hamo wrote: ↑Tue Jul 09, 2024 10:04 am
I have read a few posts (so anecdote more than data) about movie extras signing away the rights to their likeness, so they can be digitally scanned and reused in any movie in perpetuity. Might be only a matter of time before that moves beyond extras. That was a big part of what was happening with the guild strikes in the USA last year.
Theres no way it should be legal to be able to sign away your likeness, it just goes to show how the business is way more interested in money than the people involved in performing
Solicited AI bots have been turning up to Teams meetings on behalf of customers where my wife works. They're from a subscription service, they join the meeting on mute and provide a summary and transcription back to the customer.
Cdog wrote: ↑Tue Jul 09, 2024 10:19 am
Solicited AI bots have been turning up to Teams meetings on behalf of customers where my wife works. They're from a subscription service, they join the meeting on mute and provide a summary and transcription back to the customer.
Wowsers, some huge privacy concerns there, we'd never be allowed anything like that at work!!
Cdog wrote: ↑Tue Jul 09, 2024 10:19 am
Solicited AI bots have been turning up to Teams meetings on behalf of customers where my wife works. They're from a subscription service, they join the meeting on mute and provide a summary and transcription back to the customer.
Leaving out whether it's right or wrong, this is actually a great use of the technology. They've been trained well enough to be able to distinguish accents and accurately transcribe conversations, they are probably good enough to identify and separate multiple people talking at once. They're even probably good enough to. summarise a full transcript as well.
What they are likely not good enough at is inferring and making decisions based on the conversation that took place in the meetings.
Machine Learning techniques are really going to make an impact in the world of music, we're already seeing them in the modelling spaces where a well trained algorithm will likely do a better job of digitally synthesizing the sound of an amp/effect/speaker/instrument than a "modelled" amp created by older school simulators such as individual component or subsystem simulators that I'm guessing are used in some of the modellers out there. Likewise I can only imagine the leaps that we could make in creating new effect pedal type toys and DAW helpers/assistants.
All this stuff is going to be a huge help and will be reasonably invisible, certainly to the end user consuming the music by listening to it.
If I had the time, knowledge and skills maybe, I think there might be a niche in the market for a digital songwriting partner, really a tool kit. Imagine taking what we can already do with sequenced drums and instruments etc. and combine it with the ability to describe what you are after. Maybe play a chord sequence and get suggestions on voicings that could make it more interesting, generate a number of variations, be able to be prompted for the variation. Likewise, maybe start with the chord progression, sing, clap or put the rhythm into it. Maybe then sing or hum a melody, capture it and then finesse it before maybe choosing the instrumentation etc. neaten it up, or maybe even bugger around with the timings and tunings, how long before we can get the whole song to swing nicely as in the opposite of clamping it to the grid. All the while keeping track of the steps taken, allowing you to branch and rewind, go back in some areas and change what you did etc. you build up this whole workspace if the journey to write a song. All this will be easy on a decent next gen laptop and likely on the next gen of phones as well.
None of this is really AI creating songs, it's just using technology to allow creatives to be even more creative.
Obviously none of this is sexy enough to make headlines, whereas little Jonny being able to create a passable rip off of the latest Boy Band, Rapper, singer songwriter and flood Spotify makes far better headlines and stories for public consumption
I know nothing about it and don’t want too, but from comments in another forum I was in, udio is already pretty much doing the above, including vocals, I’ve heard the tracks, the style isn’t my cup of tea so can’t comment on if they were good or not, but you’d be hard pressed to pick it wasn’t a human singing….
NZTone.e wrote: ↑Tue Jul 09, 2024 1:28 pm
I know nothing about it and don’t want too, but from comments in another forum I was in, udio is already pretty much doing the above, including vocals, I’ve heard the tracks, the style isn’t my cup of tea so can’t comment on if they were good or not, but you’d be hard pressed to pick it wasn’t a human singing….
My intention is not to produce finished or even almost finished songs, more that it's like having some tools to help with the actual composing process, although I guess helping create the demos start to get into Udio etc. territory.
NZTone.e wrote: ↑Tue Jul 09, 2024 1:28 pm
I know nothing about it and don’t want too, but from comments in another forum I was in, udio is already pretty much doing the above, including vocals, I’ve heard the tracks, the style isn’t my cup of tea so can’t comment on if they were good or not, but you’d be hard pressed to pick it wasn’t a human singing….
My intention is not to produce finished or even almost finished songs, more that it's like having some tools to help with the actual composing process, although I guess helping create the demos start to get into Udio etc. territory.
Yeah, but regardless of personal opinions, I’m against it for example, but it’s just another form of making music isn’t it?
No different from going from acoustic to electric guitars really, or pianos to synths and so on.
I can see how it would be useful for creating gingles for ads, or even movie score type stuff, but that’s just taking away another human’s job.
And then there’s the other end of the scale…once the AIs have analysed every hit song in the last 80 years they’ll have an algorithm that produces a hit song and they’ll just churn out the hits…
NZTone.e wrote: ↑Tue Jul 09, 2024 1:28 pm
I know nothing about it and don’t want too, but from comments in another forum I was in, udio is already pretty much doing the above, including vocals, I’ve heard the tracks, the style isn’t my cup of tea so can’t comment on if they were good or not, but you’d be hard pressed to pick it wasn’t a human singing….
My intention is not to produce finished or even almost finished songs, more that it's like having some tools to help with the actual composing process, although I guess helping create the demos start to get into Udio etc. territory.
Yeah, but regardless of personal opinions, I’m against it for example, but it’s just another form of making music isn’t it?
No different from going from acoustic to electric guitars really, or pianos to synths and so on.
I can see how it would be useful for creating gingles for ads, or even movie score type stuff, but that’s just taking away another human’s job.
And then there’s the other end of the scale…once the AIs have analysed every hit song in the last 80 years they’ll have an algorithm that produces a hit song and they’ll just churn out the hits…
jeremyb wrote: ↑Mon Jul 08, 2024 6:44 pm
Was watching something the other day and the dude said “Your ears are analog so digital amps will never sound right!”, interesting take.
That's silly. The issue is sampling rate/ quantity of information. I'd challenge anyone to correctly determine a blind digital vs analogue test with the best available gear. What hasn't happened in my (limited) experience is digital amps feeling analogue to play. It's only a matter of time though.
Its also amusing when you consider the output, i.e. a speaker is an analog thing and your ear is an imperfect listening device...
jeremyb wrote: ↑Mon Jul 08, 2024 1:12 pm
AI tools to take out the drudgery of creative work will be the big winner I hope!
If they do use AI to make music its only going to be pop music anyway, and thats pretty much garbage these days so no one will know the difference or care, its interesting how streaming has changed modern pop music though, it all goes bang straight into the chorus, cause if you don't hook the listener in the first 5 or so seconds they will skip to the next track in the playlist...
The creativity depends on the drudgery. Art is made by work. Musicians spend long hours in rehearsal rooms and studios getting it right. Modern pop songs are garbage because they are made with formulas, to avoid the drudgery.
sty wrote: ↑Mon Jul 08, 2024 12:48 pm
A good learning exercise for yourself, which has been mentioned before, is to use one of the current AI tools like ChatGPT etc. to do something for you in a topic that you know really well. You will find huge glaring mistakes in it I'm sure. So how could you possibly trust what it tells you about a topic you know little or less?
I have done this experiment. The robots know nothing about bungalows.
jeremyb wrote: ↑Mon Jul 08, 2024 1:12 pm
AI tools to take out the drudgery of creative work will be the big winner I hope!
If they do use AI to make music its only going to be pop music anyway, and thats pretty much garbage these days so no one will know the difference or care, its interesting how streaming has changed modern pop music though, it all goes bang straight into the chorus, cause if you don't hook the listener in the first 5 or so seconds they will skip to the next track in the playlist...
The creativity depends on the drudgery. Art is made by work. Musicians spend long hours in rehearsal rooms and studios getting it right. Modern pop songs are garbage because they are made with formulas, to avoid the drudgery.
This is an interesting point, but I think creativity is in the vision, but agree that the execution takes the work, all great artists have got where they are by making tons of shit stuff first, was that because their vision was shit or the work was shit, hmmmm... more thought required.
