it says the hand versus the machine or a machine versus hand yeah let me zoom in here um I think it's exciting for people to see that there are a lot of people working with technology and art and making really cool stuff none of the so what you're seeing here those are not mine that's from the generative Art Club it's a community on slack um of I think around 500 people right now uh using technology to create

beautiful pieces um and I've started working those mediums too which if you scroll down you'll also see um and I think what comes through to me is that people's Styles and their sensibilities do translate in the machine I think oftentimes people wonder if you're using AI tools if you're using technology does that mean that the artist is kind of just gone and it's the machine doing all the work like we should get give the machine all the credit and I think when you see all these different styles of outputs you realize that it's still us as individuals that are helping to guide and determine what looks good or what we want to have as an output

exactly like something like this creating it manually would take probably 20 times the amount of time it might have taken then going to a machine and like prompting it and making edits even to what has already been created exactly right and so that's I guess what people call affordances in HCI like what does the technology afford you and uh as you rightly describe some of these things would take forever by hand and so when it becomes so much faster to produce uh with a machine then you can get into more complex shapes you can get into more complex textures maybe you have more detail or more um variations in color that you might not be able to achieve by hand you know very cool one of the things

that I remember always you know sticking it from when I took some of these classes back in the day is that they always say like you want to solve the problem the domain you're currently in and so if you're selling a visual problem you want to solve it visually and if you're solving a logic problem you want to represent that in a logical fashion somewhere along the way we kind of like you know lost track of that and so I go back to like coding there's no reason why someone should spend so much time writing between HTML CSS and JavaScript three very different languages but the output is something very visual if I want to Center something on the screen why am I not doing it in a more you know direct

manipulation style way why am I actually writing bizarre lines in English or English like text putting some numbers in when the output is Visual and that's where I think the AI tools have really like brought that all whole thing full circle because we always had low code tools and no code tools but the challenge with them is that you always just Bound by what their the definition of done is or their level of support for things it's like a box that you really if you were break that box and go in the tools no longer work you know you make one minor change manually to the output the round tripping where what what's the final output and what the low code tool generated falls apart versus AI that is

completely you know removed which even is very primitive state I'm very impressed with the output on these projects and I can only imagine what happens in the future when we decide that you know programming will is not going to be the traditional world of I'm writing some things in a text editor waiting for accomplishing build step instead just like mid-journey I'm going to describe what the upload will be you know the prompt engineering work will be more complicated I guess because then it defined more of the interactions the the states that need to be defined you know hopefully it won't be done in text hopefully be done in more maybe a slides form it maybe figma it could even be just you know framework visuals and like

various scenes and it's upload that it says okay I can see the transitions I'm going to code it up it doesn't matter anymore whether it's react or Tailwind or whatever framework it's just going to be here's the output and that's what's going to be the thing we focus on

yeah I I think AI definitely provides a lot of adaptability um and I'm excited for us to be able to say hey I want this to be on the web this is what it looks like now go and you're not constrained by webflow not having all the Integrations framework charging you an extra 300 for for something else yeah um it'd be cool to have AIA agents that can just execute for you

exactly because you know I could always think you know make whatever talk to people about who work in at least in Myspace which is like you know design tools developer tools and Frameworks is that nobody ever wakes up saying I can't wait to spend eight hours a day writing code or I want to spend all this time like you know clicking around doing things some people do and I think you know they will always have something that they'll always be able to enjoy doing just like people enjoy creating art in more traditional forms and so I want to be very successful at that but if you look at the average you know what is the most people looking for it's like I have a portfolio I need to get up and

running I need to sell a product like in your case and it sells something that I've created online you probably don't want to fiddle with like what is the most what's a secure way to process credit card Transaction what is the regulations on what these policies are around the world you just want to be taken care of you know if you're using strike behind the scenes or any other service you're like yeah it's certain level I might care but in the day I don't really care as long as it's a good user experience and I have a very friction-free way of being able to conduct what I'm trying to do and I think that's the other part I think AI might make them an interesting space

definitely yeah I think what also comes to mind is it becomes more on us to determine where we want to learn right like what areas do we want to spend time on when we're no longer required to spend the time learning something to get um say a product out the door uh we get a lot of freedom to choose okay well am I gonna spend 10 hours of my day today just learning how to do it because that's something I'm just genuinely curious about um so I'm I'm interested in how people use their time in that world where AI can theoretically do anything for you

exactly and this always has arguments made around especially in the educational community on how important is it to learn the fundamentals of like something before you jump into tools that automatically do it for you I was watching a documentary on I believe it's a very famous sushi maker and he was talking about how he spent the first few years learning how to make the the kind of the egg version of this for like you know day in and day out like they was never allowed to do anything beyond that you just got that perfect before he was allowed to move on from his instructor on more other things and I obviously always struck me as an interesting concept because you know he explained like the fundamental techniques that he

perfected in making just that kind of an item ended up translating into other areas where if you jump directly into it it would have been very difficult and I always wonder how much is of what we're learning and you know I spent a lot of time teaching developers how to do interesting things that is enjoy learning things and sharing it I wonder how much of what you know I'm doing and what people are doing today is even going to be relevant like does it even matter at this point that you know what is HTML or like basic tags or does it even matter either you know about how to do brushes and how to do like compositions in Photoshop or illustrator you know I always do question those

things ah yeah I I guess we'll see but I feel like that the knowing at the very least the concept of what those things are can always be helpful but it depends on the use case right like you're saying if

they really don't plan on becoming a developer and they just want uh to sell their jewelry or um accomplish something it may not be relevant for them to know the details of but um yeah it depends on their goals

yeah it's a you know I'm always reminded of an interesting story when I when I first joined you know as a product manager on a team that was building a design tool you know my first manager was you know well known in the industry for having built design tools and we always have endless debates on what does it mean for something to be a professional design tool you know my view was that if someone is building something they should be focusing and telling the visual output it doesn't matter what the underlying representation of it is going to be his case was of course the counter he said you know for being a professional design tool we don't expect the designer or the visual you know implementer here to know

the intricacies of the markup or what things are going to be generated behind the scenes but they should be able to know the trade-offs between choosing this particular option for how you lay things out versus the other option for laying things up because that might have implications on performance it may have implications on how does the scaling work when you're on a larger screen versus a smaller screen and those were always some interesting conversations where there never was a clear answer it depends on what the what you're trying to do but it also required someone who was purely visual to go one level deeper into technical Concepts and to be able to understand that and then I often think back about it right now where if

we just describe it needs to be responsive it needs to be performant will the AI system automatically take care of it and does it now mean that I never need to know some of these things a curse of your take on something like that

I have difficulty imagining us not having to intervene at all where we really can trust it producing a good output it's performant and we move on with our lives like at least in the visual realm I still have this sense of like okay I want more control of what the output looks like or what the image looks like um and I think that's where it's helpful to understand the tools that you're using at some level not like in hyper detail but um I think that helps in getting impressive visual images when you don't feel constrained by things that you don't know about um and things just like working for some reason and you don't know why um I think a good example of this is

actually generative art right like a lot of generative artists spend more time curating than they do actually making the artwork itself and maybe that's the world that we go into with AI where if you don't aren't interested in spending time with constraining the tools or manipulating the tools to create the output that you want then you end up spending more time curating and just running that system over and over and again until you get hopefully the output that you want but that just seems in my mind that just seems more wasteful in terms of time and a sense of like artistic pleasure right like part of art is like your intention being there at least for me um and so I find it valuable to

um be able to have my hands in creating the output

exactly it's often in case of like you know it comes to cooking as well and you know like I enjoy cooking every now and then and so for me it's like you know do I go with something that's like more you know already made like where's where's the line there like do I actually pick the tomatoes fresh from like a grocery store do I go to an actual Farm or a garden and pick it or do I go from a can of tomatoes and then start from there in many ways it's like what is your constraint right now what you're trying to optimize for and if you genuinely enjoy just you know spending time cooking it's their way of like relaxing then I'm like yeah I'm willing to spend

more time like you know not jumping to the intermediate step but going more to the base like even if the output for the most part will be indistinguishable from like you know me having done it whereas it's me buying the can that already had the tomatoes seasoned yeah but I think knowing like okay I could cook this myself or I can just get it pre-done I

think that having that confidence is really valuable or like when I'm making things I find it helpful to know okay if I don't like this output I know I could just do it from scratch um I think this is something that I've noticed a lot of Stanford students being good at knowing when to take shortcuts maybe that will be a skill that really translates in the AI setting I mean that's a skill I feel like it's a beneficial life in general it's like I went to like you know dive in and go into the details which is when to just either say you know what this is out of my league I'm going to delegate to the service this person or this product you know and I think it's a it's a good one

to have yeah

so let me zoom out a bit in the figma and see you know like other areas that we want to dive into in the same time um we can go into AI or yeah or mid Journey here too if you want yeah let's go into majority oh these are fantastic let me zoom in

yeah I mean for people for people that haven't used my journey um I'm hoping that what we're seeing on the screen here can help conceptualize what it's doing um so if I just type in cat um it will well it will produce four images of a realistic cat because that is what my journey has the most data points for it's most familiar seeing a cat in a photorealistic form um and if I is the designer I'm actually wanting a cat logo um I can add that in to the prompts to then guide it into a better Direction because mid Journey doesn't know that when you say cat you mean a logo and not photorealistic image and so on the right you can see the progression of adding in more prompts

what that will get you and then if you go down you'll see what re-rolling does re-rolling is generating new versions of the same prompts and sometimes if you're lucky it will actually create the cat logo that you had in mind um you can see on the far left it went for a watercolor style instead of a photo realistic just in case you were re-rolling for that that look

no that is actually really cool because in so many creative you know sectors right you're always you know you get into a rut you know often based on familiarity or what you've done it's often like you know at least what I do is like I have a collection of like design magazines I've been collecting over the years I'm like stuck on them you know I need inspiration I just go there flip through it and I'm like oh that's interesting that gets me out of my rut gets me thinking differently and then I apply it to whatever I'm doing right now re-roll seems like the much simplified version of it where 80 of it might not be exactly what I'm looking for but just seeing that might trigger

something in my brain that causes me to want to do something different yeah exactly I mean that's where mid-journey is really helpful for artists even today in its most primitive form um when you type in just a basic idea that you have and you yourself have a conception of what you want it will give you different outputs and sometimes you're like well actually that's not bad I don't mind that um and that those are really fun moments uh with AI tools you know there's a gray area that application take on is a lot of these visuals are based on other visuals often ones that people might have created yet I don't see any way of knowing which of the 80 people who could have had photos

at some point or represented here so yeah what do you think is the right solution

uh what do I that's a heavy I mean I think that is the question right now for artists in the space there needs to be I mean you start to see solutions for uh text-based AI for example um I think it was Bing I know Neva also did this uh where you cite the sources in the text so whenever it gives output it'll have a little footnote or attribution to say hey this was from um XYZ website um and that's better than nothing right and right now in mid-journey for visuals you have nothing so it'd be great to just even have uh something like that where it might say okay with high probability these are the artists these are the top three artists that influence

this image um because nothing's for certain with AI right like it's never like it was a hundred percent this person or rarely um typically it's like percentages so if we could just have some sense of the waiting of influences that go into a visual I think that could be a step forward yeah I think that's a great great suggestion you know I always go back to

like when you look at code for example there's a very good chance that someone wrote exactly what you did in some other form do we get them credit at this point or not like if I'm doing a simple Loop for example and it comes out by the AI system like you know at what point do we decide if this is common general purpose thing that everybody knows which is this is very unique implementation that they probably need to get credit for I think that also goes into the unknown territory so we're all trying to explore yeah I mean that's a great way to think about it like a threshold almost right like I think we do that mentally with code uh like you're describing if you

feel like a lot of your code is starting to look like something you read a couple days ago you kind of make the judgment to credit it maybe AI could do the same right if it's below the threshold of some arbitrary 50 um then it it assumes that hey I'm the one I'm the AI system that's mixing and mashing everything I should get credit for being able to pull all these resources together in small fragments and then if it's above that then you have to put down who those artists are yeah I'm looking at these logos have been generated here these look really really nice almost as if like a real human being created them because I'm thinking about like back in the day when

I had to create a logo like for example the one with the the mountain and the various colors behind it I'm like that would require me to go to a professional logo designer you know go through multiple iterations whereas here I'm like oh I see four variations of what I'm trying to do they all look fantastic and I'm sure if I just go back to Mid Journey like you know do more prompts I'll get the output I want very very quickly yeah okay so this is where I

fish like to me these are all bad um like I agree that it's good for AI um and maybe if we go back in time five years ago 10 years ago then this is amazing but like I do not think that this so to me this is equivalent of you going on Fiverr and hiring someone for five bucks to make a logo like in which case this is great this can replace it for cheap but if you're looking for something slightly more developed um I think there's a long way for AI to go no that is very fair and I

think it goes just to like when the web became more democratized but anyone could create a web page and start uploading to whatever host they could there was a gradient you know higher quality content you know had a very distinct look and feel lower quality design and content you know kind of felt that way yeah I think we're going to see the same proliferation of like a lot of content that goes through here and then we're going to see like the 110 truly good to get like you know the I think the power law applies in every sector it's gonna apply here as well a few you know percentage of individuals or organizations who created things that we were like wow this is Iconic this is

what we want to aim for and then everyone else will be able to but what does mean though is that more people who may not have the skills to create or design logos will be able to do something good enough I always say man I always go back to like you know no fault of their own when I was in school the clip part on my school assignments was very noticeable for how bad it was I can absolutely imagine my daughter's now you know about five years old I can imagine that as she goes through her education teachers who may not have like the great artistic skills will probably have higher quality clipart at least in my perception than what was available because the the building would represent

that in a more modern way it will be easier than whatever it came with whatever tool you were using back then absolutely yeah I think my for sure I

mean this having access to these kinds of tools is super powerful and definitely raises the bar for everyone um I now that I've seen the bar race at this level I'm just excited for the bar to be raised even higher right so I'm ready for AI to become something that professional designers can feel comfortable showing directly to a client

exactly and I think the you know the amount of skills he might have needed historically to be able to pull that off might be different than what you might need in the future where you know I mean I need to go to an advertising agency and spend many years to be able to do this but it's going to be more of a you know what creates the right output and I I do sometimes worried that there's always a premium that people charge for higher quality experiences and so many Industries are built around that premium being charged if we make it easy for 80 percent premium to be created very quickly by anybody will these industries survive on just a 20 being extra like how many clients will actually care that

you know I could pay X thousand dollars for this logo or I can pay ten dollars for this one and my target audience probably wouldn't know the difference between a logo design for this much and and for this much and that's an area where the speed at which so many Industries are being you know reconsidering like how are they going to operate in the future I think that I haven't seen with other things in the past Maybe electricity but I was alive for that when electricity first came out yeah no that's fair um yeah we can go to the other section if you want yeah it doesn't have a personalized UI because that to me is the one that is always very interesting and I always spend a lot of time

thinking about this because you know when dark mother Lifeboat came around you're like oh wow that's like you know a new level of customization but then we also know that different people have different preferences for how they interact with content you know how they navigate and so on today it's very much fixed you know depending on let's say you're visiting a typical application that application has a style that's been defined by their style guide that's been defined or probably years and years of refinements from like usability studies and things like that and we always know we're compromising at the end of the day you know we're designing for a happy medium like a middle you know group of people they're going to be people on

both extremes you know of varying sizes who will not find the UI or find the interface usable or you know approachable a nice way where to find AI to be really exciting is that just like we are actually creating prompts to be able to generate things you know that come with the output we can maybe even dynamically generate content as part of a person's interactions that you know let's say that I'm zoomed in at like you know 200 or 300 like I'm in figma right actually I mean 50 right now but it wasn't like 20 earlier maybe all of my UI will adapt you know this is almost like the intersection of like the work that's often done in like accessibility in these areas but actually applying it

more broadly to usability where even people who might prefer a certain style of how to navigate their application will now be able to have the represented in their app because it's not predefined it's all fluid that the AI kind of a system determines what the appropriate positioning and layout and sizing of things are because it you know before it had to be like you predefined X number of templates you pick and choose maybe which one is more appropriate here in a world where you can generate templates on the Fly there are no constraints of that anymore

yeah I I know a number of startups that are exploring the space um and actually the use case you described for ads is a very lucrative one being able to show images and text on the fly to someone who has different preferences um we haven't been able to generate like photos on the fly like that before um and it would be really cool maybe creepy but you know really cool to do that if I

you've got a Shake Shack Burger yesterday and I'm searching up like Burger shops today um it'd be great for not just the Shake Shack ad to come up but but for it to be with like a California background right like maybe with palm trees in the back because it knows I'm in California and that's something visual not just Tech space that they can play around with exactly like if it knows for example that you tend to prefer specific color scheme you know like or you specifically like a specific theme like let's say you're into music for example you know what if like all the web pages you visit happen to have like whenever there's like a header image where it's often like a generic image that it's like

specified of like you know like whatever like all these spirits and brands have they were always themed to be music themed like you go to Instagram and the default Banner in the background is of like you know what you might like could be musical instruments could be a picture of an orchestra you know all these combinations imagine an entire experience of interacting with all of the interfaces whether it is on your screen or on your like thermostat or all these various areas is very hyper personalized to exactly what you like because it's all generated entirely on your preference yeah

um I think in some cases it's really cool to have that functionality and adaptability in others I do Wonder like where does branding come in right because one of the benefits of consistency is that people recognize or have a unified experience of what a company does and what their branding looks and feels like and so to me that's kind of the tension do you want hyper personalization to give people like a really personalized touch or do you as the person providing the platform or the service care more about how you're presented to the user if they're spot on you know there's always a tension there between you know

the first example I remember distinctly was there used to be a product called Winamp which was like an MP3 player that you could use like locally in a machine to play music and it has idea of skins to make actually personalize it on all these various formats and I'm like wow there's like a very you know it's very big into like you know geometric kind of art like you know Mondrian kind of style and so I had like a Winamp skin for exactly that's like you know basically MP3 player was very much blocky in that sword and then first like Myspace came out where the idea was when people are visiting your profile it's really to see you and not the brand of Myspace itself

and so people personalize it all kinds of you know cool often horrific ways as well like black text green blinking you know backgrounds and Marquis you know all the stereotypes have like really bad design was what was done but that was okay because that's what the person wanted to Showcase about themselves whereas now you see like you know Facebook and so on there's a very limited level of personalization the content that you post can be different friend but that interface is often displayed around is very varied but I wonder how much the technology that limits that kind of customization because can you imagine unit testing like one million variations and like making sure that automation Works in accessibility works like all the various contrast guidelines you need to operate

around the world all these things work no longer matter because the system will automatically generate it I wonder if that might unlock some interesting layers of creativity

yeah yeah no I think you'll have just like how the internet provides so provided so much more data and so much more information I think the same will happen with AI tools like you just get overwhelmed by the number of potential like the number of possibilities um and I think in our example for making ads more personalized or platforms more personalized um it really is about curation and monitoring maybe that becomes what's important if AI can generate a bunch of different options maybe the human comes in and has to approve or adjust what looks good versus doesn't rather than creating the actual artifacts absolutely

and I think some Industries will be more readily willing to accept and investigate some of these things and others might be more tricky do I'm always imagine like you know we think of like that typical acronyms like a b to B and B to C and consumer product and so on I think consumer products will absolutely benefit from like getting this personalization whereas I can often imagine like more traditional like B2B kind of products where you might have someone going like I'm logging to my bank I can't find the submit button anymore what's going on you hope that it takes a 40 on the other side can replicate your version of the app in a way that is repeatable because for a lot of these organizations every time that

you're stuck doing something is a huge loss I can't imagine like Bank of America for example giving each person a very Dynamic UI knowing that there's no way to reproduce it you know and mistakes in areas could be hugely you know like you know bad where they might wire the money to the wrong person or you know the the UI for wiring is like going out it's supposed to end and like little things like that where consistency might be the right thing and that's going to be a tricky area to navigate as well

oh yeah I mean I can't imagine like the customer support people they'll be pulling their hair trying to work around uh these kinds of tools

well Chloe it's been fantastic talking to you about this like this you know to those people actually listening and not watching it and couldn't watch it because this is one of those like interactive kind of things where there's a great figma file that Chloe created with all the various things to navigate through and kind of add visuals to the conversation that we're also having and so before you wrap up you're currently you know studying HCI what is your long-term goal you know what is you know you got to think about what would ideal you know life and career be like you know 10 years from now if you work in this area what do you think it's going to look like

oh um I mean I love startups I think if I can work on a startup that's in this intersection thinking about design and visual questions with technology that's the sweet spot for me

perfect Chloe thank you so much I'll talk to you in a bit