Search form

[blinddrew]

Having read the very interesting article on AI in this months magazine (link for subscribers and future readers: https://www.soundonsound.com/music-business/ai-music)
I've been thinking about it over the weekend and thought I'd start a thread here to mull it over a virtual pint with other deep thinkers (and the readers of this forum )...
So a few things to kick the discussion off:
1) I thought the comment about the different uses of music was very interesting - stuff that is there to entertain vs stuff that is there to fill a gap. I'm doing a fair bit of video stuff at work now and for that all we need is background music - stuff like Jukedeck (www.jukedeck.com) looks like that could cover that base completely. Whither library music composers?
2) Being a bit of a nerd, and one who's new to a lot of the tools and techniques, I like the creative and the problem solving sides of the mix process. A tool that could level out and make a vocal more prominent at the touch of a button (or word of command) would undoubtedly be quicker, but would take away some of the enjoyment for me. But I recognise that for someone who just wants to get on with the music it could make a massive difference. I do wonder what challenges it would introduce if someone else is asked to then do a manual job on a track that has already been AI mixed?
3) What would really be interesting for me would be things that would speed up the set up process, "Computer, open Reaper, new project, recording template 2. Load vocal, bass and guitar template tracks." It could be doing that whilst i'm setting up the hardware - or if the hardware is already in place, it's doing that whilst I'm repeating the new bit I've just written trying not to forget it.
4) But the biggie for me where AI could make a difference, rather than just a voice control and some standardised presets, is at the end of the process. This was a touched on a bit in the article, but the thing about music making that I really hate, is self-promotion. An AI toolset that helped me draft promotional content and schedule it to effective windows and platforms would be hugely useful. Another tool that analysed my music, and then analysed the buying patterns to identify suitable niche markets, and helped target the right potential audiences would be the icing on the cake. And as we move further down the line of an increasingly democratised music making process, getting your music in front of the right listeners is going to be even more important - otherwise we'll all just drown in the noise (no matter how well produced) and the only winners will be those with the largest marketing budgets.

Discuss. Or don't.

P.S. I was reminded a bit of this thread but it's a bit more tech rather than AI focused: https://www.soundonsound.com/forum/view ... re#p507081

[desmond]

I was also pleased to read this and was thinking of starting a similar thread.

There has been a big push on machine learning over the past few years, and yet until recently we hadn't seen a huge range of audio software using these techniques - and it might be able to solve some interesting issues.

I still find machine learning a bit of a "magic black box" solution - I mean, I intellectually know what's going on and how it works, but it's still seems kind of incredible you can develop a model, train it, and then have it perform useful work doing things that were not practically possible by computers a decade earlier. And the tools to bring machine learning into the casual developer's toolkit are getting very interesting and accessible indeed.

And now we *are* starting to see audio apps using these techniques - to analyse and classify samples, for example. It will be interesting to see where this goes...

[ManFromGlass]

My first reaction was hearing the bells of doom tolling for media composers. Another nail in the traditional composers coffin.
But I will reread the article slowly and chew on it a bit before deciding how much of a Luddite I want to be.

[zenguitar]

I was reading SOS in the pub on my tablet tonight and I learned a lot from the article.

My iPad mini is getting long in the tooth, and the tablet subscription crashed it several times tonight. What I learned when the AI article crashed the app once more was that I really can't put the upgrade off any longer.

Andy

[OneWorld]

I was reading SOS in the pub on my tablet tonight and I learned a lot from the article.

My iPad mini is getting long in the tooth, and the tablet subscription crashed it several times tonight. What I learned when the AI article crashed the app once more was that I really can't put the upgrade off any longer.

Andy

I wouldn't put too much reliance on AI, as we don't know what intelligence is yet. Many AI apps aren't 'intelligent' at all, they just use sophisticated pattern matching. Some say that Amazon gizmo (and the Google thing as well) are AI driven, they are not. For example if you might ask it "What's the weather like in London today?" all it does is search it's database for the nouns "weather" + "london" +"today" and strings it into a sentence. A sentence is a template and so the words are put in the correct order. That being said a philosopher might ask "do you need intelligence to comment on the weather anyway?"

This research (getting a computer to understand and use natural language, as opposed to an artificial language such as music) has been around for ages, the most famous anecdote came from the last cold war when the US didn't have enough English<->Russian translators wrote code that would do the job. It could of course handle simple sentences easily "Boris will be stood at the corner of the street at 8:00 and will be holding a copy of Pravda in his right hand" No ambiguity.

So they tried it on more challenging language, Shakespeare - "The flesh is willing but the spirit is weak" and the machine came back with "The meat is ok but the vodka's rubbish"

AI generally doesn't come up with anything novel or creative, it just does what we already do but infinitely faster. That is why a machine beating a grandmaster chess player isn't necessarily intellectually superior, all it does is check all the possible moves (patterns) that have already been used and what outcome there might be.

A similar experiment on intelligence was done comparing cats vs dogs. A food dispenser was placed in a room. The animal handler pressed a large button and some food would be dispensed from a contraption. The dog got this right away, it saw the button pressed, did the same with his paw, and saw the food dispensed, the dog got a reward.

The cat wasn't quite so quick on the uptake, though it did 'get it' after about 3 tries. So the dog was proved to be more intelligent.

But on the 5th go, the cat didn't press the button at all. It sat for a moment then went behind the contraption dispensing the food without even going near the button, and there it gorged on all the food without faffing about with buttons. Now which was the more intelligent, the dog had been pattern matching, but the cat worked out where the food was coming from?

Jaron Lanier said that artificial intelligence doesn't exist, it was just a concept bandied about in order to get funding for research.

[The Red Bladder]

Jaron Lanier said that artificial intelligence doesn't exist, it was just a concept bandied about in order to get funding for research.

That made me smile! That sentiment is not without a great deal of truth!

I can see a great deal of pot-boiler music being churned out by algorithms and indeed, I have done this many times, when the idea is to just have some sort of 'harmonic noise' in the background. But then composers, esp. for 2nd-rate films, have been sitting down and banging out some goofy chord progressions on a piano, since the dawn of the 'talkies'.

BUT billions are being spent in the race to create the first true quantum computer and when that happens, everything you ever thought you knew about the information age will change.

We've only just begun our journey into the IT age. After just a few short years of small computers (30-40 years tops!) we now have CGI in every movie made - even home videos can feature things like CGI muzzle-flashes and wire removal - things that 20 years ago were state-of-the-art and required a £200,000 box from Quantel to do properly!

Our present-day linear computers have already made animation movies like 'War of The Planet of the Apes' possible. Weta Digital is working on a whole new set of tools and toys to make animation even more realistic and all sorts of rumours abound about what will be possible for 'Narnia - The Silver Chair' which is just going into principle photography.

After photo-realistic apes, the next step is photo-realistic people. Old B&W footage will be totally reconstructed, long-dead actors will perform once more and Mozart will be churning out even more of his second-rate operettas!

[blinddrew]

I've been kicking our organisation into the late 20th century over the last couple of years and started doing some video work here. Because of the nightmare of music licensing I started looking at some of the AI options (I think that was following an SOS article as well). I've been keeping a particular eye on Jukedeck over the last year or so and it's interesting to see how far they've come. For a lot of the stuff we do, where all we need is the above-mentioned 'harmonic background noise', they will do the job. And it's £200 for complete rights ownership.
I'm not sure I could even churn something out myself that cheaply - and I'm cheap!