Is the Future of Music a Chip in Your Brain?(wsj.com)
wsj.com
Is the Future of Music a Chip in Your Brain?
http://www.wsj.com/articles/is-the-future-of-music-a-chip-in-your-brain-1449505111
26 comments
I came here to say this. Algorithms are so horribly bad at predicting what I even like, how could they ever get 'what I want to hear right now' anywhere close?
I could easily make such an algorithm that would work well for myself. Tag songs with a mood, have different modes that switch what kind of mix I want. Add in the ability to, in the player, designate certain songs to not play in the currently playing mix, and I think I could iterate my way towards never having to skip a song again. It's on my list of things to build 'one day'.
But that's a far cry from being able to generalize over what people that aren't me would be like. It's a hobby of mine to ask random people how they consume music. We'd essentially need the power of a full-on programming language to make object frameworks with customizable predicate logic for each individual person that they would have to learn and implement, to improve upon the skip button. The domain is that hairy.
Mind reading is not for the faint of heart.
I could easily make such an algorithm that would work well for myself. Tag songs with a mood, have different modes that switch what kind of mix I want. Add in the ability to, in the player, designate certain songs to not play in the currently playing mix, and I think I could iterate my way towards never having to skip a song again. It's on my list of things to build 'one day'.
But that's a far cry from being able to generalize over what people that aren't me would be like. It's a hobby of mine to ask random people how they consume music. We'd essentially need the power of a full-on programming language to make object frameworks with customizable predicate logic for each individual person that they would have to learn and implement, to improve upon the skip button. The domain is that hairy.
Mind reading is not for the faint of heart.
Spotify's Discover Weekly has been really good -- for n = 7.
Hmm, I listened to it for a few weeks and it was basically the same playlist every Monday. I'll pick it up again - thanks for the recommendation.
So there's several different questions here:
* whether machines will be capable of composing some kind of appealing music. This is already possible, and likely to get better/broader.
* whether machines will get so good at this (on demand, no less) they will crowd out contributions that are largely human-crafted. This seems uncertain to me, maybe more likely at the commodity and conceptual ends of the market, maybe less so in craft-niches.
* whether music is really a personal experience or a social one -- or, since it's obviously both, to what degree people want both things when.
* whether machines will be capable of composing some kind of appealing music. This is already possible, and likely to get better/broader.
* whether machines will get so good at this (on demand, no less) they will crowd out contributions that are largely human-crafted. This seems uncertain to me, maybe more likely at the commodity and conceptual ends of the market, maybe less so in craft-niches.
* whether music is really a personal experience or a social one -- or, since it's obviously both, to what degree people want both things when.
I think this really depends on your definition of appealing. Machines are already capable of _mimicking the aspects of songs which are currently popular_. I highly doubt that a machine could, say, invent a new genre of music that had widespread appeal.
The usage of machines in music currently leads to a proliferation of basically glammed out pop rock with pitch perfect voices metronome timed beats and wildly predictable song anatomy. The author mentions that Sinatra's voice has been reconstructed and has released new hits... But isn't 90% of Sinatra's appeal the broken, manly _imperfection_ of his voice?
In short, I hypothesize that machines could adequately "fill the void--" ie. play your running music, play your driving music, give you some background tunes to study or code to. But to anyone that actually pays attention to music it will be wildly subpar. Imagine a machine putting off the same lyrical heat as Kendrick Lamar or Ghostface? How can a machine possibly get on a novel level like that when it's basic coding is necessarily to average the means?
The usage of machines in music currently leads to a proliferation of basically glammed out pop rock with pitch perfect voices metronome timed beats and wildly predictable song anatomy. The author mentions that Sinatra's voice has been reconstructed and has released new hits... But isn't 90% of Sinatra's appeal the broken, manly _imperfection_ of his voice?
In short, I hypothesize that machines could adequately "fill the void--" ie. play your running music, play your driving music, give you some background tunes to study or code to. But to anyone that actually pays attention to music it will be wildly subpar. Imagine a machine putting off the same lyrical heat as Kendrick Lamar or Ghostface? How can a machine possibly get on a novel level like that when it's basic coding is necessarily to average the means?
You go to a concert and rather than listening to a live band, it just plays the same song in everybody's head really, really loud so that you can't hear anyone talk.
Silent Disco?
Been thinking about something else: stream music that's playing in your head now, so that others can hear it. No AI, no "adaptive" crap, just what's being composed in your mind right now. Whether this is typical or not I'm not sure, but I almost always have some music playing in my head, which I can't always transform into a track, as I'm just a hobbyist music maker, not a pro.
If I were rich I'd establish a big fat prize for this technology, seriously.
If I were rich I'd establish a big fat prize for this technology, seriously.
I imagine that would be quite difficult. As far as I'm aware, we currently do not know if your brain is doing the same thing as mine when we hear songs in our respective heads. So first you have the problem of extracting the information. But then you have the potentially bigger problem of translating the information!
We can be fairly sure that when you hear music in your head, nothing is actually vibrating in your head to produce the sound. My guess is that you have a bank of "instrument memories" and "note memories," and you can imagine new music by smashing a note memory into an instrument memory. (Similarly, try to imagine President Obama saying "My fellow Americans: even electronic brain pancake crystal elderly." It's surprisingly easy to combine your memories of his voice with your bank of known words.) But what guarantee is there that my instrument memories are the same as yours, or that they are stored in the same way? The mapping is not clear.
We can be fairly sure that when you hear music in your head, nothing is actually vibrating in your head to produce the sound. My guess is that you have a bank of "instrument memories" and "note memories," and you can imagine new music by smashing a note memory into an instrument memory. (Similarly, try to imagine President Obama saying "My fellow Americans: even electronic brain pancake crystal elderly." It's surprisingly easy to combine your memories of his voice with your bank of known words.) But what guarantee is there that my instrument memories are the same as yours, or that they are stored in the same way? The mapping is not clear.
If your "instrument/note memories" thesis is correct, then it's not too hard to imagine that Brainr's initial setup would involve a calibration process to work out what your* particular mapping of sound memories is shaped like.
* to give this putative "plays the music in your head" app a terrible name
* to give this putative "plays the music in your head" app a terrible name
I've also dreamed about such a thing, which would open new worlds for performance and composition. I think the technology isn't even remotely close to allowing this, though. The closest I can think of is research that has been done where MRI imaging and analysis were used to reconstruct images being watched by a subject. Results were very rough, though:
http://news.berkeley.edu/2011/09/22/brain-movies/
And you have to consider that nonexistent music playing in your head has a special "unfinished" quality that might make it impossible to just translate into sound. My take is that when you process sound, it goes through multiple layers or depths of analysis, and imagined music takes place at an already deep level of analysis - could be very problematic for this project.
http://news.berkeley.edu/2011/09/22/brain-movies/
And you have to consider that nonexistent music playing in your head has a special "unfinished" quality that might make it impossible to just translate into sound. My take is that when you process sound, it goes through multiple layers or depths of analysis, and imagined music takes place at an already deep level of analysis - could be very problematic for this project.
Makes me think about: https://en.wikipedia.org/wiki/Musical_hallucinations
I wonder what actually qualifies as a musical hallucination? I listen to some fairly experimental music. I often hear sounds that remind me of that music, to the point that I perceive those sounds as music themselves. John Cage taught me this.
Want to remember a nice song ? Thanks to DRM 5.0 implants only licensed persons will be humming away. In release 6.0 non-licensed songs will also be cut from the audio feed automatically.
Haaaapy b.. what, how did that go again ?
Haaaapy b.. what, how did that go again ?
Not to mention that this could be quite an effective torture device which makes the entire idea horrifying.
I was thinking quite a bit about this recently and it is closer than many think. Already techno, ambient music is made almost automatically.
To help you understand, there are two ways, one is using so called 'expert systems', this is computer algorithms guided by humans, and other would be artificial intelligence, where computers completely generate music on their own.
For the most part, ES is done today and works well, AI can be done for ambient music. Once we plug our feedback, how we feel and what kind of music we would like to listen to, that is when things will get really interesting.
To help you understand, there are two ways, one is using so called 'expert systems', this is computer algorithms guided by humans, and other would be artificial intelligence, where computers completely generate music on their own.
For the most part, ES is done today and works well, AI can be done for ambient music. Once we plug our feedback, how we feel and what kind of music we would like to listen to, that is when things will get really interesting.
We are not going to be able to put a chip in the brain to detect your mood for decades, and even then will it be worth doing?
Surgery is invasive, dangerous, your body is corrosive, putting stuff in your brain will have side-effects and what are you going to get out of it? Music to suit your mood?
I totally agree with Nick Bostrom[1] on this one, it's not happening any time soon.
[1]http://www.amazon.com/Superintelligence-Dangers-Strategies-N...
Surgery is invasive, dangerous, your body is corrosive, putting stuff in your brain will have side-effects and what are you going to get out of it? Music to suit your mood?
I totally agree with Nick Bostrom[1] on this one, it's not happening any time soon.
[1]http://www.amazon.com/Superintelligence-Dangers-Strategies-N...
You don't need to put a chip in my brain to detect my mood, simple skin detector, like your watch should be able to do that.
Also, brainwaves can be detected with the headband, these are general rhythms, so you can't get specific centers, but it is getting more useful every day.
Anyhow, I think I understand where your skepticism originates, it make sense, I think a lot of smaller advancements would allow us to come closer to this.
Anyhow, I think I understand where your skepticism originates, it make sense, I think a lot of smaller advancements would allow us to come closer to this.
Any examples of techno/ambient generated by ES? I knew the ES had a big uptick in popularity in the 90s but I wasn't aware that people were still actively using it for things other than customer-service and similar question-asking interfaces.
Some of the stuff on Warp is computer-assisted - in fact it has been for a long time. But we're really only talking about semi-random loop cut-ups and effects sequencing.
Ableton has a small suite of algo-comp packs. And there are things like WaveDNA's Liquid Music.
I'm not aware of anyone making complete commercial tracks using ES - at least, not with a high profile career. (I suspect Bjork does a bit of this, but I'm not sure how much. Holly Herndon certainly does, but isn't very mainstream.)
Strange fact: people have been attempting algo-comp for decades, and it remains incredibly obscure and ignored, even though there have been literally hundreds of languages, toolkits, and frameworks. That's partly because most of it isn't very musical - at best it sounds like a random mash-up of music slices with no overall sense of form or mood, and at worst it sounds like musical nonsense.
It could also be because successful systems become so complicated they're only really useful to one composer. No one else has the time or the patience to learn them.
But more, musicians seem to have a block about the idea - e.g. WaveDNA have literally had a lot of hate for LiquidMusic.
Even though many genres are very formulaic, abstracting those formulas in a useable way turns out to be very hard, especially if you're aiming for a definite mood instead of a random musical squiggle. Most people who make music would rather do it by hand than have a computer do it for them.
Ableton has a small suite of algo-comp packs. And there are things like WaveDNA's Liquid Music.
I'm not aware of anyone making complete commercial tracks using ES - at least, not with a high profile career. (I suspect Bjork does a bit of this, but I'm not sure how much. Holly Herndon certainly does, but isn't very mainstream.)
Strange fact: people have been attempting algo-comp for decades, and it remains incredibly obscure and ignored, even though there have been literally hundreds of languages, toolkits, and frameworks. That's partly because most of it isn't very musical - at best it sounds like a random mash-up of music slices with no overall sense of form or mood, and at worst it sounds like musical nonsense.
It could also be because successful systems become so complicated they're only really useful to one composer. No one else has the time or the patience to learn them.
But more, musicians seem to have a block about the idea - e.g. WaveDNA have literally had a lot of hate for LiquidMusic.
Even though many genres are very formulaic, abstracting those formulas in a useable way turns out to be very hard, especially if you're aiming for a definite mood instead of a random musical squiggle. Most people who make music would rather do it by hand than have a computer do it for them.
Nothing super specific, I know people are doing raves with that closure music suite where they modify rhythm and music as the night progresses. I am not sure how much this can be considered as ES, kind of is...
Quick google found this, looks like what I was referring...
http://algorave.com/
This is what I was referring to
http://overtone.github.io/
Quick google found this, looks like what I was referring...
http://algorave.com/
This is what I was referring to
http://overtone.github.io/
Funny that today's TechCrunch Disrupt winner is an AI service for music generation http://techcrunch.com/2015/12/08/jukedeck-wins-disrupt-londo...
Music and cognitive enhancement are both cool. But I can't imagine "DJ in my head" as a key driver. Automation of learning stuff seems far more useful.
Music Schmusic. I'm waiting for the advertisements.
I'm reminded of the rule about headlines with a question mark at the end: the answer is always "No".
As an aside - this feels like a "submarine" post (http://paulgraham.com/submarine.html). I can't tell what they are selling though.