PRESENTING: Pioneers of AI with Dr. Rana El Kaliouby and Musician Taryn Southern

Hey, possible listeners, it's Reid. As we settle into this holiday season, we want to share something special with you all. An episode from the podcast Pioneers of AI. The show is hosted by AI scientist, startup founder, and investor, Dr. Rana El-Kaliyubi. In this bonus episode, Rana talks with Taryn Southern, an artist and technologist who's exploring how AI can enhance creativity without losing its human touch. Taryn shares her journey of co-creating the first AI-assisted music album and how she's used AI tools in unexpected ways. And fun trivia, Taryn very early on developed a digital twin with Hour One, just like I did to make Reid AI. For those who have listened to our recent three-part mini-series, Artful Intelligence, this episode is an excellent coda. So I encourage you to give a listen. And if you want to listen to more episodes like this, check out Pioneers of AI, wherever you get your podcasts, now onto the episode. Pioneers of AI is made possible with support from Inflection AI. It's not just Enterprise AI, it's your Enterprise AI. I don't have a need to feel unique or special, or that I've worked super hard to develop this one skill set. I think it's super cool that a robot can do it better than me. Taryn Southern is an award-winning storyteller, artist and creative technologist whose work is focused on the intersection of technology and human potential. A three-time Streamy Award nominee, she has co-directed and produced an award-winning documentary on the future of the human brain. It currently serves as the chief storytelling officer at BlackRock Neurotech. And as you can tell, she's at the forefront of all things AI. And it makes me feel like if I use this robot tool, then it'll just elevate my work. And it'll allow me to take on all these new skill sets that I didn't have and fill in the holes. Taryn sees AI not just as a tool, but as a collaborator. From co-writing music with AI to making a digital twin that helps her create content online, she's built a career by leveraging cutting-edge technology to create things she couldn't do alone. Like making history by composing the world's first AI solo pop album. In this episode, we talk about how to best harness AI across creative endeavors. Now, to be clear, there are plenty of valid concerns around AI as a shortcut that could put people out of work in creative industries. But rather than avoid or fight the technology altogether, Taryn seeks to help fellow artists and everyday people embrace the potential of AI. I'm Rana Alkal-Yubi, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution. Hi, Taryn. Great to see you. It's always good to see you. So excited to have you on the show. Thank you for having me. Yeah, so you and I have known each other for five years or so? At least. I'm a huge fan. I think what I love the most about you is that you're such a multifaceted, multi-hyphenated person, you're a technologist and an artist and a creator and a storyteller and a musician. It's super cool. So I want to talk about what you're up to today. But first, I want to take us back to 2018, and that was the year you released your music album, I Am AI. And that was a collaboration with AI to create the album. You were very early. Like you were very early in using AI as a tool, right, to collaborate. I don't even think generative AI was a term being used at that time. No, it was not. So tell us, what inspired you to use AI to create your album? So I was in a Google artist and residence program at the time that was primarily focused on virtual reality. We were taking all of these consumer tools like tilt brush and blocks and trying to create these really immersive worlds at very low cost. And basically envisioned this whole futuristic world that has my AI and my human self and their merging, their meeting. And it was just going to be this artistic exploration of the future of human AI interaction. And I thought, OK, if I'm making this whole world, simulated world, how cool would it be if the music in the background was created by AI? And I happen to have read an article in The New York Times that same year about this one music company that was making AI music. And so I reached out to them, just cold turkey, and said, can I play with your tools? And from there, it just became this kind of exploration with a number of different AI music tools, most of which are sadly not even in existence today because they didn't make it through the winter of 2019. What was that experience like? It was so eye-opening because I have always been passionate about music. I'm an OG YouTuber, so I was making content throughout my 20s, a lot of which was music-based. But I never knew how to play an instrument or how to actually compose music. So I would write lyrics or melodies, but I had no way of actually bringing that to life. And so for me, this was the first time I was ever able to just escape into my room alone and create a song from scratch without those skill sets. And yet feeling as though I still was exercising a huge degree of creative control, creative discretion. The same thing that so many of us are now experiencing as a result of LLMs and generative AI and we're saying, "Oh my gosh, like our creativity has just exploded." I got that little nugget of insight back in 2017. And it was so exciting because I was like, "This is it. This is the beginning of a whole new creative landscape." But I just didn't have a language for it at that time. So much has happened in the AI landscape, obviously, since you released the album. If you were doing the album all over again now, what would the experience be? And do you think the results would be different? 100%. For one, the tools are really different, both for better and for worse. So some of the tools that I was playing with back then, I was actually having to ingest my own data. What was the data? It was like these MIDI tracks, but you'd have to figure out what kind of song are you wanting to create. And so therefore, you're actually creating a subset of data that will get you that kind of result. MIDI tracks are data files that contain information about musical notes, timing, and other data needed to make a song. Essentially, it contains the instructions for making a song without the actual sound itself. So one of the tracks that I made was based on thousands of period pieces from the Romantic period, so 1800s. And so to have to create a pop song from that was a really interesting process. Now of course, all these tools, you're not ingesting your own data. You don't have a peak behind the curtain of how it's generating. One of the other tools I was using was a rule-based AI tool. So that's another area where I can't really get a peak behind the curtain. But the tools are very different now. I think we have a lot more breadth, and the quality of outputs now is much higher than what it was. So what probably took me 250, 300 iterations in 2018 would maybe take four or five iterations now, unless you want to get really into the nitty gritty. But if I were to make an album now, I would probably deviate into a lot more strange little corners of music style, just for the creative exercise. Do you actually still use AI to write music? I do. I use Suno and Udio, and I find myself just blown away. Those tools were not around when I made my album. I like to think of technology as a tool and AI as a tool as well. And like the car, AI can take us to a whole bunch of new places. But AI is more than just a tool. It's more like a creative thought partner or a collaborator. So can you give us a sense of what that collaboration looks like, and how do you build a successful partnership if you like? Sure. Whenever I work on any kind of creative project with AI, and now I've done dozens of different creative projects, I really look at my role as no different than a filmmaker who's directing a project. You have a vision. You sort of know what the end goal is, and then your job is to reverse engineer that vision into tiny little pieces. So you've got your set direction, your costume design, your acting, the composition and the musical background, the scene direction, the cinematography, and you've got to be thinking about all these little puzzle pieces and how they fit together, and be able to impart your vision in the language that each of those areas, each of those divisions of creative execution can understand. And so I look at AI the same way. I almost divide it up into my little team members, okay, you're my cinematographer, you're my writer, and I'm going to speak to each of those team members a little bit differently. I'm going to provide very specific instruction, sets of guidelines, parameters, constraints, let them know what the resources are, and then I let them get to town. Can you give us an example? Of course. So last year, I actually needed to do a demo for a company, and I wanted to do something ridiculous and silly. So do you remember Fire Festival? That festival that just really caught the media by storm and has two documentaries out about it? So I imagined that I was making a new video trailer for Fire Festival 2 because they had just announced another Fire Festival. Fire Festival was a music festival that took place back in 2017. It was promoted as a luxury event on a private island with food, art and music, but turned out to be the exact opposite. Instead, guests were given cheese and lettuce sandwiches for food, lodging in FEMA tents, and guests' luggage was dumped in an empty parking lot. Now, Fire Festival is planning a comeback in 2025. I go to chatGPT and I actually make a few different versions. One is my cinematographer, one is my comedy writer, and one is my marketer. So I'm using them for different roles in the creation of this video trailer. And with the writer, first I have to provide my objective. We're creating a two-minute comedic trailer that is designed to sell tickets to Fire Festival 2. These incorporate information, facts from the previous Fire Festival to inspire nostalgia, enthusiasm. Here's your target audience, 18 to 35-year-old influencers who don't care if it's a fire storm. They just want the association. So I give it all of these detailed instructions. And it writes this script for me. It's iterative. There's a lot of back and forth. And it was actually, I mean, it was a pretty funny script the first time around. It was pulling in specific quotes from the previous attendees as testimonials. It brought in the infamous grilled cheese sandwich that everyone knew was the food of choice for all of these attendees at this ridiculous festival. But there was back and forth because I have my own comedic sensibilities and my own constraints in making this video. So I'd say after about 10 back and forth, I had the perfect script. This took me probably 40 minutes in something that would have taken a couple of days. Now I take my script and I feed it to my chat GPT cinematographer and I say, we need to make something that is beautiful, cinematic, photorealistic, provide a shot list for me based on this script. And I want the close-ups, the mediums, the wide shots. You need to tell me which camera we are using and the lens. Then I take those very specific shots and I pull that into runway or mid-journey, any of those image or video generators to create the individual shots. And that's just two of the roles, but of course you can expand it into music and costumes and all of these other areas. So you can get as detailed and as creative as you want. What I love about this is I'm not a creative, I don't have a creative background. I'm not an artist. With AI, there's a vehicle to bring that creativity to the world, right? In a way that was not possible before a lot of these tools existed. 100%. I mean, there is no such thing as a creative limitation anymore. I mean, one of the reasons that I was so drawn to being a content creator early in the days of YouTube was because it was a bunch of people who were just taking consumer tools, breaking them, figuring out how to make content as cheaply and as efficiently as possible because that was the only way you could actually survive on YouTube. But it meant not having to play by the traditional Hollywood rules of requiring millions of dollars and dozens of people to get something done. Because that process, while it fuels beautiful art and movies, it also has prohibited so many incredible ideas from ever even making it to the screen. And so now, I mean, to see the cost of content is virtually zero minus the human capital required. It means like anyone with an incredible idea and a vision can go and make that and get it done. Yeah, and I know you're very passionate about this idea of democratizing access in the creator community. 100%, so many people just don't have the resources and that can be such a detractor from even pursuing a creative field. AI is allowing us to create projects on our own that used to take teams of people or a wide range of skills, but not everyone is on board. Concerns around IP and copyright issues have many artists up in arms. As an artist herself, Taryn has a perspective on this that may surprise you. Let's talk about some of the concerns. Some people in the creative industry are worried about AI. For example, there are thousands of people who have signed on to a statement opposing the use of their creative work to train AI models. Some big names include members of the band Radiohead, actors Kevin Bacon, Rosario Dawson. I personally think these are legit concerns around copyright issues and IP. How do you think about that? Like as you create this content, how do you think about, you know, it's probably being used to train some of these models? Totally. I'm of two schools of thought, so I'm going to be a both ant. The first is I do believe that these models should have elements of consent and control embedded into them for artists and copyright holders. Does this artist, does this copyright holder consent to the using of their data? I think that should be baked in. Is there an ability to actually be compensated? So consent, compensation and control, actually. Can they be compensated even if it's fractional compensation, just like Spotify is fractional? And then the third is control. If the artist is consenting to the use of their material, can they have some sort of control over how that's actually embedded and reused? And I think once they've figured out those three components, you're going to see any and all artists signing up because that's how you recreate the business model. I mean, YouTube did the same thing. There was a long period where you couldn't upload a cover song of a popular band or artist. If you did, it would immediately be taken down by the label because the label would say, you know, this person is, is not the copyright holder and we're not making money off of this. And YouTube figured out and other startup companies developed around us to create an industry where YouTubers, content creators could easily split revenues and rights. And it was this very easy process. And what was so funny was that a lot of these cover songs by these YouTubers were getting bigger and bigger and bigger and bigger audiences. Sometimes they were getting bigger audiences than the original artist. Let's take a Sharon's hit song, "Shape of You" as an example. The most viewed cover of his song is by Jay Fla Music, which has over 350 million views on YouTube. Another cover performed by "Walk Off the Earth" features weird instruments like kitchen utensils and a jack in the box, and that brought in 27 million views. The rock band "Hour Last Night" did a cover of the song that was shot solely on an iPhone and received 28 million views. So basically there's a whole ecosystem of artists that cover popular songs. And by doing so, they extend that song's reach. YouTube decided to lean into that. So they were making more money for the artist than the actual artist. So then of course the artists were like, "Hmm, we're collecting checks doing nothing. We're very happy with this." And so within 10 years there was this big push by the labels to try and get YouTubers to cover new releases. So we saw a complete switch just because we figured out the consent, the control, and the compensation. You know, I just invested in a company called Parada.ai and they are trying to bring to market this idea of attribution as a service. So if your data has been used to tray an AI model, whatever your data is, right, might be a publisher, a creator, et cetera, their algorithm can detect what percent of the prompt's answer is due to your data piece and then you get a cut of the revenue. So it's a new business model and I'm curious if it'll scale, but if it does, I think it solves a lot of these issues. But I also think the intent is in the right place. You know, the problem, and I'm sure you already know this, is we're talking about thousands, millions of pieces of data that are being ingested into this. So when we're talking about fractional compensation, you know, artists complain about the pennies on the dollar they get for Spotify streams. I mean, this is going to be way worse. But I think people care more about do they feel that they're actually being included in the conversation than what are they actually netting out at the end of the day. I think there's a philosophical argument here that is a tougher one to sell to the masses. And I don't necessarily think we should be using this as an argument for tech companies continuing to use all this data without permission, even just us as human creatives. We're collecting data and information around us all the time, right? Our brains do not store attribution models for all of the sources of which we've, you know, learned our various trades and skill sets. And so if we think about AI more analogous to the human brain, then we could just argue that it's just a much higher capacity. It's like a savant who's just pulling from all of these sources, but it's operating much the same way a human brain does. And I don't know, it's not like we force artists to give royalties to their mentors or the people that have inspired them to create the work that they have. Fascinating alternative way to think about this. Yeah. It's not a very popular one for artists, but I think it's real because this is actually, I mean, these systems are designed to mimic the human brain. How closely they actually do so, it's hard to say, but if we think about it in terms of that neuroscience perspective, it's an interesting thing to chew on. I also think there's a certain purity that some artists are holding on to. You know, they develop their skills over years and years of work and practice. And now you've got this AI tool that can kind of mimic that. And so this all can feel very inauthentic and a little threatening. Do you think that's a fair take? I completely understand where these artists are coming from. I can't imagine feeling like one minute, you know, you've spent your entire life working on honing a craft and then something else can generate it in under five seconds. It is threatening for me as an artist, as someone who's made my livelihood off of the film, music, television, digital media. For some reason, I just feel the call of adventure is so much greater. I don't feel that we need to be special in order to still gain tremendous love and passion for our creativity and our exercise of that creativity. I think a lot about this question of what does it mean to be human in the age of AI. I think I agree with you like AI doesn't have to threaten our sense of self, right? It doesn't. We could we could stay grounded in the creativity and the empathy and the authenticity. And we still get to decide how we apply these tools. I mean, for me, I think creativity is the opposite of just doing one single thing incredibly well. For me, creativity is an exploration, it's a journey, it's a synergy of all of these different elements and pulling them together. So by definition, using AI feels like the next step. AI may be the next step for our creativity, but can it also be the next step in our humanity? What happens when this technology actually becomes a part of our bodies and allows us to accomplish tasks just with a thought? So you've also explored a whole bunch of other creative areas. One of those is what I'd argue is the ultimate human-machine interface, so brain-computer interfaces. You created a documentary called I'm Human, which explored this idea of BCI. Many people will have heard about brain-computer interfaces through Elon Musk's Neuralink. But can you define what it is, what are some of the cool applications of it, and of course some of the ethical concerns? Sure. So BCIs have actually been around for quite a long time, several decades, they've actually been implanted in humans since 2008, but Elon's got a very large megaphone, so that's great. It's great for the whole Neurotech space because now everyone is aware, oh my gosh, there's people on this planet walking around with a tiny chip in their brain. But it's true. So BCI is essentially a tiny implantable device that goes inside the brain and it allows you to connect the human brain to a computer. So anything that you and I can do with a digital device, someone else can do with simply their thoughts. So there's approximately 43-plus humans currently implanted with the technology of the company that I work for, I think around the globe, around 55 total, 60 perhaps. The early trials around this technology are absolutely awe-inspiring. You know, you've got people who can operate robotic arms, who can control a cursor on a screen, can type, can create art, can feed themselves with a robotic arm, can operate a wheelchair, can sense touch through a robotic arm. I mean, all of these things are possible now through the use of BCIs. Is the first kind of application area of the space for people who perhaps have challenges around doing these things, motor challenges or neural or sensory challenges? Yes. It's mostly people with motor limitations, so people with paralysis and ALS. Those are the two primary indications, but of course there's other neurological disorders that are also being explored with BCIs, but this technology is so impactful for them. I was recently on a call with a patient who is fully locked in with ALS and uses his BCI to work every single day. He's able to show up because of things to COVID and all of this movement towards completely virtual work. He can work every single day, and he's just thinking out what he wants to say, and the BCI decodes those thoughts, decodes the motor intention, types it up, and there it is, and it's sent through. It's wild. I mean, that was not possible five, seven years ago. And do you think the idea is eventually this will become like a consumer grade technology where, you know, I'll be driving my car and instead of using voice, I'll just be thinking about, okay, these are the, you know, write the following seven emails, and then it sends it off? It's hard not to imagine that it will eventually go in that direction simply because every technology goes in that direction. The question is how long will it take and what will be the assimilation process along the way? It's one thing for people to become comfortable with physical enhancements of their body, which we already see a lot of that happening. It's another thing for people to get comfortable with the idea of altering the thing that represents who they are. And so I'm very curious to see how the public will reconcile with these issues and kind of how it will be accepted. Well, what are the ethical kind of, how can this go wrong? Lots of ways. Yes. We actually established an ethics advisory board at BlackRock Neurotech, which is the company I've been working at for the past three and a half years. And so we assembled ethicists, philosophers, legal experts, patients, neurosurgeons, kind of everyone in every category to sit around and talk about the issues. What's the top two? I mean, it depends. Are you talking about an immediate timescale or on the 30-year time horizon? Because the 30-year time horizon is what people are generally focused about in the news. They'll say, "I don't want companies reading my thoughts and then owning my brain data." That actually belies the actual, like what's actually possible right now scientifically. But in the next five years, there are other ethical concerns, such as, what happens if a startup who's raised a lot of money goes under and a patient has a device in their head that they can't have extracted or that they can't get updates on? I mean, those are actually the issues that matter now. But of course, the media sticks to more of the Black Mirror narratives, which may be relevant 20, 25 years from now. Where does the world of AI and BCI intersect? I mean, it intersects everywhere. We're using AI in a lot of different ways. You're using AI to decode the brain signals. I mean, you're taking in a lot of brain data, and then you're having to convert that into sort of a digital code that computers can understand. Sometimes there's other algorithms being used to fine-tune a model around language assessment. ALS patient I was speaking of actually has his own voice that's being used. And so that also has AI-enabled capabilities. So it sort of depends on the patient and what the application is, but you actually can't use these devices without AI. Working with AI can be intimidating for people unfamiliar with it. So what's the best way to get started? And what other AI tools has Taryn used to expand her creativity? Here's a hint. One of these tools looks just like her. That's after the break. So I love your posts on Instagram because you make AI accessible. You're always exploring these new technologies that are coming out and new use cases. If you were advising one of our listeners on who are maybe not so familiar with AI and they have a fear of getting started, what would your advice be? The most important thing you can do with any sort of emerging technology is just pick one. I always say chat GPT just because it's the easiest. And just start using it to help you solve random problems. So next time you have that thought of, oh my gosh, I can't figure this thing out or I need to research this thing, you know, ask chat GPT, ask the AI system and you'll start to learn all of the different ways that you can apply this in your life. So last week I was having some girlfriends over for dinner, I'm not a cook, terrible. And one of my girlfriends had come over and said, I really want you to teach me how to use AI. And I said, well, lesson number one, I'm just going to use it for the problem I have right in front of me tonight, which is that I have all of this random food, but I don't know how to reheat it properly so that it actually tastes good. So I just took a video, you know, uploaded it to chat GPT was like, tell me how to cook all of these different food items for someone who cannot cook and burns everything and try to put it, you know, make the instructions so that I know what time I'm putting it into the oven when so that nothing's cold when the dinner party is supposed to start. I was a little more articulate when I asked the prompt, but you get the idea. It was like total dinner party for dummies instructions for my exact food items. I was like, okay, you're going to roast the Brussels sprouts at this time. And then five minutes later, you're going to put the chicken in so it made it so easy for me. I didn't I took all the work out of it and my friend was like, I would have never even thought to do that. Yeah. I just learned something. I'm a terrible cook too, but I love hosting. I'm going to try it the next time. Yeah. It's so cool. So you also created your digital twin. I did. Yeah. What was that experience like? How do you put your digital twin to work? Well, it's so funny because now it's so easy to create digital twins, which is amazing. I made my digital twin back in 2020 with a technology called hour one and back then you had to ingest a lot of data. So I think I was sitting in front of a camera for quite a while to get the necessary amount of data to create the digital twin because now I can just do it with like a minute of video. Yeah. It's wild. It just is a fun experiment and to see is it possible for her to sound and look like me and released a couple of videos. Now I am playing around with actually making scalable social content with my digital twin. I am a creation of artificial intelligence. That's right. I don't exist in the 3D world. I exist in pixels 1280 by 720 to be specific with the permission from the real Terran Southern AI Terran. You know, it's weird though. I have such mixed feelings about this as a creator because on one hand I want to interface with people that I know are real on the other side. However, as a content creator, the part that I like the least is filming. I never liked filming. I don't like having to put makeup on and do my hair. You know, you get it as a woman. It's super annoying and time consuming and just like I want to spend my time doing other things. I'm not I'm not a talking head. So that part would be thrilling for me to just completely erase from my workflow. I think this is going to be one of those things where I already know a couple of content creators who haven't filmed a single video in three months that they have completely automated their process with their digital twins and it looks amazing. Now that I know I can see it, but most people I've already asked certain friends and they would never be able to tell. So that friend clearly doesn't disclose that it's an AI. Where do you fall in that line? Would you disclose or not disclose? I would disclose. But I think once we as a society figure out how we feel about these digital twins, they're going to be everywhere. And then at a certain point, the disclosure will become mute, I think. Beyond the creator economy, what are some areas of technology or in particular AI that you're passionate about? I am pretty thrilled about what's currently going on in the robotics world. I know a lot of people are terrified. I cannot wait for my robot home assistant. Are you kidding me? I would 100% if someone said to me right now, would you rather have a car or your robot home assistant? 100% my robot home assistant. I want a robot that will clean up after my messes, that will keep my closet organized at all times, that will ensure I'm taking my supplements every single day, that is making healthy food go, you know, I just bring stuff home from the farmer's market and it chops it up and it makes it into these beautiful, easy to eat smoothies. All of a sudden, I'm getting all the nutrition I need rather than doing what I'm doing now, which is going on Uber Eats and just getting whatever is fastest. I think this will make a massive impact in our well-being and I'm so here for it. That's for sure. All right, so I want to try something with you. I'd love to do some rapid-fire questions. Oh, no, I'm not good at these, but I'll do my best. Wearable tech, yay or nay? Yay. AI companion, yay or nay? Yay. Oh. Okay. I mean, like, romantic companion? Well, okay, AI friend and AI romantic companion. Yay and maybe yay. Okay. But, like, only with the human companion. Okay. Chat, GPT or Claude? Claude. Okay, should you have Chat, GPT write the first draft and then you edit or you write the first draft and then Chat, GPT edits? Both, depends on the use case. Very cool. This was so fun. Thank you so much for joining us today. Thank you for having me. I love how Taryn is always experimenting with emerging technology. If you're curious, I encourage you to follow her on Instagram, where she shares a lot of her learnings. Speaking with Taryn makes me excited for the bold future AI is helping us create. So, I want to know, have you used AI to bring a creative idea to life, like making a song, a piece of art, or even a poem? Was the experience surprising? How did the results make you feel about creativity and the future of technology? Tell us all about it. Leave us a voice message at 601-633-2424. That's 601-633-2424. Or connect with us on social media. We're at Pioneers of AI on all major platforms. And don't forget to rate and review this podcast. We're a new show, and this really helps other people find us. Thank you. Pioneers of AI is a Wait What Original. Our executive producer is Yves Tro. Our producer is Rachel Ishikawa, and our associate producer is Jordan Smart. Our senior talent executive is Stephanie Stern, Mixing and Mastering by Brian Pugh. Original music by Ryan Holiday. Production support from Timothy Lou Lee. And our head of podcasts is Litao Mulat. You can join the conversation on LinkedIn, Instagram, TikTok, YouTube, and X. Just search for @PioneersOfAI. Thanks so much for listening. [BLANK_AUDIO]