The number of responses that you're getting that some people would prefer a rote and repetitive act like folding bloody laundry over the act of creation makes me think I accidentally stumbled through the fun-house mirror equivalent of Lewis Carroll's looking glass.
People are different. I doubt the majority would prefer that over _every_ creative output. But I would 100% prefer 8h of folding laundry over inventing stories (aka creative writing) and I bet there are dozens of people who hate making music just as much as doing laundry ;)
Robotics is capital intensive and doesn't scale the way a software startup does. It's akin to starting a new car company but for a product category that has not yet been successful in the market - very risky. Then combine this with inevitable competition from China, who has dominated recent robotics-adjacent consumer product categories: drones and robot vacuums. I don't blame investors/founders for being hesitant.
Edit: Oh yeah, and then there's safety/liability. A robot capable of doing laundry is also capable of breaking your finger(s) or poking your eye out. Just imagine the TikTok "challenges" kids would come up with interacting with such a device.
I dont think you understand just quite how much harder a laundry bot is than simply stealing everyones work and training a model to learn an approximation of the source distribution.
Where would you even get the training data for starters? Not like people tend to wear mocap gear to do their chores yanno? Not like theyve been doing it the last few centuries, cataloging and digitizing everything dilligently.
Solving washing machine bot would get us a lot closer to AGI than I think llm's / music / art bots will in the forseeable future.
You make some good points, and I think I agree with them. I know robotics tends to be harder [1], but I don't understand how much harder it is. Visual-Language-Action models [2] have been gaining traction recently, along with training in simulated environments. Fine-tuning VLA models for specific purposes may reduce the data requirement. Plus, it's pretty cheap to record someone washing the dishes.
I didn't down vote you, but I assume it's because at best, you arent saying anything that adds anything interesting, and at worst you're just being contrary. You hated playing the clarinet, so what?
> I’d argue that some people like putting doors on cars in an assembly line, that it provides for their family
What? That doesn't mean that they inherently like "putting doors on cars" or working in an assembly line - they're doing that task because it puts food on the table.
It's like watching a bunch of people bobbing up and down in the ocean after their boat capsized and saying "wow they must really like swimming!". uh... no. They're doing it to avoid dying.
This sounds nice, but how many people out there actually want to do "writing and art" purely for their own enjoyment? I know they are out there, but I wouldn't actually place a bet that that number is bigger than people who enjoy doing laundry.
I think people are missing the forest for the trees - if AI does become the ultimate media wish-fulfillment machine, and I have no need for the creative visions of others - what else is there in life besides consumption and chores?
> This sounds nice, but how many people out there actually want to do "writing and art" purely for their own enjoyment? I know they are out there, but I wouldn't actually place a bet that that number is bigger than people who enjoy doing laundry.
Quick clarifying question, do you think the number of people that enjoy doing laundry is comparable to the number of people that like making art?
Making art to share and impress others is one thing, but OP's quote implies that it's only for people's self enjoyment.
I don't doubt that there are people who enjoy painting and never plan on displaying anything. But I also don't doubt that there are people who enjoy doing their laundry.
Why would you assume people only want to do art for self enjoyment? That was not my interpretation of the quote at all
I mean, if you have ever been around theater kids for a minute you’d know they are the most extroverted people on the planet and bringing joy to as many people as possible is exactly what drives them
Creating music that you yourself love -- even music no one else will ever hear -- is a peak human experience that I wish everyone got to experience.
The joy of creation is truly wonderful for its own sake.
Can AI help that creation, just like other "synthetic" techniques for making music (not least - the synthesizer)? Definitely.
But for me at least, the less of me (i.e. other people's samples, stock sound effects, DAW gimmicks) that's in my music, the less rush and joy I get from it. The less it feels "mine". But that's all very subjective.
Which takes me back to why this is such a dismal take. Why wouldn't this dude point to how making music is a creative joy, and AI can help bring that joy to more people by lowering the barrier to entry?
Love your point here - the non verbal-ness of it feels like an increasingly rare thing in our lives (especially in tech).
Music and I suppose other traditional 'fine arts' (painting, drawing, sculpture, etc) are these windows into creation that's not as mediated by language, which is arguably the defining feature of humans.
I suppose dance / ballet / etc would be another similar space too, a kind of physical creation not mediated by language.
Towering above all other art forms, though, remains music - with such an immediacy to it (both in terms of how it hits your senses and how easily it seems to convey emotionality). It's no wonder that it looks like it's the top form of 'art' made and enjoyed by humans these days.
Despite all its worth, Suno doesn't approach the problem well.
After a few laughs and cheap replicas, people realize that it's damn hard to produce a good sounding, creative piece. Suno almost always adds noise and you can feel most of this is coming from fingerprinting.
With Suno and Udio, you lose control. You can generate starters and helpers but sooner or later, you want real control. That's not editing a section and having a conditioned piece of garbage seemingly fitting to the rest. No, control means completely changing progressions, sudden but calculated change of beats, removing any instrument for the shortest time and putting it back with razor-sharp studio detail.
I know a few of these are already addressable, you can take the output, separate into channels (if it's simple enough), quantize, edit and have a good one. Yet, you're not really supported anymore. What should have been was these other core music production software to get cheaper and/or far more effective.
Suno and Udio is a top-down approach. Maybe one day Logic Pro, Ableton, Melodyne etc. will fill in the details up to this point, coming from the ground up with AI, I don't know. We're not there yet and it just brings down the mask of mainstream music industry with its all-repeating shallow beats marketed to hell. Hearing mainstream was awful but it suddenly got even more awful.
I don’t believe any of these companies are seriously trying to create a tool for musicians/artists.
When they say they’re catering to “artists”, they want to say that you can become an artist by promoting their models for the small fee of X.99/mo.
Their goal is to train a model of the well-trodden one-shot text<->art form, and just sell prompting access directly to the mass market consumer audience. Way less work, way bigger market (and less discerning, too)
That's... an odd clarification. I mean, yes, a lot of people drop out of music because it's hard to get good at it and stay good at it. But how does Suno AI solve that problem? How does Suno make it easier for someone to learn how to express themselves through a musical instrument? How does it make it easier for people to enjoy the craft of sound?
This doesn't work as a clarification because it doesn't explain why his original comment was relevant when talking about his company. What would explain it is if he meant exactly what the interviewer took it as: that he believes more people want the end result of musical creation—the musical piece—more than they do the creative process itself. Because that's what Suno AI does do: it shortcuts the path to statistically average sounds.
> I mean, yes, a lot of people drop out of music because it's hard to get good at it and stay good at it. But how does Suno AI solve that problem? How does Suno make it easier for someone to learn how to express themselves through a musical instrument?
The same way DJing does? That's a form of music "creation" which involves less skill than mastering a musical instrument.
Pedantic but important distinction - most DJing is really more comparable to "performative art" then it is to creation.
And the answer is, while Suno/Udio does make it easier to create music, it does not make it easier to express oneself through music. The disconnect between the input medium (text) and the output medium (music) is simply too vast to allow for much individual control.
DJing is not a form of music creation, it's a form of music playing.
If you mean free-form mixing and music creation based on samples and synth some DJs do on stage, then yes that's music creation, and it involves skill.
Interestingly, even in music schools there is a kind of snobbery from composers that performers are lesser musicians
That is, if you are world class pianist and tour the world playing classical music, you are still lesser than a composer who creates their own works. Is an actor not an artist because they didn’t write the lines and aren’t directing the scene?
Interpretation is art. Performing is art. That includes DJing at some level, because at the end of the day we are all using the medium of music to express ourselves and bring that experience to others, regardless of who wrote the notes on the page or put the track on vinyl
> you are still lesser than a composer who creates their own works.
That concept goes back to "There's nothing remarkable about (the piano). All one has to do is hit the right keys at the right time and the instrument plays itself." -- Johann Sebastian Bach
DJing is absolutely a form of music creation. Maybe not always in the moment of performance but the curation, mixing techniques, resampling, etc. wildly confident takes on this all throughout this post.
"Music" in this case has such a high quality bar. Truly, the delta production value of popular music today vs. even 20 years ago is huge. So our standards are higher, and if you're a solo musician trying to reach these high standards you will most likely burn out and hate the process.
But maybe I'm over-thinking it. Musicians have told stories of giving up everything for their music over and over again. It can be one of those tortured loves.
I don't know, I agree that production value has jumped in the last decade but these trends are often cyclical. The overproduced boy bands of the 50s and early 60s gave way to aesthetically simple folk singers like Bob Dylan, and the shiny, glittery pop-rock of the 80s got swept away in a cloud of secondhand smoke by Grunge. I believe another one of those counterculture, indie backlashes is due to hit soon.
It is really crazy the level of knowledge that goes into production today and I don't see it going back. New grunge will re-amp in the computer if for no other reason than because software amp sims are cheaper than hardware and your settings/values are instantly perfectly recallable in a sim.
Even modern metal requires insane specific technical knowledge nowadays. Dance genres are to the point I don't even understand the differentiation (in xyz the kick drum transient click goes 'clihhhhhck' but in xyy it goes 'cllihhhhhckk').
Arguably the indie scene is bigger than it’s ever been. I think the issue is more that, like social media, streaming music services like Spotify and YouTube keep everyone in their bubble. Even if you try to break out the algorithm pulls you back in.
Radio was a much different paradigm that had to respond to changing desires or risk losing revenue and going out of business. So it had to react to listener desires with much more finite resources and that led to (I think) larger cultural shifts within music than we see now.
I see these tools as another instrument or a synth. It should let 'me' create something rather than just do one-shot text-to-music. I would like to have a piano with intellisense (predict next few notes after I play my dumb keys), does this exist yet?
Academically, at least, yes. One of the latest versions of this kind of idea I am aware of is https://arxiv.org/abs/2306.08620, although at more than a year old, this work might be out of date. I am sure there is ongoing work commercializing these kind of approaches, but I don't know of any in particular.
In the sense of modern DAWs/VSTs, sure, but in the sense of the synthesizer as a standalone instrument, its absolutely not designed to reproduce the sound of other instruments. Its an electronic device that outputs specified parameters. Whether that happens to coincide with parameters set by other artists/instrumentalists to 'reproduce' other tonal profiles, that's a byproduct, not a feature.
The spirit of the comment you're replying to is admonishing AI's use of pre-existing melodies, harmonies, chord progressions, and musical themes from songs that were fed into it.
Be it Beethoven (which is in the public domain so hey, free game) or Sabrina Carpenter (very much not in the public domain, but arguably not musically complex), its taking pre-existing tunes and regurgitating it into something else.
Try this: go out and take some sounds off the internet, don't pay the people who recorded those sounds and make a synthesizer out of them, and sell it. Let the people whose recordings you stole know what you did, and see how that works out for you. The lawsuit will be straightforward.
Now, go out and take those same sounds off the internet, don't pay the people who recorded those sounds, and use them to train an AI, and sell it. You're still a thief, but now it's much harder to prove you're a thief, and lots of AI investors will invest money making sure that courts decide you're not a thief.
My digital piano literally has soundfonts designed to mimic other instruments. A Fender Stratocaster, a Rhodes electric organ, etc. All of them are clear and blatant ripoffs of instruments.
The recordings may be legally protected. But if I use a sound engineer to listen to those recordings to fake these sounds, it's not that much different than an AI doing it.
Amazing. You've found a way to devalue the labor of both the artist AND the sound engineer in one go. At least you're consistent in not wanting to pay people for their work!
Yeah, but there's many, many synthesizers that don't work like this - and many people actually wouldn't even use the term "synthesizer" for this kind of instrument.
Well now you can't go without mentioning sampling machines, a way of taking audio samples from others (or yourself, any audio sample) and using it as an instrument.
.....which also did lead to numerous legal battles around sampling and artists paying royalties for sampling, so in general "society" didn't go "oh sampling is fine to do all willy nilly" out of that, it made artists pay other artists for using their work.
Lots of us enjoy playing an instrument, improvising, etc. Even if an AI can do it better, I don't care, because just listening is not the same experience.
This was not meant to be a statement about people who enjoy making music but that in general people don't suddenly become more willing to make their own music just because there is an AI tool for it.
It still takes the will and desire to do it even if the barrier is so low as it is now.
Speak for yourself. I find practice extremely soothing in a strange way: it requires utmost attention and awareness of body while also maintaining a critical ear and a willingness to continually problem solve. Also, you need a lot of persistence to grind out enough repetitions to build muscle memory. This sort of skill/competence-building is very grounding for me.
And that says nothing about actually playing with other people, or improvising.
I didn't mean that people who are making music don't like making music, I meant people who don't want to play and instrument or sing or user a DAW or any other kind of conventional music making don't suddenly become interested in making music, even if they do enjoy listening to good music
IMHO the real odd part is "I think the majority of people don’t enjoy the majority of the time they spend making music.", making it sound as if there are tons of people making music despite not liking it. That's a lot weirder than just saying that some people don't like making music.
As a drummer myself I don't mind much about the type of music Jojo Mayer plays for, but not only he's reached an unmeasurable level of technicallity in his playing but also he's uttered a gem of wisdom that applies so hard for this era:
"If we surrender the thing that separates us from machines, we will be replaced by machines. The more advanced machines will be, the more human we will have to
become."
(I found out this quote by pure sheer of luck while doing some stuff with ConTeXt - it happens it's included in its sample texts)
> "It’s not really enjoyable to make music now […] It takes a lot of time, it takes a lot of practice, you need to get really good at an instrument or really good at a piece of production software. I think the majority of people don’t enjoy the majority of the time they spend making music."
I'm also a musician myself (see e.g. http://rochus-keller.ch/?p=1317) and I have many friends who are (professional) musicians. Learning an instrument is a lot of work and sometimes indeed tedous, but I think every (true) musician can confirm that nothing compares to the joy when the band "clicks" and the musicians toghether experience the moment when great music arises. The only reason that (professional) musicians occasionally lose the joy of making music is their existential dependence on their profession and the pressure to get a decent income with the mostly low-payed gigs (if any at all). But that has nothing to do with the music itself.
Tools like Suno are more suitable for and more likely to please people who have not learned the craft of music, and ultimately tend to lead to musicians having even more existential stress and thus losing the joy of music even more quickly. But we can't stop progress.
PS: I'm surprised that and I don't understand why the post was flagged; it is definitely a topic that's interesting for "hackers", and as engineers and technology enthusiasts, we should not close our minds to the implications of our work.
I love AI and I love making music, but I find music generators like Suno vaguely gross. It can make better music than me and I probably couldn't tell it's AI, but it's an emotionless experience. Text-to-music feels more like replacing humans than empowering them.
The biggest tell by them is how gleeful they are about the possibility that AI could lower the bar of competence. If they had the skills, they wouldn't need to be frothing at the mouth for AI.
I don't think human-replacement is most people's goal. I think ordinary people want to express themselves and generative AI is a promising technology to help with that. Unfortunately, I think current paradigms around LLM/genAI products favor automation over creativity and human involvement.
All that said, human-replacement is certainly the goal of many businesses.
> think ordinary people want to express themselves and generative AI is a promising technology to help with that.
Ordinary people aren't training AI, and building bots to rip training datap from ever nook and cranny of the net. The people who are are specifically targeting not areas that everyone can agree there is utility in aspirationally automating away (laundry, low skill drudgery), but the very jobs that yield the biggest drain in the conventional "money vacuum" that is the modern business.
Abandon the naivete. Listen to what they do, not what they say.
"It's not really enjoyable to make music now [...] It takes a lot of time, it takes a lot of practice, you need to get really good at an instrument or really good at a piece of production software. I think the majority of people don't enjoy the majority of the time they spend making music."
- Mikey Shulman, the CEO and founder of the AI music generator company Suno AI
From the article: "It’s not really enjoyable to make music now […] It takes a lot of time, it takes a lot of practice, you need to get really good at an instrument or really good at a piece of production software"
Music creation (like many things) is an intensely personal process. The feeling of accomplishment COMES from the blood, sweat, and tears that you put into crafting a new piece. It does not come from the ability to type "4-voice hurdy-gurdy fugue in d-minor (vocal accompaniment by Grandmaster Flash)" into a textbox and sip your latte while an AI cranks out song after song on an assembly line.
it makes me sad that having to get really good at something is presented as a negative. also where is the "now" coming from in that statement? now that there's so much slop?
An AI assistant lessens the fear of the blank page. It gets you started creatively so that you can work and it offers ideas which, even when they are bad, can lead you to something creatively satisfying. The important thing is that the AI is only an assistant and not the complete author. The human has to guide and tweak and rewrite a TON.
I wish AI companies' CEOs would leave creatives alone. They really know nothing about artists' journey, struggle, development, ups and downs... the whole story that makes artists interesting and lets fans connect with them.
I'd love to be able to upload audio / have it analyzed in the DAW and then for ai to create riffs on it. SUNO is amazing at this - but it's useless without the midi output / DAW integration.
Art is an expression and representation of its creator. AI can neither self-express nor self-represent, so it can’t make art. It can create pretty images or pleasing melodies, but it’s still not art.
I picked up the cello about a year ago and already it's more enjoyable than most other things I do. The idea of replacing that with AI is silly. Why stop doing something I love?
I firmly believe that use of AI in music production is solely for the sake of monetization. Bona fide instrumentalists are playing their instruments, practicing, getting better, playing with other instrumentalists, and not implementing AI unless generating something they intend to use to make money. No actual musician I know is using AI in a serious musical context.
> And so that is first and foremost giving everybody the joys of creating music and this is a huge departure from how it is now. It’s not really enjoyable to make music now […] It takes a lot of time, it takes a lot of practice, you need to get really good at an instrument or really good at a piece of production software. I think the majority of people don’t enjoy the majority of the time they spend making music.
Hot take: he's not wrong, but his phrasing is weird.
His quote was not about "musicians" but people with no musical background - he's not even assuming the hypothetical person can even play an instrument. And yeah, for a person with no musical background or knowledge, making a song from scratch would be pretty miserable.
As a counterpoint though, the kinds of people who do want to make music generally have self-selected themselves into a group who have some abilities to do so.
Counter-counterpoint: I literally just had a conversation with a friend (and amateur musician, mind you) and he said he just wished for a Metalcore version of a classic Christmas playlist. He's not going to go through all of the effort of making that by himself, just for himself to enjoy.
Music is becoming extremely personalized, and even the concept of a band is dying. So this very well could be the future of music.
Not only that, but getting to authenticity in musical expression is hard and kind-of the point if it, one of the big challenges of composition is I think to get as far as possible from what an AI could have produced based on current music, current taste, current everything. Case in point: Eurythmics "sweet dreams" written when the band was ready to give up and found their voice, or Damon Albarn's letting it all hang out with his Gorillaz departure from Brit-Pop...
How much self expression is there in an artist having engineers and producers completely define their sound, and in most cases having songwriters that produce the skeleton of lyrics/melody?
Music is about a bunch of people in a room sharing a feeling and a beat and creating and enjoying something beautiful. People are addicted to their media feeds, their screens, their ragebait and clickbait, and do you know why? Because they don’t have real people, friends, in the room sharing real time. Music can do that, AI is a perversion of it.
The latest wave of capitalism is that they have successfully invaded our personal relationships and replaced and monetized them. Let’s not do that to music too.
There's something so dull at this point when people mention "impact", "billions of people", etc. Just another thirsty CEO trying to extrude money from the system. Who can blame him. Idk, it's just boring is all I can say.
But I don't think he's wrong necessarily. A friend of mine always said, "The artist of the future will only have to point." I imagine things will continue to progress towards that until we arrive there, wherever that is or whatever it flips into.
In other words, I don't know where it's going, but it's definitely going there.
People who do make music sure don't like it when Ai is trained on their copyrighted recordings without their permission and for $0 compensation. What is happening is the music version of the game played by amazon with its vendors. See a hot new thing trending, fine tune train you model on it and then release an ai "artist" to compete with a real group to steal their popularity. Just like Amazon sees a particular flashlight is popular they come out with their own version of it and bury the original while highlighting their own product at the top of search.
Well they don't have to worry, because the RIAA is incredibly litigious and if the AI produces an existing 5 note progression and someone shares it, that's considered copyright infringement.
I'm surprised not to see anything about David Cope, who is the real master of AI produced music. Cope works mostly within the classical tradition, using AI to produce work in the style of great composers like Bach, Mozart etc. To my ear, it lacks the brilliance of the originals, but some of it is quite good. Certainly several orders of magnitude better than Suno's AI slop.
I think some people will never reach a sufficient interest in music to where they can produce songs that reach their level of taste. But those people can be far more experienced and interested in a different craft, like programming. So services like these are there to make up the difference in interest for those who don't want to try hard in one domain, but don't see their lack of interest as a problem. It's just a consequence of limited attention budget to devote to so many crafts, and serves as a form of automated outsourcing.
Cultivating interest in something you're not interested in is hard. Especially when you care more about N other things already. People who use services like these to produce "slop" wouldn't have produced any songs in the first place otherwise, and I feel a lot of them couldn't force themselves to if they tried. At one level, if the creator is personally satisfied with the output then the bar has been cleared, regardless of what others say. And it's hard to judge someone for choosing programming over music but wishing they could dabble in the other with some level of efficiency. At some point, if you put the bar low enough some people are going to be intrigued, push button and receive song, as nothing will stop them from doing so anymore. That's the point of the bar being lowered.
However, I think the calculus changes when those people start to give themselves the air of human artists and upload their songs to streaming. This I believe deserves critical observation since in my opinion art is more than being a human pumping out songs like a factory and watching numbers go up. It's a form of elaborate communication between parties and peer communication is fundamental to our species. So when machines speak for people, a lot of social assumptions that have existed for centuries are overturned. Anyone could have been a human factory pumping out songs detached from the artistic process of reflection the past few decades, but spending hours instead of seconds producing things in such a vacuum is pretty soul-draining, so it was a disincentive for people wanting to do so.
I believe the two types of creative output (handmade and AI) have different "qualities" to them. I do not think either can be equated in terms of this "quality". One is deliberately produced and one is randomly stumbled upon. I think we need a new term for the AI era that encapsulates this "quality" of intention in creative output, without the negative implications of a word like "slop".
I absolutely can't stand this sort of drivel from people who have never had the pleasure of partaking in a truly creative activity and experiencing how that feels. To them it's just something to be optimised away and automated for the 'benefit' of bringing it to the masses.
“I don’t want AI to do my writing and Art so that I have more time to do laundry
I want AI to do my laundry so that I have more time for writing and art”
This guy clearly doesn’t know any musicians. The starving artist archetype exists exactly because people will sacrifice everything to be a musician.