Apologies for the clumsy title, but I’m not really sure how to sum this up in a snappy way, and I know “AI” isn’t actually the proper term (but is fast becoming the commonly accepted buzzword for this type of thing).
What I’m talking about is the possibilities that technology is now giving us to unearth old music and make it listenable for us. Of course it was the Beatles single that got me thinking on this course, but I’m not really wanting to start another Beatles thread.
The thing is that, whatever you think of the artistic and cultural merits of the song Now and Then, the technological aspects of it are astonishing. This was a hissy, unlistenable, home demo quality tape from over 40 years ago, with Lennon’s voice buried behind a clumsily played and intrusive piano. And yet, like archaeologists unearthing a tomb, the boffins managed to extract an isolated stem of Lennon’s voice in startling clarity.
When you think about it the possibilities are endless. There are so many bits of audio ephemera out there for all our favourite bands that you can start to see this is the tip of a very large iceberg. I’m particularly taken with the thought of those hours and hours of early Pink Floyd concert recordings, mainly recorded by keen audience members under less than perfect conditions, buried in hiss and distortion and mostly unlistenable. But now the potential is there to extract and clean up all the separate audio elements. A pristine, surround sound mix of 1969 concert staples The Journey and The Man? Yes please.
And you can apply that to any artist of your choice. There’s bound to be one which piques your interest. Early Kate Bush demos? Miles Davis’ Birth of the Cool with same presence and clarity as his 60s albums? Robert Johnson sounding like he is playing in the room in front of you instead of via a scratchy, tinny gramophone?
There’s even (believe it or not) an audio recording of the Quarrymen playing at that church fete in 1957. It’s barely audible, muffled beyond belief, and in the background of lots of crowd noise – but it’s there and potentially we are now able to extract the performance and get it to sound as good as the Get Back sessions.
On the one hand, I’m desperate for this kind of stuff.
But on the other hand, I’m wary. I think the power of music, pop music in particular, often comes from what is left out rather than what we let in. It exerts a pull on you because you are constantly craving more. Like many, many people, the Beatles were the band I became obsessed with at a formative age. And as I gradually found out the existence of bootlegs and outtakes, and then we had the Anthology albums, and now an ever-growing body of remixes and other goodies, I pretty much gobbled up as much as I could afford. But it’s the kind of stuff that, to be fair, I only listen to once for the novelty. I tried listening to Anthology 3 yesterday, for the first time in many years, and found myself pressing the skip button a LOT. Bu whenever I go back to the original albums (as I did with the White Album last night, which started this train of thought) I’m thrilled by the focus and the immediacy, the condensed nature of it all, stripping songs right down to their essential elements. (The White Album has a reputation of being a lazy, unfocused, messy thing, but the constant, perfect chain of two minute thrills on display belies that view. Even Revolution #9 is a titillating joy, and over long before you tire of it.)
What’s my point? I suppose it’s that as we get MORE and MORE we’ll start to lose appreciation of that slim vein of the Really Good Stuff. And we might actually forget stuff like the White Album, in its original condensed glory, exists. And that maybe less is more and it’s dangerous to tap into a potentially unlimited well of music.
I suppose it’s a similar worry to the chronic indecision that music streaming services create. I’ve more or less stopped using music streaming services now: all the music in the world (well, kind of) and I could never find a thing I wanted to listen to.
I don’t know. I’m just suddenly pessimistic about it all and think it might be a slippery slope.
If you think that’s confusing, wait until the software starts producing entirely new tracks by your favourite dead artists which sound as good as or in some cases better than them.
We moved in the space of a human lifetime from an age of music scarcity to an era of abundance. Ahead lies the glut.
I think we’re alrady in glut conditions with some acts.
Obviously inferior demo versions of songs? Multiple live recordings of bands who play their songs exactly the same every time. What’s the point of that except to squeeze more money out of their audiences?
Multiple duplicate live versions – much of the joy of a live version is it is different. Bum notes, missed cues, repeated verses etc. AI-ing them would render the live version obsolete
Yeah very good point. With live music especially, there’s an ethical question about how far you take the airbrushing process.
Thinking about it, I think I’m generally okay with a bit of smoothing and mixing to make it a decent home listening experience. And to be honest, glaring bum notes I’m fine with patching over, just… because.
But I draw the line somewhere long before fixing them into pristine, note perfect performances that sound just like the record.
It’s Pandora’s Box, innit? I think if you asked me to push a button to generate hours and hours of new, AI-generated Beatles songs, my curiosity would get the better of me. But I just KNOW (or I hope I know) it would muddy my relationship with the original albums, and it would be a bad thing to have in the world.
But yeah, I agree, it’s probably coming, and very soon.
God knows how it will feel when the technology inevitably arrives. The shift from physical media to streaming was definitely a wrench, although I’d say the upsides have outweighed the down, from a listener’s perspective at least. As ever, change was the difficult part.
The change I’ve observed in myself over the years is that as my consumption patterns have shifted, my locus of attention has done likewise. Whereas I used to be fascinated by the back story to various bands and musicians – the mythologizing of their genius – nowadays I care far less. My relationship is with the various noises they made, and my interest is in how those noises make me feel.
I’d expect AI music to be another step in that direction. I’m sure some fabulous music will be produced, but watching software churn this stuff out will probably be a hammer blow to our lingering notions of a unique human genius capable of such acts of creation. There will be no Mojo cover stories for your new favourite album, because it’ll have been written and performed somewhere off in a colossal server farm, well out of sight.
How you feel about the above development probably depends in large part on whether you invest first and foremost in the artist or the art (and in what combination). As has been discussed at length in the last fortnight; is a song better because it is by the Beatles? Does the weight of their history contribute to your enjoyment of/elevate said song? Would you enjoy it just as much if it had been released by (ooh, let’s say) Scouting For Girls?
In some ways it’s the corollary of the issue that arises when our favoured artists do horrible things in real life – does it impact our enjoyment of their art? Should it? It’s separating art from artist all over again.
Interesting reaction from Rick Beato.
I saw that video come up but haven’t clicked it as I don’t really like Rick Beato (he waffles and I don’t like his smug smile…). But if it’s relevant to this, I might give it a watch!
I’ve rather tired of Rick Beato. I liked his “What makes this song great” videos, but as his audience has grown, he’s believed he is the authoritative voice on all things musical. His dismissive attitude to current Spotify chart music is annoying (“there’s nothing new here” – Yes, but you’re 61 not 11. It’s not aimed at you)
while he praises Now & Then for having 3 minor chords in a row (Bm,Em,Am) which is a pretty standard chord progression (and the demo was written over 40 years ago)
Yeah, I think I agree with all you say here. I got tired of Rick Beato very quickly, after hearing how great he was! And for much of the reasons you mention.
Although, having said above that he waffles/rambles, I have to concede that in this video he succinctly puts into words what I was trying to say about the Beatles! They are “a band who knew when to stop” (or words to that effect). That’s exactly it. Their records are masterpieces of saying just enough and no more, and doing it in 14 songs under 40 minutes.
Oh I like Rick. He’s really enthusiastic and has loads of knowledge which he shares without irritating ads or product placement. His opinon on current charts is no more or less valid that anyone else’s as I’m sure he’d agree. He’s right on the minor chords thing too. Not many songs do use 3 minor chords as far as I can remember.
I only came across him a few months ago, when someone recommended his interview with Keith Jarrett. I can see why musicians like to be interviewed by Beato, he shows both knowledge and emapthy. This interview is also quite moving – to see Jarrett now unable to play with his left had due to a stroke, but still able to improvise a beautiful off the cuff version of Desfinado which sounds like he is playing with both hands is in a way uplifting.
I also liked the video called something like college professor to deli server, where he tells of how he decided to take the risk of becoming a full-time musician in his thirties, and it all went wrong, is both honest and funny.
A pedant writes: it is MAL, not AI. Nothing new created, just an assembly separated into its component parts.
What Bingo describes above is AI. New (artificial) content created.
You know, I actually considered name checking you in my post, because when I was typing it I could sense you would be chipping in to say this! Yeah, I totally agree. I think on the whole I’m a Lapsed Pedant though (where do I hand in my card?) – I used to be picky about stuff like this, but now I just think, if the crowd goes with something, why try and swim against it? “AI”? Fair enough. “CGI”? Knock yourself out. “Vinyls”? Why not. It all feels like nails on a blackboard, but life must go on!
“It all feels like nails on a blackboard, but life must go on!”
Afterword t-shirt.
You will probably know this Fenton. Is there a AI plugin to “unclip” clipped parts?
**returns to horribly clipped bass part and tries to make it sound OK**
I’m not sure it is AI, but there’s a plugin available for (I think) Adobe Audition called something like Declipper. A pal uses it to recover brickwalled CD rips.
Death Magnetic is probably beyond redemption, however.
I’m on my daily walk and just pondering whether a lo pass filter would do it. Filter off the graunch?
Nails on A Blackboard. TMFTL
Presumably there was some quality control undertaken in crafting ‘Now and Then’? I think it’s quite a sweet song, though it sounds and feels very similar to Real Love, and there was probably a similar atmosphere and mood around its (re)creation.
Just because something can be produced, doesn’t mean it should or will be. As it becomes easier and easier to record and release all the scrapings off the studio floor, the long tail rule of diminishing returns will surely apply. There is an audience for everything, but not necessarily for everyone.
The Pokemon Principle is only for the obsessives and for the extreme AWers.
I like the idea of being able to hear and see old recordings in the best quality possible, while recognizing they are modern exercises in nostalgia. As a mash-up fan, the real creativity for me is in how they are mixed and drawn on to create new audio/visual material.
The industrialization of music production will result in machine-tuned product that ticks audience-tested boxes, but I suspect, like all mass-produced stuff, will miss something tiny and almost imperceptible that emerged from generations of musical evolution and has never been labelled – the equivalent of micronutrients, or surface hiss on a record.
Great topic, by the way. Thanks for starting it, Arthur!
I am sure a LLM-type music generator exists – or is in the process of being made, whereby, having been fed with all the, say, all the Steely Dan and Donald Fagen/ Walter becker albums, could generate a Dan-sounding song about whatever it has been instructed to. (Some might say they are called “Monkey House”, who are great). It’ll be OK, sometimes ostensibly great, but not QUITE on it. It’ll get better, though we will still prefer human-made music. Some musics will be/ are easier to generate and sound more convincing than others, as most music is adequate for what it is, just like most art or graphic art is. Frankly, i’d rather a proper Beatles-bot make a new track than the piss-poor scrapings that was the recent release. What would be wrong about having a Stevie Wonder-bot generate music based on his 60s and 70s output with Les McCann and Oscar Peterson influences, thus sidestepping the years of stagnation?
I think we’ve wandered away from my original question into sci-fi territory about AI recreations of dead bands, but that’s fine! It’s all part of the same issue.
Some interesting comments to ponder above: ” What would be wrong about having a Stevie Wonder-bot generate music…?” (Vincent), “Is a song better because it is by the Beatles? Does the weight of their history contribute to your enjoyment of/elevate said song?” (Bingo), “The industrialization of music production …. will miss something tiny and almost imperceptible that emerged from generations of musical evolution and has never been labelled – the equivalent of micronutrients, or surface hiss on a record.” (Salwarpe)
Salwarpe, I think you’ve articulated it well actually. I think this is the fear I have. We can be rationale and scientific about it, and conclude that if our ears can’t tell the difference between an AI song and a real song, then why should we care? But that “tiny and almost imperceptible” loss is the thing I think that will gnaw away at us.
It’s the same uncomfortableness I get when you see these sci-fi movies where someone is on a spaceship billions of miles from earth, and they hit a button to start a VR simulator to make them feel they are in a forest or by the sea. I feel like in that situation my brain would just reject it and I would crave earth even more, even if I couldn’t quite work out why.
Or to take a more earthly example, I do have sympathy for these people like Neil Young who decry the move to streaming and digital music. Full disclosure, I’m sure in a blind test I couldn’t tell the difference between an analogue music source and a high bitrate MP3… but (godamnit) because I know there MUST be a difference it would eat away at me and I would feel tricked.
But but but. Maybe the issue is KNOWING that the difference is there, even if you can’t put your finger on what it is? Like all technology, as the Younger Generation ™ come through, they’ll care less and less about this, won’t they? In a decade or so, will it just be our lot (backward-looking music fetishists!) who care about whether a song has been “written” by an “actual artist” or created by an algorithm? For me, some of the doomsday fears by the old guard now seem laughable non-issues to me (such as: CDs replacing LPs, musicians not playing live on TV, sequencers, synthesisers, sampling, etc etc) so I do think this kind of debate will be seen as antiquated eventually.
I’ll check back on this thread in a decade to see if I was right!
I think Sal’s post is ultimately an appeal to a form of authenticity, even where that authenticity is essentially imperceptible.
I’m less convinced. There were people who decried the advent of recorded music as no substitute for live performance, and we got over that hump pretty quickly. Human beings tend to wring their hands over new technology for a short period and then adapt themselves to it, I don’t see why this will be any different – particularly, as you say, amongst the young.
The last couple of decades have demonstrated, again and again, that if you give people what they fundamentally want at a comparatively low (or even zero) cost they very quickly move on from wondering about how it was produced, or whether it’s sufficiently authentic. People are quite happy to eat sausages without ever seeing how they’re made. Whether they should wonder more about those sausages is another question entirely.
Well – penultimately, (as the thanks to Arthur is the ultimate paragraph) and kinda sorta, in that authenticity is appealing – something we like to hang onto in times of change, but is also a marker of slow change as opposed to rapid transition –
I think authenticity will emerge as the veneer of human culture smeared all over new toy as we paw it with our greasy mitts. There will probably come a time when AI feels very authentic as we explore each nook and cranny of what it can do for us – all the unintended consequences of shed-based tinkers discovering aspects that were far removed from the algorithms of the original founders.
Though ultimately I will take ‘ultimately’ – as I rambled all over the place before getting to what might be taken as the point. It is ever so with me.
Once you realise authenticity is debateable then the less you care about it. The Long And Winding Road was messed about with against the author’s wishes and a tiny minority think it a shame and know the story but the vast majority accept and take it as it is and believe The Beatles made it that way.
As some charlatans once said: nothing is real.
Monitoring what the Younger Generation(s) are doing, and doing the opposite, seems like the best bet, and has done for a significant amount of time now! I was certainly doing it forty years ago.
‘Now and Then’, much like my purchasing of a Dusty comp. for £1 on Saturday, simply drags me back to the original recordings of the 50s and 60s, many of which (on Kent compilations, for example) I have never heard.
With mass culture, I think there is always a conflict between automatization and authenticity. At this point I would post a clip of Michael Palin turning the laughter dial on the desk in some old Monty Python sketch – but I cannot find it. Stimulus/response. Canned laughter and the audience at home will join in. When you see the man behind the curtain, the Oz illusion disappears. But until then, we are utilitarian pigs, happy in what we believe is true. It’s suddenly got a bit Matrixy here.
I think until machines come preloaded with oxytocin, we will always prefer the human touch, when available. Neuroscience and brain chemicals are a fascinating area of research, but the more that is understood about the brain parts, in particular the hypothalmus, the hippocampus, the amygdalla, the more easy it will be to affect and stimulate experiences and memories in ways that seem authentic.
I also think it’s important to get out in the fresh air and go for a good walk. That way I realize the world is a big place, with fantastic natural forces that are far more powerful that are dreamt of in our philosophies. Hooray! Show!
MAL or similar’s ability to isolate separate voices or instruments on a master recording would presumably allow multitrack versions of recordings to be made which now only exist as two or four track? XTC need that for English Settlement, as the multitracks are still AWOL.
It does indeed. Sounds like a thrilling prospect, doesn’t it?
There are certainly many, many albums where I wish the original isolated master tracks were still available for a spruce up. To take one example, Cream’s studio albums always sounded pretty poor, and would be a great candidate for this treatment. Yes’s Going For The One album and Mike Oldfield’s Incantations as well: both crying out for proper surround remixes, but sadly impossible without this technology.
The downside being: I bet Peter Jackson’s millions of lines of code and server farm won’t come cheap. Those XTC BD/CD sets won’t be 20 quid any more.
Oh, true. But it will presumably become cheaper, and more accessible, by the time they’re up to MAL-9000.
And the minute English Settlement’s been MALed, the tapes will turn up.
The use of new technology in any field has an affect, and music is a prime example. The introduction of recordings, and then the introduction of electrical amplification and so on had a huge impact on the way music was listened to and enjoyed.
I have absolutely no problem with using this ‘ere MAL thingy to manipulate poor recordings into listenable music. The various live Beatles recordings, particularly the Star Club tapes, would be fascinating. As Giles has said, it’s only a development of existing demixing tech, which he used to great affect on Love to create a fabulous new work from broken down tracks. But…they did add some new bits to that in terms of orchestration…was that wrong? Or was it ok because humans did it?
As for using AI to create ersatz work by artists, I really don’t see the point of it besides being a moderately interesting excercise.
Thought experiment for the blog.
In five years time, whoever controls the Beatles estate announces that they’ve stumbled upon one final great lost recording. Miraculously, it’s the entire band playing a previously unheard song, circa whenever their peak was. The recording is in great condition and needed minimal restoration. Word of mouth around the quality of the song is sensational.
When it’s eventually released, alongside an appropriately tearjerking and tastefully manipulative video, you’re blown away. The song is masterful, full of what made the band great and more besides. It moves you profoundly, sending you right back to your youth. After a few weeks of listening you conclude that it stands among the very best of their work.
A year later it’s revealed that the entire song was fabricated from nothing by cutting edge AI.
Do you stop listening to the song, stop enjoying the song or carry on as before?
Does your answer change if instead of a Beatles track from the archives it transpires that your favourite contemporary artist’s brilliant new album (yes, I know that will break the immersion here for some, but let’s imagine you have a favourite contemporary artist) has similarly never known the touch of a human?
Some audiophiles over on the St*v* H*ffm*n forum have sort of lived through this scenario when they discovered their previously loved AAA vinyl was sourced from the digital files used to make the SACD versions.
Some analogue purists cried foul. I just think they sound great, whatever.
Argh don’t do this to us.
To be truthful, we probably shouldn’t care, I suppose. But I think once I knew it was fake, I would go off it. And I wouldn’t be able to help myself.
I’ve actually had a similar experience recently. The Mike Oldfield fan community know about some unreleased recordings he did in the late 70s which (apart from a couple of notable exceptions) only exist in very ropey quality as background music in a documentary film. (‘Reflection’, if you are at all interested). They are a bit of a holy grail. Well, last year or so, someone suddenly released on YouTube, without fanfare, what appeared to be pristine stereo copies of these recordings.
When I first heard them, I was thrilled. It was astonished to hear these in their full glory, and was so glad someone had finally obtained copies.
But no. Turns out someone had done a (quite brilliant, and almost note perfect) copy in their own home studio and playing their own instruments, and that’s all it was. (I don’t think it was an intentional scam: the guy just didn’t think about it and didn’t label it properly when he shared it).
Well, after that I’ve barely listened to it. Now I know in my head it’s not the real deal, it just doesn’t hit me in the same way. And in fact all the glitches and not-quite-right bits stick out.
I don’t know what my conclusion is, but in answer to your thought experiment, I think I would have to say I need to know something is authentic.
I think Authenticity is overrated, but on the other hand I think Outright Fakery is to be strongly deprecated.
If you can’t come up with something that is recognisably your own, then you’re not really creating.
That’s interesting, Arthur.
To be clear, I don’t think there is a correct or incorrect answer to the above question, but I do think it has the potential to reveal something about the different ways in which we each consume music.
This gets us into the realms of “real or fake” objects. If the watch you inherited from your father, turns out to be a cheap copy, does it mean less to you?
I like these thought experiments! I don’t think that would mean less to me, no, because I think in that instance the value of the watch would come from a sentimental value rather than a financial value. Wouldn’t it?
I’m looking forward to seeing this stuff pushed to the extremes. On the one hand the Beatles Star Club tapes cleaned up would be fab to hear. Also maybe there’s music and historic speeches from the 19th and early 20th century might be another area for this technology.
On the other hand, what horrors are going to be perpetrated? Stuff along the lines of Kenny G duetting with Louis Armstrong. The industry will do anything for a dime and I can’t wait to hear what they’ll do. What other delights await us???
If you think I won’t one day be in here posting nothing but the songs of the Rolling Stones with vocals performed by Gilbert Gottfried you are sorely mistaken.
Some would say that we’ve already got that with Hackney Diamonds.
😂
The danger with applying AI to historical political speeches etc. is that history could be changed. Or at least people’s perception of it.
The UK TV series “The Capture” illustrated an apparent ability to put words they didn’t say into people’s mouths on live TV, actions they didn’t do on live CCTV footage. Even easier with historical audio/video.
https://en.wikipedia.org/wiki/The_Capture_(TV_series)
Whilst not wanting to post spoilers, something similar happens in the latest MI fillum.
Here’s one – generating dialogue for silent films?
I am not in favour of this. The genie is out of the bottle and who knows where this will lead.
I can see all sorts of tripe being resurrected by the corporations who see a further opportunity to rinse their gullible audience.
Now and then wouldn’t even feature in the top 100 Beatles songs but is being praised in the same way as the second coming would be. I think people should get a grip on reality.
Beatles fans like buying Beatles stuff, even if they already have it. It makes them happy, so live and let live I say.
Bad enough that old silent movies are being colourised these days, let alone having dialogue/sounds added too.
I’m sure there are areas where AI will be used extensively in music (and is starting to be already). This is for more ‘production line’ pieces like some pop and catalogue/soundtrack music that needs to be done quick and cheaply.
Will be interesting to see what they can do with MAL. If it means some of the masters that have gone missing or have been destroyed, such as in the Universal fire can be worked on that may be a good thing.
No doubt it will be another thing to be charged against future royalties by The Man.
Get yourselves ready for Robert Johnson in HD stereo and Louis Armstrong’s Hot Five in surround sound, because if it can be done then someone is bound to do it.
Some years ago, there was a sound engineer, Australian I think, who had a series on Radio 3 playing “stereo” versions of bands like Duke Ellington from the thirties.
He had discovered that when recording was to disk, it was common practice to set up two devices in case one failed. As they would be set up in different parts of the studio, they could be combined to space out the sound in a kind of stereo effect. He also cleaned up the recordings in other ways, and at least to the casual listener, they did sound as though they had originally been recorded to tape as stereo.
Back in the ’30s the recordings would have been cut direct to disc. Tape recording was very much a post-war thing.
Bing Crosby was given the first two tape recorders to arrive in the USA, brought over from Germany. Bing gave one of them to his pal Les Paul, because he knew that sort of techy stuff interested him.