Luke F. Walton Love Music More Episodes Ideas For Unlocking Upsides To AI Through Augmentation And Automation

Ideas For Unlocking Upsides To AI Through Augmentation And Automation

Love Music More · hosted by Luke F. Walton · Solo episode

The podcast feed is titled Love Music More with Scoobert Doobert. Scoobert Doobert is Luke F. Walton's primary music project.

Jump to section
  1. Listen
  2. Topics discussed
  3. Host note
  4. Selected moments
  5. Selected excerpts
  6. FAQ
  7. Transcript

Listen

Sponsor the show.

Topics discussed

Host note

Figuring out how to actually use AI as a musician is vexing, so I dig into where it's worked for me: vibe coding tools like Cursor, and a little actual Python, aimed only at automating the boring parts of the process, not the creative ones.

I walk through what that automation looks like in practice, and then push further: what if AI could help me invent a new instrument, not just speed up an old workflow? I ran into real limits along the way, including a model that falls apart past roughly 600,000 of something, and even that failure told me something useful about where the technology's edges are.

The more I pull on this thread, the more interesting it gets. I'm inviting you to try it, and fail at it, with me.

Selected moments

  • Introduction to Vibe Coding 0:57 I discuss the concept of vibe coding and its significance for musicians looking to integrate technology into their creative processes.
  • AI as an Artistic Tool 1:32 I argue that instead of seeing AI as a threat to creativity, musicians should view it as a tool to enhance and expand their artistic output.
  • Automation of Boring Tasks 3:02 I reflect on the repetitive tasks involved in music production and how they can be automated with AI, freeing up more time for creative work.
  • The Role of AI in Music Creation 5:16 I explain my approach to using AI for automating drudgery in music rather than the creative elements.
  • Introduction to New Instrument Creation 6:46 I share my vision for how AI could be used to create entirely new musical instruments, pushing the boundaries of how music is made.
  • AI Limitations and Strengths 12:02 Reflecting on AI's limitations, I conclude that its weaknesses can inspire new solutions and creative approaches in music.

Selected excerpts

I really think for the practical use of the modern musician, trying to figure out some way to use this crazy technology is kind of vexing.

~1:30 in the full interview

What I'm trying to do is kind of write some stuff for myself to take away some of the boring parts.

~2:16 in the full interview

So that's one trying to just like stay up on it enough so that I can solve these problems when other people solve them or solve the ones that I need to solve on my own.

~5:19 in the full interview

But what if the AI helped me invent a new instrument?

~10:32 in the full interview

They get really really bad after about 600,000. But it reveals something about AI's limitations.

~12:45 in the full interview

The more I bark up the tree of augmentation, the more I find interesting and intriguing and powerful.

~14:19 in the full interview

FAQ

What is vibe coding and how can it help musicians?

Vibe coding is a concept in music technology where musicians use coding tools to create new instruments and augment their creative processes.

How can AI be used in music creation?

AI can automate tedious tasks in music production, allowing artists to focus on the creative aspects and potentially even assist in the creation of new musical instruments.

What are some limitations of AI in music production?

AI can struggle with maintaining context in long sessions and it often reveals limitations that can inspire unique creative approaches.

Transcript

0:59 What's good everybody welcome back to love music more today I want to talk a little bit more about technology and where AI is going not necessarily just song creation AI something far cooler than thatinstrument creation AIThis is vibe coding. This is not

1:35 Using Suno or any of the song generators not throwing hate there. There's a lot of fun to be hadBut I really think for the practical use of the modern musicianTrying to figure out some way to use this crazy technology is kind of vexing.

1:51 It's kind of difficultIt's kind of weird andthat's frustrating because It's been positioned as this thing that's going to take away our livelihood take away our art andIn reality, it's a tool that we should be able to use to make more art to make more interesting art and then like maybe to automate some of the boring stuffThe reason why I put it that wayIs I've been reading a book

2:19 I don't do that and often not often enough clearly. I can't even speak EnglishBut the book that I'm reading is talking about automating the boring stuff using Python andOne of the reasons why I'm trying to learn the Python computer programming language is becauseArtificial intelligence is really good at using itYou're able to integrate AI into Python scripts

2:43 But really what I want to figure out is how can I take away some of the most boring parts ofMusic creation and get them off my deskBecause I don't know if I'll ever be able to afford an assistant engineerIt'd be cool if I could because there's like that mentorship shipMentorship opportunity that comes along with that and that would be really wonderful and

3:09 It would be fun to pass a torch if I ever you know how to torch pass to mebutAll said and done a lot of what the assistant engineer does is stuff that is just super annoying like the little tiny babyTuning by hand clipping out the wrong breathsColoring and organizing your sessions

3:35 Taking your tracks and then making the mix sense getting you set up to be creative because it's really hard to

3:41 Have the organizing hat on do all of the organizing especially for mixing and then to take that off and then put on the creative hatYou almost have lost some of the energy because when you confront a song for the first time it tells you somethingIt is a really special moment that I like to capture as a mix engineerTo be like okay, what does this?

4:05 With fresh ears actually evoke what am I trying to get the song to sound like?Where is it different? Where's the Delta? Where is what I'm?Like smacked across the face of this is what the song wants to beVersus this is where the song is and then boom. I have my road mapBut if I'm two in the weeds and I'm thinking about like oh god, there are 300 tracks.

4:30 Oh no Then they're not in order. This isn't in line. Oh these toms are completely out of syncOh, the phase is wrong over hereNow I'm thinking technically rather than thinking creativelyAnd that seems like a damn good place for AI to step in andSolve some of that problem. So I'm trying to think about that.

4:48 So that's one handThat's what I am hoping that the promise of AI is I'm not trying to like write Computer programs for the music industry or somethingMight be fun, but that's not really what I'm trying to doWhat I'm trying to do is kind of write some stuff for myself to take away some of the boring partsAnd then also like just stay up on it enough so that when these tools start to bubble up for example

5:09 Riley Knapp who I had on as a guest he made a wonderful company called Stemport that has saved me quite a bit of timeIt uses AI to sort stuff

5:19 Listen to the podcast. He'll explain it better than I can but essentially you're able toOrganize your stems make them understandableSo you're able to hit the ground running rather than just like being smacked up against the face with the wrong sort of thingsYou get to smack you get smacked by the creative forces rather than getting smacked by the technicalThe technical debtYou have to get work your way through

5:42 So that's one trying to just like stay up on it enough so thatI can solve these problems when other people solve them or solve the ones that I need to solve on my ownThat's just kind of like baseline stuffBut I think that that's a really fascinating and fun way to think about AI rather than automate the music creation processAutomate the drudgeryIn the same way that Python

6:02 Can do that and some of the tools that I have help getting you part of the way thereAI can get us furtherBecause it's more flexibleAnd music is chaoticEvery song is very differentYou can apply the sameStrategies to a lot of different songsBut a lot of them start in a very different placeAnd so so much of what I think I bring to the table is the vision of how to get it from point A to point Z

6:27 And sometimes those intermediate letters are spent hours of just likeMaking up for a poor recording or decisions that weren't made early in the process that should have beenAnd that's okay I think that's part of just like working with underground people is everybody's on a different stage of the journey

6:46 And hopefully AI can help us you know skip some of the worst stages of that journeyThe other thing though this is what I'm really excited aboutAnd this is what I want to talk to you about more today because we've covered the other stuff in previous podsIs instrument creation

7:01 So I really think that AI and vibe coding is going to allow people like me to write programs that are not computerPeople in the same way like I use computers a lotI'm very digitally native I can think inDigital sorts of ways and workflowsBut I don't have a ton of chops in terms of like systems integration orYou know all the fancy words of how do I make an application actually workBy taking

7:33 Some of that power through the magic of programs like cursor if you're interestedThat's a really fun and a lot easier to use than Claude Code and stuff in my humble opinionBecause you're able to assign tasks to different levels of AI's so smarterDumber and based on how hard the tasks are and you can go back and forth

7:54 And you can check its work as you learn more about coding you unlock more of the ability of these agents

8:01 All this is happening very fast and I anticipate in the year it's gonna be even more approachable than it is right nowBut what do we do with that?

8:09 I think we make new musical instrumentsI think that's where we should go andSome of the things that are really promising to me are computer vision contextual contextual Like workflows like maybe collaborating with robots or robotics something that goes beyond midi manipulationSo if you're not familiar midi is theStep-wise communication tool that powersSynthesizers it was released at the NAMM show for the first time

8:39 I think it was 1983 andIt kind of unlocked the digital to analog conversationOf being able to use things called midi controllers to be able to control all sorts of thingsAnd it's part of the reason why synthesized strings sound good nowIt's not just that they've been able to record these string players betterBut the composer is able to quote unquote

9:03 Perform by using an expression pedal or a bunch of different kind of knobs and things that you can turn to crank up different emotionsMidnote so the notes evolve just like an actual string player would play I was actually just working on thatthis weekend for something that I'm scoringandAll of that got me thinking of like Imogen Heap has her glovesThat's another like kind of innovative way of

9:31 Using midi a different kind of controllerBut what if we think even further outside of that box the different ways that we could create instruments through movementor through subtle movement?Or maybe even anticipating your next move

9:47 Like for a while I was thinking about how interesting it would be to train an AI to likeWrite lyrics like I do or think of song titles that might be interestingSo then when I get sit down to do a writing sessionI already have some song titles sorted outIt's been fun. I've done some of it.

10:06 I'veHad it write some chord progressions and things like that But I always just end up changing it and I don't know how much further along it necessarily gets meIt might even make me a step behind because now I'm not thinking creatively and freely. I'm thinking about like how to fitThe prompt that it's giving me.

10:23 But anyway, I think it's good to explore And I'll continue to but yeah, like I think that's all kind of small thinking

10:31 Like okay. Yeah, like the AI can write your lyrics like no, I don't really want thatBut what if the AI helped me invent a new instrument or what if the AI knew what I was going to play on guitarBecause it's listened to all of my guitar solos or I solo into it for like you know an hour a day for three monthsAnd then it does an adaptive delay thatThinks ahead of me

10:56 I'm just kind of swip on but like those kind of ideas I want more of that and I hope people out thereMaybe I'll come up maybe I just came up with oneBut like there's got to be something elseThat's beyond a synthesizerThere's got to be something else that's beyond a controllerLike why does every music effect have to be invented already

11:19 Why does every instrument have to be invented already and the ones that I usually see are really complicated or really sillyCan't we figure something else out?And it's not even necessarily saying that these AIs would be like a collaborator or writing the songIt's more just like opening our minds as to how to interface with theseThese flexible mini intelligences

11:46 One final point as I've been using them there is a lot ofhype around the million context window so tokens you can think of asyllable or two that's a tokenSome of the LLMs like Google's Gemini can handle a million of themAnd I'm here to report to you know they can't they get really really bad after about 600 000But it reveals something about AIs limitationsAnd every time you find a limitation you can find a

12:18 Different use case or a different strength just like a human being you knowLike hey you might not be the best singerBut because you are not the best singer you're going to write certain types of songsYour weaknesses can define you as much as your strengths in a good waySo applying that same thing to AIIt's not very good at remembering thingsIt's not very good at acting for a long period of time with a lot of context

12:45 Like a good example of context isYou walk into a room and you go wait. What was I doing?That AI does that midway through a coding session

12:54 It's like I'm just going to delete everything I'm panicking. It literally. I've seen it do that be whereGot to put guardrails around these thingsBut then it makes you think okay, what what am I bringing to this conversation?I have them decades-long context windowI remember lots of thingsI don't remember everything but I remember way more than the AI doesSo maybe there's something there in terms of

13:19 It's a lower contextIntelligence that has limitationsSo maybe we can use it not to recreate ourselvesOr to make ourselves obsoleteBut to be like little tendrilsLittle extensions of our intelligence something that's more empoweringSomething that's more flexible something that's more funAnd so for all those reasonsI am dive-in-deep.

13:43 I'm really trying and I encourage you too as well I know that it's scary but I and I also don't think that and this was my original intent was to like try all the music creation toolsThere's gonna be cool stuff.

13:58 I'm gonna try moreBut the more I bark up that tree the less I find in itThat's interesting that like really does anything for meAnd the more I bark up the tree ofaugmentationSlash automating the boring stuff The more I find interesting and intriguing and powerfulAnd so I just want to invite you to go and vent a different instrumentAnd vent a new audio plugin do something that you think you're incapable of doing

14:26 Because these tools are gonna make you capable if not now very soonSo catch that wave manGet out and paddle outI hope you enjoyed this podcastMy name is Scoobert Doobert.You can listen to my music at scoobertdoobert.pizzaI promise it is not written by AIBut maybe someday it'll be collaborated with AIOr some cool delays or something. I don't know.

14:50 Maybe you'll make itIf you do hit me up at scoobertdoobert.pizza If you enjoyed this episode rate a five star and share it with your friendsSome great guest episodes coming up right around the cornerAnd behind the last cornerSo stick around and subscribe and I will see you next week

The transcript above is a reviewed transcript of this episode (JSON). The recording is the Spotify episode.

Ask a question for a sourced answer, or type a page name to search. Try: