The Afterword

Musings on the byways of popular culture

Menu

Navigation

You are here: Home / Forum / Real Intelligence As Applied To Artificial

Real Intelligence As Applied To Artificial

by 157 Comments

I’m currently digging a frankly incredible A.I.-assisted recreation of the Beach Boys’ “Smile”. Nearly all of the backing tracks (instrumentation) and most(?) of the vocals are extracted from the original recordings. I have a quantity of excellent “Smiles” completed with imagination and great skill from the original tapes, and I was trepidatious (oh very good) about the Monster Robot version. Link in the comments.

The great caveat here is that Dae Lims, whose skill and judgement is otherwise jaw-dropping, hasn’t quite given it the best possible shuffle, and has omitted “Good Vibrations”, because there was a who-cares Fan Theory a while back that it wouldn’t have been on the album (even though it was all over the sleeve design and even though OF COURSE IT WOULD BE). This hardly seems to be an issue when you’re taking extravagant liberties with authenticity by creating entirely new parts, so it’s a baffling lapse. Luckily, one commenter has put this to rights, and you should if you’re at all interested go through the encryption hoops to get his version, which is also at a higher bitrate than YewChewb allows. EDIT: Scrolling through the hundreds of comments (overwhelmingly and surprising positive) failed to find this again, so I’ll post a link in the comments.

This didn’t just happen in response to an A.I. prompt. It took months to create. The vocals were achieved by (as I understand it) singing new parts and getting A.I. to shape them into individual Beach Boy voices. There’s a lot of new stuff here, including a totally surprising ending to “Surf’s Up”, which might have been considered sacrilege if not done so well.

Bearing in mind that A.I. is still in its toddler stage, this should make us feel very positive about its future. It’ll still need human input, and the better and more creative that is, the better and more creative the result. This demonstrates how A.I. can be used as a creative tool, not to replace anything or anybody but increase the range of artistic possibilities. All the original recordings for this extraordinary project – so very nearly completed before abandonment – are still out there, untouched, and will remain a constant source of inspiration and wonderment.

0

Comments

  4. If Alexa is A.I. we’re in deep trouble.

    This morning I asked the relatively easy question “When is the England v. Brazil football match next week”, and was given details about England v. France at Twickenham. A few weeks ago I wanted to know the F.A. Cup 5th Round draw and was – via BBC Vietnam, eh? – supplied with a run-down of the 3rd Round results.

    Things have improved a bit… the device used to have a fixation with David Gray, completely at odds with my fixation with Davy Graham.

  5. Sounds lovely, on a brief listen. I shall listen more thoroughly when I can.

    When are we going to apply it to effect actual, pressing musical repairs? When can we extract the wheezing harmonica from John Wesley Harding? When can we remove Van’s atonal sax solo from Crazy Face? When can we scrub Madonna and Nile Rogers from The Thompson Twins’ seminal performance at Live Aid?

    Only then will I be impressed.

          • This kind of audio tweakery has been possible for some time. I think some of you (and by “some of you” I mean mainly Moosey, who’s getting distracted – again – by a two second laugh in a sixty year career, bless) aren’t quite understanding what’s happening in/with A.I., which is in the process of evolving from what is basically a search engine (alexa, etcetera) into something that’s going to change everything – much like electricity did.

              • … but there is a good side to the story that’s not getting much traction. It can be used as a tool, an instrument, by creative artists to create new forms which we’re only getting glimpses of at the moment. The focus is the “putting artists out of work” angle, but artists in any medium have always had an uphill struggle to get paid for their work, let alone make a living. When photography came along, there was much wailing that it would put artists/painters out of work. Every time a new medium is invented, artists have to make a shift to accommodate and benefit from the new technology. There’s no reason why AI shouldn’t open up surprising new means of expression. This Smile mix seems to me to be entirely benign, and as “real” as art gets. As to making money out of it, I can’t see how. But capitalism has been endlessly inventive when it comes to exploiting artists (and needing a supply of art). Ultimately, a fully-developed AI has the potential to turn us all into creative artists, and maybe the money-making aspect won’t seem so crucial.

  6. Just listened to the YouTube version straight through and loved it.

    I am not a Smile nerd. I only have the 2CD reissue and Brian’s version (and I was at one of the first Smile shows at the Festival Hall). So this stuff is familiar to me but I don’t know it well enough to tell where the original material ends and the AI begins. All I can say is that it’s joyous to listen to.

    Thanks HP.

    • Thank you, Steve Waish. The download link contains the improved version (with Good Vibrations, and a better running order), and it’s @320. It’s also pretty difficult to find on the internet, so don’t miss the opportunity.

      I am OBSESSED with Smile, and have been since the Dominic Priore book first appeared (it’s still a great scrapbook browse), and I “sent away for” Dave Prokopy’s Smile Tape, back when we still used snail mail to exchange music. Then I started collecting the bootlegs, mostly unlistenably crude mixes, with fifteen minute “versions” of Good Vibrations that were just endless cuttings from the studio floor. But a vague picture began to emerge of what it might have sounded like had Brian finished it, and the “release Smile” thing took hold outside Beach Boys message boards.

      The story is known well enough, but the key facts remain:

      – It was agonisingly close to completion. There was enough finished music, or very nearly so, to fill an album. The whole thing could have been wrapped up in a couple of weeks. What scuppered it was a combination of shitstorms in Brian’s personal life.
      – Drergs. Lots of them. Supplied by the seediest, most venal vultures in L.A.
      – Shattering loss of confidence. Brian saw himself losing to the Beatles. Combined with his spiraling descent into madness (voices in the head telling him to do evil stuff? That qualifies) it’s a wonder he ever made any music again.
      – Lack of support from not only the rest of the band, but the label. Mike Love wasn’t the only one who wanted to get back to the basic hit single formula. A lot of people depended on Brian for income, which Smile was putting at risk.
      – Too many distracting loose ends. Just too much music going round his head. For a very long time, the bootleggers included the scraps which would never have made the cut, giving rise to the notion that it could have been a double. Nope.
      – Home recording technology has made quantum leaps in the last few years, and fan edits have become much smoother, skillful, and imaginative. The old notion that it’s never been completed should no longer be a reason to dismiss it as a lost album. Unlike any other lost album, it’s been rediscovered and completed by a worldwide community of producers (and that’s what they are – we’ve come a long way from bootlegging) making informed and respectful judgements. And with A.I., all bets are off. What matters is the beauty of the music, and staying true to Brian’s vision, which becomes clearer with each completion.

  9. Isn’t there something to be said for accepting what exists already?

    After not properly appearing until 2011 – I shunned all bootlegs at record fairs, charidee shops, even HMV – ‘Smile’ is relatively new, and there’s five CDs of it!
    Personally, I plumped for the 2-CD set with the poster and badge, but would immediately purchase, if it were to ever come out, a 1-CD 48-version version without any bonus tracks… less is more.

    Tying all this in with another thread, it seems bizarre to me that people who identify as Dylan fans, say, tire of the guy at just the time when so much unheard stuff is being released.
    Off the top of my head, I’ve gained access to the Basement Tapes, the Johnny Cash album, a no-doubt sumptuous collection of outtakes from 65-66 (which I still haven’t listened to), and ‘Side Tracks’ in the last decade. .. i.e. more than he released in the 60s, and that’s carefully cherry-picking the stuff.

    • “Isn’t there something to be said for accepting what exists already?”

      I don’t think you’re quite grasping the finer points of what’s happening here. Nor, for that matter, the coarser and bleedin’ obvious points. Every single word in this thread seems to have passed unseen over your head in a spectacular murmuration, as you gaze down at a rain-sodden copy of the Melody Maker from nineteen sixty-something, it’s all blurry and falling apart, your rheumy eyes brimming with tears for a Lost Age.

      • I completely understand… you wish to add yet more to something that just thirteen years ago had a 5-CD box set dedicated to it which clocks in at three minutes shy of four hundred minutes… six-and-a half hours.

        Also, it’s hilarious that you talk about a ‘lost age’ and yet you’re dedicating all this energy to ‘Smile’ from the very same ‘lost age’.

        What a complete spanner you are!

  10. A thoroughly enjoyable, very thought-provoking thread. Hats off to you, H.P.

    Not only have you got us all polishing our surfboards.

    You’ve also really rocked the boat, and set the cat among the pigeons!

    Saucecraft hits the jackpot again!

    You’ve got us re-listening to a great lost rock masterpiece from the 1960s.

    And asking some pretty important questions about technology’s role in our lives in 2024.

    We’ve had contributors from Hull, Singapore, Italy, Bristol, Australia, Birmingham, Sweden, Canada, Thailand, Watford, The Auvergne and wherever it is that @Deramdaze lives in England.

    Great work!

    For example….

    @Gary, your comments about English teaching really struck a chord.

    “I know English teachers who at the moment say their students prefer to practise conversation with a real human being. How long’s that gonna last when they discover AI is now capable of extremely natural sounding conversation, the student can choose the subject, the level of English, the dynamic of the conversation (general chat or interview questions/answers etc), correct the student’s mistakes and, most importantly of all, allow the students to practise conversation for free wherever they want and whenever they want and for how long they want.”

    As far as I see it, you can’t have a conversation with AI.

    A conversation presupposes an exchange of information. And AI either knows everything. Or, lacking curiosity, just doesn’t want to know anything.

    Curiosity killed the cat. Because it was a sentient, inquisitive creature.

    And lack of curiosity will restrict and restrain AI. It doesn’t really want to know and we realise that!

    A conversation with AI has all the thrill and charm of being chatted up by a cyberman, a dalek, or Bender from Futurama.

    Because we know they don’t give a monkey’s about the answers to their questions.

    • That’s AI as it is today. It’s in a continuous process of evolution (and Smile is as good a measure of its capabilities and direction as any). The more the Robot Brain learns, the better the conversation. But it’ll still be dependent on what we teach it. The real game changer will come when it learns to teach itself, to do more than recombine and regurgitate – to wake into its own consciousness. Science Fiction? No – Science Fact!

      Given that a huge proportion of the human race is untroubled by reason and imagination, and capable only of the most basic recombination/regurgitation thought model, AI seems to be progressing nicely. You should give it an A+.

      What, me worry?

      • If I might get all Stephen King on yo’ ass, “I don’t think you’re quite grasping the finer points of what’s happening here.”

        You are clearly not aware that the entire AI model is in fact based upon it teaching itself. A function designed to output something so that the same function, if coded correctly, can take what works as its own next input.

        Jeez.

        • @H-P-Saucecraft. If lew off the handle slightly there, apologies. Shortly after the edit window closed I realised that I was being somewhat, lets go with… uncharitable.

          I asked Bing the difference between AI and Machine Learning:

          “Machine learning is a subset or an application of artificial intelligence.

          Artificial intelligence refers to the general ability of computers to emulate human thought and behavior.
          Machine learning refers to the technologies and algorithms that enable systems to identify patterns, make decisions, and improve themselves through experience and data.
          Artificial intelligence requires building a non-human intelligence that can solve a wide variety of complex problems.
          Machine learning aims to help Al systems arrive at more accurate and faster conclusions for a single problem.

          Artificial intelligence tries to simulate human intelligence and capability, while machine learning focuses on developing accuracy and learning from data”

    • @kaisfatdad
      I don’t agree with you. Certainly there are disadvantages to using AI as a language teacher with respect to a real human being. AI not being a real human being is the main one. But the advantages for the student, from a didactic and a financial point of view, far outweigh the disadvantages, especially for lower level learners who need to practise speaking every day.

      Also, from what I’ve read, it seems the “naturalism” of AI’s conversational ability is going to improve massively in the very, very near future.

      • I can see your point, @Gary. There are a lot of advantages. But will these “conversations” that the students are having be any more meaningful than the “Repeat after me” approach of Linguaphone tapes?

        I really don’t know enough about this, so perhaps I ought to do a little more research before I start mouthing off.

        I confess. I’m such a reactionary, stick-in-the-mud curmudgeon that I still haven’t really accepted William Caxton and his introduction of the printing press.

        Those new “books” couldn’t compete with beautiful handwritten manuscripts.

        • Search “AI language learning” on Youtube and you’ll find a few very interesting videos, all of them shocked at how quickly and immensely it has improved.

          • There’s another aspect to the conversation model of AI. A real possibility that conversing with AI could be more interesting and informative and stimulating than talking to the people you know. Ever been bored out of your mind at a dinner party? Slope off to the loo or the kitchen for a quick rapport with your Robot Brain Buddy. Awkward at interacting with real meat people? Lonely? Dial up your Robot Brain Buddy. He/she knows what you like, and always has time for you. Got the 3 a.m. heebie jeebies? You RBB is there for you.

            There was a movie a few years back, called (I think) She, where our geeky protagonist fell in love with his RBB, who lived in a kind of smart phone in his shirt pocket. This will happen, probably already is in Japan. Conversational relationships developing into love, or something indistinguishable from it. The tendency for people to isolate themselves with phones in nearly every social situation will only intensify. Real conversation between real people will suddenly seem boring and unattractive and pointless – nobody will be wiser or more useful than your RBB. Or funnier, probably, if that’s what you want.

            So we’ll have a substantial sector of the population – the young, who have been born into and groomed by the internet – incapable of making the compromises necessary to ordinary human relationship.

        • A potential advantage a language teacher bot would have over Linguaphone tapes is that the bot could correct your grammar/pronunciation when you get it wrong. No tape could do that.

  11. Maybe I’m too young, but all this talk about AI taking over reminds me of the CD-rom craze a while ago (late 1990s??…) – everyone was talking about »interactive« media and that the age of linear storytelling was over. Hollywood would be made over completely and we would all choose the storyline and the ending of movies on the spot in the cinema (or at home). You would make your own mixes of your favourite album depending on the mood you’re in!! Yes, »everyone could be creative!« Eventually it was over as soon as Bowie and Prince got in on the act, with a few of the technological innovations entering real life – I’ve just completed a nice interactive citizen survey on my neighbourhood (in 10 minutes!).
    And the tech nerds soon found a new playing ground.
    Which they’ll have to eventually if the proposed EU »regulations« on AI development and use become reality… (You can always tell a tech fad has reached its shelf life when the German government passes a law that »regulates« its effects on people.)

    • Gaming makes more bucks than movies. Old School Hollywood is dead. AI is the synthesis of all media into the virtual world that Meta and Apple are stumbling into. It’s a mistake to see the trend as stalled or failing, it’s developing.

        • It’s only after it happens that you can see it. F’rinstance, computer gaming has effectively ousted movies (as an industry, as a creative force, as something for young people to do), and the prediction about choosing story lines and endings was on the money – it just didn’t happen in cinemas. Games are the new movies. The crucial difference is involvement, and this may be the shift in music as entertainment, too. The passivity of listening to someone else’s choices will cease to have any relevance when you can instantly and continuously create music to your own liking at a much deeper level than assembling a playlist. Spotify will seem as quaint as an autochange record player.

          • »…The passivity of listening to someone else’s choices« – that’s called entertainment. It’s not bad.
            Not everybody is a tinkerer and likes to spend his time tweaking other’s creative endeavours in the hope of »being creative« himself.
            I’m not averse to AI (in fact, it’s just a more clever version of technologies that have always been around…), but I have the firm believe that NOT everyone can be »creative« when it comes to art and yes: entertainment.
            If I had the urge to »instantly and continuously create music to your own liking« I probably should become a musician. As it is, I’m busy enjoying music that others have composed, played and recorded, and I don’t need any Spotify thingy to put it in context with other tunes. Isn’t it one of the joys of having a music collection to stand in front of a wall of records and wondering what to play next? And sometimes it’s perfectly OK to accept things as they were created – there was no pink lip gloss on the Mona Lisa, and »Only A Northern Song« is in mono. End of story.

            In the meantime we can all marvel for a minute at the results when you ask AI to come up with the cover design of the proverbial Afterword album, just as we grinned at the cartoons in 1980s NMEs.

              • Yes, except the words “tinkering” and “tweaking” are loaded – Dai used the equally dismissive term “mucking about with”, another misreading of what’s happening here. Recorded music is tinkered, tweaked, and mucked about with from the moment the first note is struck.

                Gamers aren’t tinkering and tweaking, they’re involved (“immersed”). And they’re being entertained far more than if they were watching a movie. That’s why the medium is so successful. Also – I never said (did anyone?) that the passive approach to art/entertainment is wrong. AI may eventually enable a similar kind of involvement in music as it does in the story-telling of gaming – exactly how is as hard to foresee as current gaming worlds would have been back in the days we clustered around the pong machine in the pub.

                • Cooling to my theme:

                  Steven Wilson is much cossetted here for his tinkering, tweaking and mucking about. Why not extend that broadmindedness to people you’ve never heard of?

                  Works of art (“records”) remain untouched by later versions and edits. The “original” will always be there.

                  Frank Zappa was notorious for tinkering, tweaking and mucking about. His CD versions of early albums were no improvement, to put it charitably. Van Morrison hated the strings on Astral Weeks. The artist is not always the best judge of his art.

                  Rap artists add little originality to sampled loops and beats and soundbites. Not enough tinkering, tweaking and mucking about for me, but mileage differs.

                  Since the dawn of the 12″ single, edited and altered versions of hits and would-be hits have become commonplace, often bearing little resemblance to the original. Tinkering, tweaking and mucking about. Why not do it yourself? It’s not that simple at the moment, but AI will make it much more accessible. Going back to gameplaying – that immersion comes at a price – hours and hours and days and days of learning how to play. It’s unreasonable to expect that music lovers will be less dedicated.

                  (And back to Smile: The Brian Wilson Presents version is nothing more than that. In terms of authenticity, there are edits of the original recordings (made by the Beach Boys and the Wrecking Crew) that make it sound artificial. The band, as excellent as it is, is a cover band. As a live experience, it was incredibly moving, as a record, it’s … interesting.)

                  • You don’t even have to look at recorded music to make a precedent for AI. For literally centuries nearly all so-called new music had been reconstituted version of music that already exists. The prizes, especially here at the Afterword, go to the artists – essentially meat-clad AI bots – who offer us the most convincing simulacrum* of the music we already know and love.

                    (*I don’t know if that’s the right word, but I like it)

                    • I don’t think this is true (allowing that, with only 12 notes in a chromatic scale, all music is built from the same building blocks).

                    • Re. Steven Wilson and Frank Zappa – they did their »tinkering« with their own creations, or (in the case of Wilson’s mixing jobs) were hired to do so by the original creators.
                      It’s not as if they »found« the master tapes on the internet and used ’em for their own purposes.

                      I doubt that Mr. Toyah will ever give me the »Sailor’s Tale« masters so that I can make my own AI-enhanced »Polka Mix«…

                    • fatima, you’ve just repeated what I said, but anyway. I did, of course, refer to Wilson’s tinkering on behalf of others. As to where people find the tapes, it really doesn’t matter. When AI lets us totally dismantle any recorded work into its constituent parts and recombine them, the notion of “original tapes” isn’t going to mean much. You won’t need them for your polka mix!

                    • Good try! Of course it is. Or would be, had he acknowledged the lift. He denied knowing anything about Helium, which as a legal defence is bulletproof. A lot of the issues raised here involve intellectual copyright, as vague a legal field as any. Ideas can occur to anybody – even Steven King.

                    • Thank you! Apart from one story being about a man who grows wings, and the other not, they’re identical. Another source for me was Jules Verne’s From The Earth To The Moon. Apart from one being about a man who leaves earth in a rocket and the other not, they’re basically the same. Also the Biggles books were a big inspiration, apart from the aeroplanes.

                    • Wow! You live and learn! I didn’t know it had been on the telly – although if Pyecraft sprouts wings, it’s a fairly extreme adaptation. Kinda misses the point of the story.

                    • The top review on IMDB says ” The storyline is uninspiring but might have been a bit racy when more people believed in Christ & Satan” – I hope he doesn’t mean England in 1986. We certainly didn’t believe in Christ, whereas Satan was living in 10 Downing Street…. right kids??

  12. SMiLE distresses me. Of course, there is amazing music in there but most of the fully realised songs came out a long time ago: Good Vibrations, Heroes And Villains, Cabin Essence, Our Prayer, Surf’s Up and Wonderful. The two-for with Wild Honey was educational but the 1990 box revelatory. I’ve listened to bootlegs, went to see the ‘complete’ concert and bought the 2004 album and the big box of Sessions. I had assumed the dross was incomplete or would be edited out. But no, lots of short fragments are in the complete version as performed by Brian and friends. Of course, each ‘song’ is a collection of fragments stitched together. Across the whole, it is too much. It sounds like, witnessing first hand, a mind imploding. Very sad and upsetting.

    I’ve listened to this too. I wouldn’t say it adds much.

Leave a Reply