SC-090 · Expert
Music, AI, and the Human Touch
Guest: Pedro Sarmento, Postdoctoral Researcher at the Center for Digital Music
Summary
Pedro Sarmento is a postdoctoral researcher at the Center for Digital Music in London, where his PhD focused on guitar tablature and deep learning. A guitar player himself, he argues AI tools should assist music making rather than replace the human contribution to it, and doubts that the connection people feel to human-made music, including reasons unrelated to sound like an artist's story or the causes they support, will ever disappear.
Sarmento draws a line between music as a creative experience and music as a business. He does not think AI threatens musicians' sense of creativity, but expects real disruption on the commercial side, especially for functional uses like advertisement music, where a company might pick a tool over hiring a composer. He compares this to chess, where machines surpassed humans decades ago without stopping people from playing or enjoying the game.
His PhD research favored symbolic formats like guitar tablature over MIDI, because tablature also captures how a note is played, such as hammer-ons, palm muting or string bending, detail MIDI and sheet music miss, even though AI music research has historically favored piano over guitar. He sees one-on-one AI tutoring as a genuine democratizing use in music education, and argues that as AI narrows the gap in technical skill, storytelling and the reason a human performer is on stage will matter more than ever.
As of the episode's release on 16 January 2025.
Key takeaways
- 01Sarmento argues AI poses little threat to music as a creative experience, but will change the business side, especially functional uses like advertisement music where a company might pick a tool over hiring a composer.
- 02He compares AI in music to chess, where machines have far surpassed humans for decades without stopping people from playing chess or caring who plays it best.
- 03His PhD research used guitar tablature instead of MIDI because tablature also captures how a note is played, such as hammer-ons, palm muting or bending, detail MIDI and sheet music miss.
- 04Sarmento sees one-on-one AI tutoring as a genuine use for democratizing music education, since studies show one-on-one tutoring gets better results than being taught in a class.
- 05He argues that as AI narrows gaps in technical skill, the reason a human performer is on stage, their story and intent, becomes more important than ever.
- 06Sarmento expects that once mainstream AI tools cover the simple uses, more exploratory and experimental approaches to AI music will emerge, as new tools rather than faster versions of old ones.
Guest
- Pedro Sarmento, Postdoctoral Researcher at Center for Digital Music
Questions this episode answers
Does Pedro Sarmento think AI will replace musicians?
Sarmento says AI poses little threat to music as a creative experience, since the reasons people love a band, such as their sound, story or the causes they support, go beyond anything a machine can copy. He does think AI will change the commercial side of music, especially functional work like advertisement music, where cost matters more than artistic intent.
Why did Pedro Sarmento's PhD focus on guitar tablature instead of MIDI?
Sarmento explains guitar tablature captures not only which notes are played but exactly how, since a guitarist can play the same note several different ways. He says most AI music research has focused on piano and MIDI instead, even though guitarists make up roughly as much of the research community as pianists.
Does the history of AI beating humans at chess prove AI won't stop people from making music?
Sarmento uses chess as an example: machines have far outperformed humans since the late 90s, yet that never stopped people from playing or caring about chess, and players have even improved by learning from what the machines discovered. He is not certain the comparison holds exactly for music since music is art, but treats it as a hopeful signal.
What does Pedro Sarmento see as a genuine use for AI in music education?
Sarmento points to one-on-one AI tutoring as a strong use case, since studies show students get better results from one-on-one tutoring than from being taught in a class, and AI can offer that kind of attention at scale. He also sees value in AI adapting practice material to a student's own taste in music.
Why does Pedro Sarmento think storytelling will matter more as AI improves?
Sarmento agrees that as AI narrows the gap in raw technical skill, the story and intent behind a performance become what actually sets a human artist apart. He describes a shift away from valuing pure technical polish toward valuing what an artist has to say, which he thinks could open space for more exploratory music once mainstream AI tools saturate the simpler use cases.
I still think there's something special about music being made by humans.
Episode notes
The human element in music remains essential, even as AI transforms the creative process. In this episode, Pedro Sarmento, a postdoctoral researcher at the Center for Digital Music in London, discusses how technology can support, rather than overshadow, human creativity. From his journey as a guitarist to his work with AI and guitar tablatures, Pedro explores the balance between tradition and innovation, music education's future, and AI-generated art's philosophical implications.
Highlights:
- The intersection of AI, music, and creativity.
- Pedro's practical approach to integrating AI in music.
- The significance of human artistry in an AI-driven world.
- Innovative AI tools enhancing musical craftsmanship.
Sponsored by Allfeat: Decentralized Blockchain Solutions for the Music Industry - https://www.allfeat.com/
Produced by Amplitude Ventures Consulting: Partners in Early-Stage Music Tech - https://amplitude.ventures
Topics
- Music AI Research
- Guitar Tablature
- Human Creativity
- Music Education
Transcript
Transcribed from the recording by the production team. Names and terms may be misspelled.
Read the full transcript
Jakob
Hey guys and welcome back to the Sound Connections podcast. Today we're gonna have a topic that I've talked about a few times on the podcast, but I love to talk about it and that is AI and music. Pedro, welcome.
Pedro Sarmento
Thank you, thank you, Jacob. Thanks for having me.
Jakob
Pedro, I follow your profile on LinkedIn and you're a postdoctoral researcher in this space. And I thought, this can be really interesting. As I talked to you very briefly here before the podcast, a lot of the interviews I've had has been on a very, let's call it theoretical basis or from a very business perspective where your angle, as least as I perceive it, is very practical. Just before we go into the very, very deep end of all this, could you briefly explain who are you and what do you do?
Pedro Sarmento
Yes, so I'm a postdoc at the Center for Digital Music in London. I just finished my PhD there. My topic was mostly around guitar tablatures and AI. I'm also a musician myself or a guitar player. And as you were saying, and thank you for that, I do also think that my technique is a bit practical. Getting a feedback.
Jakob
I think it's practical. Yeah, I don't understand what that might be.
Jakob
Okay, so it's an internal feedback here in the program actually. Yeah, there's an internal feedback here, so. Okay, if anything happens, I'll just mute my mic. I'll take control of it. You don't need to care about it. So now I know the solve. Just mute my mic. There's some sort of feedback loop. Okay, so I'm just gonna ask the question again and then you give it a go again. So who are you and what do you do?
Pedro Sarmento
So I'm a postdoc researcher at the Center for Digital Music in London. That's where I did my PhD as well. My topic was on guitar tablatures with deep learning. And yeah, thank you for saying that the practical part of it, I also think that I'm a guitar player myself and my whole angle with the PhD was coming from a perspective of a guitar player, I believe so. And trying to work on AI tools that would somehow help and well, It's a bit of a cliché to say this these days, but not to replace humans in music and that's what I would try to do.
Jakob
Yeah, because that's a really interesting topic, not to replace human music. Why is that important?
Pedro Sarmento
Well, personally, because I think, well, for one, I very much enjoy making music. I would say that most of the music that I enjoy listening to is also made by humans. I still think there's something special about music being made by humans. And the way I would like to contribute to this space would be with tools and approaches and techniques and assistance for music making, not... to erase our very unique contribution to music making, would say.
Jakob
Yeah, I've been struggling with this concept so much. And for people who listen to the podcast, have probably heard me say it a few times. Like I obviously have been following, Suno, Udiio and whatever programs out the most. definitely prefer Suno because it just generates faster and this space. Yeah. Anyways, that's a deep talk, but, I've, I've been making music with my kids. my oldest son, he's very good with lyrics. He remembers everything. So we've sitting, you know, at night. made up lyrics, I sit 20, 30 minutes after he's asleep, finish the lyrics, make a song, play it for him the next morning. And I've sort of been contemplating, you know, what is the future of creativity? Because I'm, to be honest, I'm more excited about creativity than music per se. Like creation for me is the interesting aspect. And I've sort of been... It's sad, I don't know if that's the word, but I've been a bit confused about what does music take me. And I think I've just, had a period where it was just a bit like, music is going away somehow. But then again, I'm just changing my mind because I'm as confused as everyone. I was watching TV and I never do that. But there was this program on Norwegian television that was called Every Time We Meet, which is basically just artists playing music for each other. They're old catalogs. And I was so touched by it. like television, totally set up, everything is artificial, but it's still like, humans playing human music for other people. And it just struck me again, like, nah, I'm actually doubtful that human connection, even on a fabricated format like TV, would go away. I think there's still gonna be space for, and much more space than I first anticipated. How do you feel about all that? Why do you believe that there's room for people in the future?
Pedro Sarmento
I for once I 100 % agree with that. I think and I'm pretty sure that there won't be a replacement of this very human feeling that you get when you're listening to or you interact with some sort of human made music. think I also even going a bit further, I don't think that these new AI tools enter that sphere so much. the problem comes when you talk about income and remuneration, not so much about creativity or the sense of, you know, art making or human art. There are very interesting studies actually that talk about what would be machine-made art, not art made by machines trained on human art, if you follow what I'm saying, but to get a sense that art is at its core a very, or a unique human capability, at least as we understand it. So I don't definitely think that humans or this feeling that you get... from listening to human music will ever be lost. And I'm not even touching on points such as, say for example, you like a band, sometimes you like that band not only because the way they sound or how they play, but for many factors, maybe because of who they are, maybe because of what they do live or the kind of causes even sometimes that they support. So there's a whole very biological kind of aspect of your attraction to music. that I think AI poses no threat to.
Jakob
Yeah, I think you're right. But also when I look at sort of the progression, and again, I use Sunu the most in examples just because I believe that they do the best. And for example, they came out with V4, I guess, four weeks ago, which is their new model. I did a trick, I've been using Moussino since it sort of came out and I took a lot of the V3s that I have made and basically remodeled it with V4. And hearing the difference from compositions I've made or prompted, whatever you want to call it, to a V4 setting where I've spent hours and hours and hours per song, prompting it just right and changing as much as I can, to then transfer into a V4 and just keeping that uniqueness of... melody and sound that V3 had but then into the improved auditory experience of V4 and it's like, fuck, like, I don't know what to think about this because... The way that I say it is, I've worked in publishing for a few years and I say, if I was a publisher, I would hire that guy. Like I would sign him. And when technology comes so far that it competes with, you know, people who've been creative for six, seven years, then I don't know what the future holds, if that makes sense.
Pedro Sarmento
Yeah, it does make a lot of sense and it of touches upon what I was saying. When you said you would hire him, that would implicitly mean that you would, I guess, pay him as well. So that's when money, remuneration, income comes in. there, I do agree with you. I think there will be changes. There will be some sort of slight revolution, if you want, because...
Jakob
Does that make sense?
Pedro Sarmento
If you're looking at music from a more... Well, I don't want to say functional as purely functional. There's 100 % a lot of creativity that goes on in prompting a model, trimming, playing with tools like Sono. But my take on it is that if you're, at least at the stage that we are, if you're a professional musician... it's hard for you to maybe express what you want to express, specifically coming from the way or the workflow that you are used to by just using words. I think this type of approach of music making, it's great for people that want to be creative in the music space and maybe without tools like that wouldn't have an opportunity to be so. But it's quite disruptive if you think of professional musicians, bands, the way they want to express their sound, their music. It's usually hard to put in a sentence or two paragraphs. So I think we're talking about two different ways and also that both of them can coexist in this space. On the other hand, I also think that if you're doing music for advertisement or for more... Functional or commercial application like I was saying maybe you would think twice in hiring someone versus doing it yourself on Suno or audio and this will I think make some changes in the way or more on the market side of things not as much on the creative side of things I don't think personally that musicians feel threatened from a creativity perspective. I don't think any creator or musician feels like, well, these machines are more creative than me, even because there's hardly any competition when it comes to creativity, right? But there will be, in the functional space, there will definitely be some changes, I believe.
Jakob
Yeah, and one counter argument to that though that I'm sort of exploring myself right now is I'm an over enthusiastic person and I always try to do new stuff. And four months ago, I decided to start programming because there's a program that came out the cold cursor. And, you know, I had done HTML a bit when I was younger and I did study one year of electrical engineering like yourself, but dropped out. So I had like basic... programming experience, I understood what it is and I understood what a terminal was and stuff like that. But that was about it. now Amplitude Ventures has around 13 developers, full-time developers in our company where we build AI agents. And I'm really good with context. I'm really good with conceptualization. And my developers are really, really talented. I can beat any of them in speed, any of them. And I don't have a development background, but what I do have a background, I have a background of creation. So I can build full-scale products in two and a half, three hours, cut the deployed function of the internet. And none of them can do that. But of course, it's not going to be sustainable products that could be full-scale digitalized and SaaS products. So I'm not saying that I can do that. But if you compare that to the future of creators. I believe that's where the shift is going to be in like the practical aspect for creativity. There's going to be people who understand how to tell stories in a really freaking fast manner. And I'm a firm believer, I'm often wrong, but I am a firm believer in, in oftentimes quality comes from quantity. make with good intent, just making shit tons of stuff. And then you can sort of sort out the good parts. And that's sort of where I'm trying to. get at with music, I believe, even though we can't transfer that directly, that there might be some truth to it when it comes to future creators of music. That might be, again, I'm just using myself as an example, that's very musical, understanding lyrics, he might create in a fundamentally different way than I did when I learned music. And I think there's also gonna happen a shift there that's gonna change the way music is created completely, if that makes sense. So even though musicians might not be threatened,
Jakob
I just believe there's going to be a new type of musician coming and they should be threatened by them rather than technology, if that makes sense. When it comes to market opportunities, when it comes to who's going to take the production assignment, who's going to work for that artist is my assumption. don't know, but that's the parallel I can create myself. Does that make sense? does it resonate a bit with you or?
Pedro Sarmento
It does, it does. I can agree with you on that. I would still think that despite those changes, my take is that both the more traditional approach to music making and these new novel way of creating music will probably coexist. You raised an interesting question as well, which by talking about your son, which made me think of, well, how would or could potentially these tools affect... music education in the future, right? And you have other, if we start diverging a little bit away from tools like Sono, which, Udio, that do what we call audio music generation, prompt based audio music generation, you do have other tools, AI based tools that are used sometimes, or currently in high school for music teaching. I'm talking about... tools like Rave, if you're familiar with it, or Newtone, which are AI tools that work on the timbre space. So a use case would be you would give this model a melody from a violin and you would ask it to morph it into some sort of crazy sound that you've trained your model with. And if we start seeing the AI in music, I also think and know that due to, well, the mediatic press and the sheer size of what's behind these companies. Suno and Udio have been dragging the dialogue around themselves, but I would like to also take the opportunity to say that there's way more in the AI and music space than just Suno and Udio. And many of these tools can be seen as more easy to integrate in this traditional workflow. Hence why I'm saying that I'm positive that both worlds could coexist because yes, Suno and Oudio are super amazing, powerful tools these days, disruptive tools. But in this space, you also have tools that are more catered towards the traditional workflow and that musicians are exploring and also investing in. And sometimes as someone that is very, very keen on both the music side and the AI side of things, it pains me a little bit to see or saddens me a little bit to see that
Pedro Sarmento
Maybe the debate is getting so inflated because, well, we are focusing so much in this audio generation bubble that we are maybe giving a bad name to everything else around. Not that they have a bad name, but you understand what I'm trying to say.
Jakob
Yeah, I totally understand. again, I'm just in the moment right now and going back to the development use case, again, that's sort of where I'm right now. There's this new product on the market called Lovable, which is basically the build full products just through prompts. And then there's Cursor that I use, is much more like, it's still programming, it's still terminal, it's still like local deployments. And the interesting reflections about that is that any developer would still choose the tools that are AI enabled where they have the freedom. Because at the end of the day, if you want to build something really meaningful, really stable, really scalable, the influence on every single detail of decision can be very, very important. And I think that goes back to creativity and music. Going back to traditional producers, Quincy Jones just died. He was not a one man army. He knew how to direct. modern day Quincy Jones might not direct, you know, full orchestra or like a big band or the same way you might, you know, produce a lot of tools so that makes sense, but still with the same intent. And if talked about making good music that sustains for a long time, what brings something new to the market? I think the intent of the creator, even what he disposes of talent or opportunities is just so incredibly important. And I don't think that AI will take that away. But it will definitely eliminate, in my opinion, the midsection. Like the people where they don't have a lot of intent, they're just good at it, know, to some degree. I don't think there's gonna be room for that in the market. I think if you wanna exist in music, you've really gotta mean it, if that makes sense. That's my opinion.
Pedro Sarmento
I agree with you as well. I would go even further as to, if we focus, for example, on also trying to think of, discussing a little bit this distinction between the AI being some sort of creative entity or a tool. If you think of the applications of an AI in music as what people are more comfortable in calling tool. The work of producers and engineers in editing, for example, those types of tasks, often boring tasks, I don't know, in metal drumming, for example, it's very common that you spend some time very meticulously aligning everything perfectly to the grid. That's, in my opinion, it's a bit of a boring task. I'm not a producer, but I've seen people working. It feels boring. So if you have some sort of system, intelligent AI system that does that, I would call it an AI and music system. I would still think that it improves everyone's lives. Maybe. but the guy that spends his day doing that. But even him or her she, there might be some time saved there that could be put potentially into some more creative tools. I think it's also important to understand that these are tools at the end of the day. I still struggle with the notion, especially knowing what's under the hood. the notion of giving some sort of sentience to these machines. Well, I'm not saying it wouldn't be possible in the future, but as someone that kind of knows the insights of these things, it's still a bit weird to think of them as some sort of, you know, an entity that has creativity built in or intention, if you will. But on the other hand,
Pedro Sarmento
This also raises super interesting philosophical debates, right? If you can imitate something to the point that no one can distinguish it, is that enough for authenticity or no? How does someone deal with these super incredibly and Kenny Valley feeling realistic tools?
Jakob
Yeah, and... I think, and again, I like to make predictions because it's fun, but I think there's gonna be a future where people need to reason for the human to be there. Let me explain. Like, what is the intent?
Jakob
Two seconds, I'm just waiting. looks like the internet is poor or something like that. Can you hear me?
Jakob
There we go. Something else. Okay, okay. I'll just continue where I let go.
Pedro Sarmento
Sorry. Predictions, you were saying. Last phrase I heard you saying was that you like to make predictions.
Jakob
yeah, okay. So I kind of like to make predictions and one thing I'm playing with right now is I believe that in the future with music, the human needs a reason to be there. And let me explain. Right now, when you see a concert or perform music, you would expect there's, you know, multiple instrumentalists somewhere there. But I believe as technology gets more ingrained in people's life, people need the focal point. the relationship point, which means I believe actually the individual becomes even more important. So if there's someone on stage or if there's you know, human playing something, like why are you playing it? Why are you not using the machine? And you need that intent. And oftentimes the artist is like an obvious answer is you need someone for the storytelling to resonate with you. You need another human being in order for their lyrics, their music to resonate with you. And I think... Honestly, that's going to happen as technology gets more ingrained. you really want to think like Elon Musk and Neuralink and everything, I think that's going to be even more important. But that also means that, again, coming back to my point, I think people need to know why they're doing what they're doing. I went to very prestigious music schools. I a bachelor's and master's in some of the best institutions in Europe. even there, people didn't know really why they were doing it. It's everything I love like sure. Okay, fine. And then what like there wasn't enough intent in my world to survive what's going to happen. And I think that's going to be the survivors that is the people who can be a focal point have something strong to tell have a reason to be there in the music until that story. I might be taking it too far, but that's sort of how it sees right now. Like it good enough is not going to be good enough. Like it needs to be best.
Pedro Sarmento
But it's very interesting, that point, but I wonder, you know, are we talking about the music business or making a living out of music or music per se or both? One argument of, at least one story that I like to tell myself when thinking about these problems that you raised is chess. chess is one of the paradigms of AI in the late 90s, so we cracked it. Machines are way better at playing chess than we will ever be able to, and that did not stop humans from playing chess. Quite the contrary, it has still a very strong interest on this activity, and we have even gotten better from... by learning a few tricks from the machines, I believe. I don't know if it's comparable or exactly comparable to music because again, we're talking about different things. We're talking about art here. I would say as a musician, I'm okay with doing music just because I like it. It may be a very unsatisfying answer, but... You know, just as a guitar player, example, sometimes just the act of the physical activity of playing the strings, touching the strings, the muscles and having the pleasure of nailing a passage from A to Z seamlessly, at least I like to think. That gives me some sort of reward that I appreciate. This is a very biological... inspired overview on what music maybe is doing to me. but I think that something can exist just by the fact that people enjoy it and not necessarily because they have these new entity there that could potentially do everything that you just did, better than yourself. it will probably be a bit shocking and demotivating at first. but I don't think that will kill
Pedro Sarmento
human's desire or human's pleasure from music. But that's on the music making side of things. If we're talking on the other side of the spectrum, which is when you need to make money out of that, when it's a business, when you need to be the best or your revenue or income depends on the quality and your efforts, that's a different story and that I can meet you halfway. And halfway because it's, I believe, still a bit hard to define what is better or worse in musical terms, at least from how I see it. Of course, you said something interesting at first, which is the quantity argument, which is, of course, with the tools that we have right now, no human musician can compete with an AI in terms of quantity. If I start doing 24-7 music making, it's impossible. not the same scales but then I guess quality will also come in a way and sometimes not even quality just how pristine is your production, how innovative your ideas are. Sometimes you just want an artist to do something for you or you commission an artist just because of that particular subject, that particular subjectivity on that particular human I think and that Sure, it could be mimicked by a machine pretty easily even. But I mean, if you can, why would you have, if you can, if you, if you can hire hands on Zimmer to, do music for you, why would you train the machine, the machine learning model to perform a hundred Zimmer for you, if not for the business side of things? That's, that's my take on it. Of course I'm, I'm downplaying the
Jakob
Yes.
Pedro Sarmento
the money side here. I'm just trying to make a very clear divide into, well, money is at the end of the day the main issue because I still think that if that's not the case, money, accessibility, I would still prefer a human taking the lead, a human producing the art because at the end of the day, it's human art. These machines are mimicking human art.
Jakob
Yeah, You talked about chess and I don't know chess that well. I've seen the Gambit Awards ever is called and you know, I live in Hawaii so Magnus Carlsen is obviously a talking point. but that's the extent I know about chess, but I do know about running a lot just because I used to do that and I'm very passionate about it. So this might be a stretch, but I'm just trying to sort of create an analogy. Why do I enjoy running? You know, if someone was biking, they would do it faster. Is that a car or a plane? Like, it's not about speed. What is it? Well, it's in reality, it's about feet or storytelling or mastery to some degree. When I look at, know, I live 20 kilometers from Sanes where Jakob Ingebrigtsen, which is probably the world's best runner right now. And why have I followed him for 10 years? Just because it's such an interesting story. Do you know what mean?
Pedro Sarmento
Mastery.
Pedro Sarmento
Yeah, exactly.
Jakob
I'm not connected to his running times. I'm not connected to anything but the story of the story of this 11 year old kid that started running with his big brother, Sarah, who peened champions and then he comes 10 years later and is the best in the world. So why do I enjoy running? Because it's incredible storytelling. It's about people going out of, you know, the ghettos of people coming out from poor countries with nothing and becoming a global icon. So I think to a big degree, you can argue music the same way. That it's not about being quote unquote the best or the most impressive or the cleanest mix. It's about being a person that people can grab the story to and fall in love with. But. without being professional runners, they would not be able to dedicate the time and effort to become those. And I think that's the key thing for me. And that is why I agree with you that the finances is essential because any person that has that special thing needs years and years and years of cultivating it. And without money, that's simply not gonna happen. So I think actually at the core of everything, money is the is the issue because money enables dedication. Money enables that journey to create that storytelling, if that makes sense.
Pedro Sarmento
It does. Although you touched upon, well, the need of money for some sort of music education, let's say, for you to become proficient of some sort of musical activity. That, I would argue, is another very positive use case for AI in democratizing. access to music, not to music making per se, but for example, there are works, we have some people working on that at our center, where AI tutors for music, once again, not eliminating the need for a human to conduct and correct some sort of mistakes, but maybe making it easier for you or more engaging for you to learn things that you want to learn instead of practicing the same scale a bunch of times, practice the same techniques and at the same time keep the same engagement as you would and those strike me as nice positive use cases for AI in this space.
Jakob
I agree, barrier of entry but also creativity and how people learn. I think your research in sort of guitar tabloid must have covered some of this. What was some of the findings you did in your research when it comes to learning or understanding the instrument or how people adapted to the traditional way of learning music but through technology? What was your observations there?
Pedro Sarmento
Can you give me one sec, I have some people speaking outside that probably is, let me just tell him to...
Jakob
Okay. I use some AI tools so I'm never going to hear it. So if it doesn't disturb you too much, not a... No, no. We use some AI tools that we're never going to hear that. So if it doesn't disturb you too much, I can just ask the question again. So through your research with guitar taploids, what's your observations about learning using technology to learn this traditional instrument?
Pedro Sarmento
is it? Okay. I was very aware that this would ruin everything.
Pedro Sarmento
Beautiful. No. Yes.
Pedro Sarmento
It was not a particular focus of my PhD, the learning side of things. I do think that on that topic, I would like to highlight that in our space, in our AI and music research, it's very common to distinguish between the subfield that I mostly work on, what we call the symbolic side of things, where you use MIDI, tablatures, sheet music to represent, to abstract musical information into some sort of data features that can be understood by a machine or the audio side of things where you use spectrograms or raw waveforms. I personally think, well, I have to kind of believe this, I'm a bit biased towards this, but I think there's a lot of value in working this symbolic approaches, mostly because... If you want to render this, if you want to interact with this, hardly you would listen to a midi file on its own, commercially, professionally experience a midi file with a VST without anything else on it is maybe, or most of the times, bit machinic. So working on this symbolic domain often leaves space for an extra... human collaboration. On the side of education, this would be the next steps of my PhD. Sadly, it's only four to five years, but one thing I was very interested in was precisely what I was telling you before. So with tablatures, we have not only what notes are played, but where those notes are played in the instrument, because the guitar, contrary to the piano, you have different ways of playing the same notes and with these tablatures or my work you also have ways of capturing you know different ways that you can play a given note, different techniques. You just hammering on doing some sort of legato? Are you palm muting? Are you using some sort of more advanced technique like tapping for example or bending a string? These sort of specific
Pedro Sarmento
Our guitar-specific playing techniques are well captured by these format, the tablatures, not so much by traditional scores, sheet music, or even MIDI. And MIDI is the standard for most of the research in these AI symbolic music space. So my journey was a little bit, of course, this was not a journey done alone, but was try to... take the focus a little bit from MIDI, which don't get me wrong is great and a lot of progress on the field was done because MIDI is so ubiquitous and because we have so many tools and packages and scripts that work on MIDI. But I was always very surprised when thinking about researchers in our community or people engaged in our community. I would argue always that we would probably have as many piano players as guitar players and But there was a massive discrepancy in terms of research done focusing the piano and the guitar. I think historically because the piano would be easier to model precisely because of that one way direction between a symbol and a note on the instrument. But on the guitar side of things and the education side of things, tablatures, I think with the work I've done, it's possible or the next steps would be to cater to a student's needs or interests, musical interests for example, you like a specific type of music and your guitar tutor told you to practice scales until your fingers bleed but instead of that maybe you can practice the same motions in some sort of metallica kind of adaptation of riffs or I don't know, maybe if you're into jazz over some jazz chords and some jazz passades that kind of alleviate a more boring approach to the instrument. And those ideas could be extended further to the point where you'd have a system that could perhaps assess how well you do in terms of guitaristic features. How much are you able to keep steady on the beat?
Pedro Sarmento
Consistent are you when attacking nodes and provide you solutions and recommendations for studying approaches, I would say.
Jakob
Yeah, I think it's a lot of valid points. It's really interesting. When you talk about so tablature and I play guitar when I was younger and you know, was on guitar pro all the time and try to learn, but the truth of at least when I use it, but that's, call it 13 years ago, 14 years ago was it was difficult to know what to do. Like even the knowledge about what I need to learn. And I think maybe again, to the point going back to AI as sort of like more. democratized access of creating things, I think we also need to make it easier to access learning in order to keep the jewelry of playing the instrument. Like the same way that social media might have sort of given some sort of dopamine addiction to people, I think to some degree learning also needs to adapt to that. Like people, can't just expect that people have the same amount of patience and diligence that they used to have. Like... I like academics to some degree. I've, you know, I have eight years now in my Belgian academics, even though I've always been working full time on the side, finishing up my second master's now. But I believe it's a hopeless form of learning. Like honestly, some people might be able to, but just looking around at my classmates or just people around me, like this is not adapted to how people learn now. My, almost a bareness to say, but I'm... my primary source of learning is TikTok.
Pedro Sarmento
Mm-hmm.
Jakob
Honest to God, my whole algorithm is comedy and learning. That's it. And that is my source of learning. And then if I want to go to something, go deep, I listen to a podcast. Like I don't open a book. When I sit at a class, I'm just bored. And that's also the negative side of technology to some degree and dopamine rush and whatnot is I'm not receptive to that type of learning anymore.
Pedro Sarmento
Yeah.
Pedro Sarmento
Mm-hmm.
Jakob
That needs to be taken consideration into the future instrumentalists as well. Like how would they be receptive to learning the joy of an instrument and maybe becoming an exceptional craftsman than then can be a storyteller. But I'm just gonna assume, and you would know a lot more than me, that needs to be taken into consideration as well.
Pedro Sarmento
Definitely. You touched on a lot of interesting points. If we want to bring AI into the equation, I think education, the educational system is... along with the healthcare system, probably the two main systems where AI can be more impactful. There are many studies that show that when you have a one-on-one tutoring system versus being immersed in a class, your results are often better with this one-on-one approach. AI kind of enables this to some extent. Like yourself, I... do spend a lot of time engaging and trying to extract knowledge or learning if you want from tools like ChatGPT or Cloud. And just by having this one-on-one kind of conversation with this machine that I can prompt and ask, well, sometimes it's just because I want to be, or go as deep as I want, or ask the dumbest questions that I want to pose. That's one of the points. that I think will make a good use case for AI. On the music side of things, if we want to trace back a little bit before the AI enters the picture, you already kind of see how musicians, well, I'm very biased and deep into the guitar space, but you see, especially, I would say in the last 10, 15 years, you have... witnessed a next level of guitar playing and it's often from people that you know they come up online they are on instagram or they come up on youtube with amazing chops and what used to be these guitar icons from the late 80s the late 90s that had impeccable technique they would do these master classes and record these videos all that information because
Pedro Sarmento
I believe because everything is so readily available now either on TikTok, YouTube, Instagram. You can so easily pick up a lick, see someone playing it, replay it so many times. I think that impacts and impacts the learning and the quality of the performance that we have right now. And you can see that with this new generation of guitar players with crazy skills that are very unusual to what a traditional electrical guitar player. would be accustomed to. So I think we are entering that stage that you are talking about, that people are getting more knowledge, quicker access to knowledge, and those results are showing now. On the downside of it, I think you also touched a very important point, which is we struggle now, I struggle for sure, spending time quieting down, grabbing something that would require my attention undisclosed, uninterrupted for long periods of time. Books are, I would say, the perfect example for this. I think I still read a lot, but the distractions are definitely there and it's hard for maybe if you're not as disciplined... The trade-off between getting your knowledge for something so interactive and fast-pacing as TikTok, for example, versus a book is hard to justify. And of course, I still think that you have unique and valuable information in books. It's definitely worth spending the time reading there. I have great time reading it. But I do understand when some people that maybe are not... as like-minded tell me, look, I can have the same information with interactive TikTok videos. I'm mostly curious about that approach as well. And I think there's benefits in both. But at the end of the day, I think what this prompts is again, a shift in paradigm brought by technology.
Jakob
Yeah, it is an interesting space and...
Jakob
Think what I get from this conversation today is that there's most likely going to be a place for love for music. Maybe the approach to learning will be different. It sounds like it's still quite unknown from everywhere I talk with, also you with like, what are the financial implications and what are the practical implications of the financial implications in AI music? And I think it's very unknown. But again, the hypothesis I'm working with right now is I believe it will be somewhat close to the ones of developers. And the reason why I try to take that angle is when I do my research in music, music financing, startups, whatever, it's just across the board. It's so hard to figure out what's happening. As you say yourself, there was a lot of research on guitar and MIDI. If you just look at the industry at whole, there's not a lot of research in music. Like if you look at just, know, compared to what there is in any other industry, gaming, MedTech, whatever, it's still a small field, which means it's also really hard to have a fast-paced moving academia and prediction on what the future will bring. That's at least my theory. So for me, it's easier to bring power holes to where there is a lot more talk about it, where the most clever people in the world go out and make statements. And that is the belief that I have within development is, might be wrong, but there's a lot of people who... who think this as well, is that the developer as we know them will be gone in five years. I don't believe that rings true to music, but I'm gonna assume there's a good part of it that will ring true to music, which I think the separation between the musician, professional musician, amazing craftsman, and the hobbyist is gonna go even further. and everything in between is going to belong, practically speaking, to the hobbyist category. That the entrance into the professional craftsman storyteller is going to be even harder. That's my assumption. I might be wrong. Because of the financial constraints that will be in the middle.
Pedro Sarmento
I can, well, that for me, analogy with programmers leads me to, again, the more functional side of music making, if you will, which it is an important aspect, an important chunk of all music making, for sure. I tend to have also a more... maybe brighter positive take on it. Maybe it will be like chess in the sense that it will make us be better musicians as well in a sense to overcome. If the threshold is so high these days, maybe it will just push us to be even better composers, performers, arrangers. I think that also How should I phrase this? From a musical perspective, if say for example, as a musician, a performer, a live performer, there's still this threshold right now that No one is, at least from now and from what I've seen, debating or talking too much about it, which has to do with... The AI isn't yet at the stage where a live or a session musician would feel threatened. Well, maybe not a session musician, a live musician. it's hard still to contemplate or imagine robots getting on stage and performing music in traditional... instruments like humans do. But there are already projects, there's this collective from Germany, Compressorhead, which is very focused on making robots perform very accurately live. So in a sense I think that it will probably create the divide that you mentioned, maybe...
Pedro Sarmento
Playing music will get into a higher level once again. At least that was the promise as well, I think, when streaming platforms came around because revenues from CDs took a little bit of a hit. So the promise was back then that, yeah, music making, live shows, this will be the new means of substance. So the same argument could potentially be put here in the AI space. But just to wrap up my little futurology, one thing that I also like to think of is with these particular models that are very good at digesting and finding patterns in all the music that we have made, there's still one caveat to them. They are only trained on everything that existed, right? Could this potentially push artists to reinvent themselves, to try to merge, create new genres, create new bolder approaches to music, things that are not in the dataset, things that span out of the dataset? Of course, you could try to generate them with the AI tools. Most of them are catered to these combination amalgamation of genres, but that coming from... the artists themselves and trying to blend or run away from what the tools are able to provide could be an interesting aspect of it. And you also have artists that are very experimental and exploratory in this space, in the AI music. Databots have been doing, have been paving the way for us all in terms of creativity with AI in the music space. their work, think it's very inspirational. I would recommend everyone to check them out if you're not familiar with it. And that's the role of AI in music that I like to think of. It's definitely something to augment our artistic capabilities, even pushing it towards experimental avenues more than just...
Pedro Sarmento
Reproducing the same kind of thing that we've been doing just that we do it faster or quicker or cheaper. In a way, know, it's like sampling is often the argument that comes to mind, right? When it first came out, are we just grabbing things from some other artists, putting them, twisting it a little bit and calling our own? Or is this actually a... a valid means of expression, artistic tool that you can use and make your practice with. I kind of like to think that, but once again, I'm a bit biased because I'm an active part of this space.
Jakob
Yeah. It's very interesting. I appreciate so much your thoughts and sharing your other podcast. Maybe I'm left with the one last analogy, sort of because I've got to be inspired by what you said. That is, I'm going to refer back to my, I like to talk about, again, as we said, quantity can feed into quality. And when I got into, my bachelor's degree in music production, that particular school, the average age of getting in was 26 because they didn't necessarily value musicianship, but more intent and artistic expression. And when I came in... I was a bit younger than the others and let's be honest, I wasn't very great at music either. So I didn't really, well, I loved that it got in, but I was very different from my classmates. They had one like Danny Scramie's, like were big, producers. And that's sort of that school. And... In those three years, I'll say two years after, at different educational institution, I barely worked on music. I just worked on storytelling. I just worked on what do I want to say? I never became good at music. Like I'm not a good musician. I'm a pretty decent singer. But again, you know, I'm not good at music, but I've been relatively successful with my music explorations as manager, producer, publisher, and now as entrepreneur. And what did I learn? I learned how to say something. And my point is I believe that AI will make the impressiveness of skills much less.
Jakob
And really just get back to the core. So if the machine can do so much of what used to impress you, so what do you have to say then? What's your story? And this is where I believe that people consciously, subconsciously, through society or through expectations will be forced to have something to say in order to be relevant for anyone. And that's I think it gets interesting because that people at quantity are forced to have quality.
Pedro Sarmento
Yeah.
Jakob
In storytelling in order to make great music. And then you want to take that picture, then there might be amazing artists coming from AI because they're just forced to be a better version of that storyteller. Does that make sense?
Pedro Sarmento
Absolutely. If I understand it correctly, and I do agree with you, it's probably a means for us to augment our storytelling skills, no matter the craft, I would say. So you kind of make the craft less shiny, but you do put the emphasis on storytelling. And to be fair, I think that's, you know, could potentially be a different paradigm of musical quality, if this makes any sense. One where... proficiency or prowess is not as valid anymore, but the story behind it, the concept, the way you frame it, I like that. I personally less and less value what I call like the brute aesthetics, you know, how technical you are and the technicality of your productions or your music or your skills as a musician, but more... what comes out of it, the story that you tell at the end of the day, like you phrased it. So I definitely think that this could potentially kickstart some sort of divide on that space as well. And also I think that we might have tools that could potentially help us become better storytellers by giving us... new pens to write, if you want the analogy to be like that, like new tools to use to convey what you're trying to convey. Not tools that you're accustomed to, not tools that do what you do, but faster, quicker, cheaper, but different tools, more exploratory tools. I think AI is very capable of doing that. I also think it's probably not the main use case these days because Of course, the bigger players are more interested in, I think, finding a space that caters to a larger number of people. But when that is filled up or when that is done, the space for these more exploratory avenues and experimentalist approaches to AI music will emerge. And that's just new tools for the storytelling, I think.
Jakob
Yeah. Well, I think, you know, for every single person I talk with about this, it comes back to that thing, story. And obviously, I could also be leading in a steering the conversation that way, but I don't sense I do that. I think that is the natural consensus around this is it will change. We believe that storytelling would be more important. Maybe the, you said it, brute aesthetics or craftsmanship will be less important. that will probably fundamentally change educational institutions or exposure to learning that in turn will create a new generation of expressors. So. I think if you look at it from that angle, know, AI does not need to be necessarily negative on a macro environment, but still have maybe the negative consequences on a personal financial level for many people. But that's what disruption brings and music is not the only industry that's going to be changed a lot. So, but Petr, thank you so much for your talks. I think this would be really interesting. I hope that people go on LinkedIn and follow your work and your articles. I believe what you're doing is super important. I've become a convert of academia and I haven't always been that, but I do see that the importance of being thorough in understanding. So thank you, Pedro.
Pedro Sarmento
Thank you very much for the invitation. It was great, Jacob.
