Some thoughts about AI acceptance

[pekbro]

My friend Craig, who's a respected artist in the film industry uses AI for inspiration, he says it
gives him ideas he had not thought of on his own, then he paints the stuff by hand based on that.In his case though, he is a better artist than any AI hands down, even at photo-realism. No joke.

There are some truly incredible artists out there!

AI is limited for photo-realistic imagery. AI images are static. It has a hard time creating a sense of movement. Characters for example, are posed, and not in motion.

Yeah, I truly wish I could imagine the world with the same level of detail as Craig does.
Sadly, I can only imitate him, one thing I have in common with AI I guess.

[BONES]

Unfortunately, there's no further info on how those creative professionals used AI. Some of them could be using AI to reference software manuals.

It being Adobe, it's reasonable to assume they were asking about their own AI tools in Photoshop, Illustrator and some of their other products. They have their own AI, Firefly but the most recent updates give you the option of using third party AI instead of theirs, if you prefer.

I use Firefly pretty much every day at work. One of the most onerous tasks a graphic artist has to do is cut out a subject from a photo. 10 years ago it was something you could spend most of a day on if it was an important job, something that would be seen often over a period of time. Today Firefly can do it to the same standard in 30 seconds. Just 6 months ago, when it had been around for a couple of years, it still required the artist to spend 15-30 minutes tidying it up but today it pretty much nails it every time.

I also use it to extend the background of an image several times a day, so that we can frame things better for a graphic. That was something we used to do manually, using a combination of the clone tool and simply copying and pasting of parts of the image. It could take an hour sometimes but, again, Firefly doe sit in seconds.

But Photoshop and After Effects have had smart tools for a while that aren't AI, tools that can recognise and isolate shapes in an image. Sky Replacement can find the horizon in an image on it's own and the Roto Brush tool can track a shape in a video clip pretty well. But they've been around for years now.

Adobe have also linked their AI to Adobe Stock, which most subscribers have free access to. You can use it to generate your own images or to edit/enhance anything you get from Adobe Stock. (If you don't know, "stock" images and video have long been part of graphics work. It's the graphic equivalent of using samples.)

Anyway, the point of this long exposition is to say that within my profession, moving from "traditional" tools to smart tools to AI tools to using AI to create images form nothing feels like a gradual, natural evolution of the process. Will it put a lot of artists out of work? Yes, it already has, but not as many as Moore's Law has, as we're able to do so much more in any given timeframe than we could even 10 years ago.

Its the same with music - moving from analogue hardware to digital hardware to using a computer to sequence your hardware to doing it all in your computer with a DAW and plugins to using AI tools to letting AI generate complete songs feels like a completely natural progression if you've been involved with the process as long as I have. Again, I reckon Moore's Law has had a bigger impact than AI (so far, at least). It happened over a longer period of time, so it was less disruptive, but its impact overall has been far greater.

To me, the hue and cry over AI seems deeply rooted in (small) egos. People feel threatened, they feel they may be unable to compete, blissfully unaware that they were never able to compete anyway. I mean, seriously, where is all this human creativity you keep bangin' on about? I'm not seeing it. All I'm seeing is tiny people with tiny minds doing tiny, inconsequential things that nobody has any interest in (myself included, of course). Wake up to yourselves, it was all over for you long before AI came on the scene.

[BONES]

BTW Supreme Court just ruled (or ignored ruling) about AI art being copyrightable. So it's the same with music then, it's not copyrightable so anyone could take your AI stuff and do whatever they want with it, like resell it.

That's not what it means. If what they do with it results in a loss of income or reputation for you, there are other remedies available. e.g. People selling fake Rolex watches aren't being charged for copyright offences, they are being charged with fraud. So they can't just take your AI generated mp3 and sell it on but they can copy an AI generated riff or rhythm and use it in their own work. All of which seems perfectly reasonable to me, much as you can mostly get away with sampling someone else's music and using it your own work (fair use). I'd actually be quite flattered if someone did it with any of our music, AI generated or otherwise. it wouldn't bother me at all.

Normally, I'd have had to go into the "rubber stamp" tool and remove each bit and replace it with what should be there. I'm actually great at this, but it takes a lot of time to make it look perfect.

The Healing Brush would be a much faster way to do this and it's been in Photoshop for 20 years or more. I reckon nit would be faster than using AI.

But why are these people in positions in the first place? My working theory is that, like who we make our president, we pick coworkers who seem like you'd have fun hanging out with.

Sorry for the digression but I think this is something male bosses are terrible at but female bosses tend to do very well. Women are really good at picking people who fit in with the rest of the team. Men, in my experience, rarely even consider things like that. Every workplace I've been with a woman in charge has been a far better workplace than those with men in charge. Turnover is noticeably lower and people tend to work harder, not for the company but for each other. Nobody wants to let their colleagues down. OTOH, in workplaces run by men, the environment tends to be more competitive and cut-throat. Some people burn out and are discarded and others move on at the first opportunity. All the best bosses I've worked for over the last 30 years have been women.

AI is limited for photo-realistic imagery. AI images are static. It has a hard time creating a sense of movement. Characters for example, are posed, and not in motion.

You do know that you can generate AI video, too, and that Ai has absolutely nailed human movement, right? It struggles with other critters, ducks tend to walk like people. That's started to change, too, just as AI overcame it's problem with hands.

[VariKusBrainZ]

Some thoughts about human acceptance

https://m.youtube.com/watch?v=SbEqMkxEzvA

[BONES]

More the other way around, really.

[bermudagold]

1. Will local AI tools be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local AI tools?
3. If the answer to 1&2 is yes, will acceptance/adoption of local AI tools pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused app I've seen with a localized AI option is resing...are there any others?

[zerocrossing]

My friend Craig, who's a respected artist in the film industry uses AI for inspiration, he says it
gives him ideas he had not thought of on his own, then he paints the stuff by hand based on that.In his case though, he is a better artist than any AI hands down, even at photo-realism. No joke.

There are some truly incredible artists out there!

AI is limited for photo-realistic imagery. AI images are static. It has a hard time creating a sense of movement. Characters for example, are posed, and not in motion.

Yeah, I truly wish I could imagine the world with the same level of detail as Craig does.
Sadly, I can only imitate him, one thing I have in common with AI I guess.

None of that is really true. AI video is all the range. It's often a bit clunky, but it has made a lot of progress.

My illustration technique is to quickly mock up something rough in Maya, just to get an idea of proportion, lighting and general composition. Then I'll illustrate over that, though I'm no photorealistic master. Usually I'm just sketching something up to communicate a concept. We had a concept art guy on staff at the last place I worked at who was an amazing digital painter. I have no idea how he worked, but it was very good. I had to laugh that on our last game we pitched, the crappy art director (really a technical artist who failed up) just used a bunch of AI in the pitch deck. I figured our guy must be busy with something, but later I talked to him and he admitted he was just sitting around waiting for a task.

(we were not awarded a contract for the game)

[ghettosynth]

1. Will local LLMs be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local LLMs?
3. If the answer to 1&2 is yes, will acceptance/adoption of local LLMs pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused AI I've seen with a localized LLM option is resing...are there any others?

Do you realize that he sounds like a DJ talking about voice leading? Do you, or anyone here, think that this is really what people who use AI productively are doing with it?

Yes and no is the answer to your question.

Oh, and his dumb analogy is not insightful. He's doing a very typical KVR move "I know something about something so I'll analogize that something to the thing that I'm butthurt about and show that my strawman won't hold up and my butt will feel better." He should stay in his lane.

f**k me.

[bermudagold]

1. Will local LLMs be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local LLMs?
3. If the answer to 1&2 is yes, will acceptance/adoption of local LLMs pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused AI I've seen with a localized LLM option is resing...are there any others?

Do you realize that he sounds like a DJ talking about voice leading? Do you, or anyone here, think that this is really what people who use AI productively are doing with it?

Yes and no is the answer to your question.

Oh, and his dumb analogy is not insightful. He's doing a very typical KVR move "I know something about something so I'll analogize that something to the thing that I'm butthurt about and show that my strawman won't hold up and my butt will feel better." He should stay in his lane.

f**k me.

that's the question...who has quantified how big the gap is between enterprise, small business, and consumer use cases?...yeah the analogy is tenuous in its oversimplification...several of the folks in the comments agree with you and point out the holes...he's a sensationalist for engagement...but that's the majority of what social media has become at the end of the day...lotsa folk look to him however on the intersection of music and AI topically

[zerocrossing]

1. Will local LLMs be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local LLMs?
3. If the answer to 1&2 is yes, will acceptance/adoption of local LLMs pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused AI I've seen with a localized LLM option is resing...are there any others?

ReSing isn't an LLM. It's a Vocal Singing Model. I don't know if something similar, like Synthesizer V, is processing locally. I should turn off my ethernet and find out. Whatever it's doing, it seems to happen pretty fast, but not instantaneously. If I type in a word, there's like a second where you can see it redraw the audio.

[bermudagold]

1. Will local LLMs be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local LLMs?
3. If the answer to 1&2 is yes, will acceptance/adoption of local LLMs pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused AI I've seen with a localized LLM option is resing...are there any others?

ReSing isn't an LLM. It's a Vocal Singing Model. I don't know if something similar, like Synthesizer V, is processing locally. I should turn off my ethernet and find out. Whatever it's doing, it seems to happen pretty fast, but not instantaneously. If I type in a word, there's like a second where you can see it redraw the audio.

whatever the semantics...its using AI that you can train locally to create your own instruments
"Unlike cloud-based tools, ReSing works directly on your computer, eliminating the need for data transmission and queuing. Additionally, ReSing is the first platform of its kind that allows you to model your voice or instrument on your device for personal use or lease it for profit."...Also allows you to "Import voice RVC** models from popular open communities"

[zerocrossing]

1. Will local LLMs be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local LLMs?
3. If the answer to 1&2 is yes, will acceptance/adoption of local LLMs pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused AI I've seen with a localized LLM option is resing...are there any others?

ReSing isn't an LLM. It's a Vocal Singing Model. I don't know if something similar, like Synthesizer V, is processing locally. I should turn off my ethernet and find out. Whatever it's doing, it seems to happen pretty fast, but not instantaneously. If I type in a word, there's like a second where you can see it redraw the audio.

whatever the semantics...its using AI that you can train locally to create your own instruments
"Unlike cloud-based tools, ReSing works directly on your computer, eliminating the need for data transmission and queuing. Additionally, ReSing is the first platform of its kind that allows you to model your voice or instrument on your device for personal use or lease it for profit."...Also allows you to "Import voice RVC** models from popular open communities"

I guess Synthesizer V is the same, though it's not transforming a recording, but instead using text on a piano roll. I personally find it sounds significantly better than ReSing, and while it's more of a hassle to get your vocal set up, the ability to really edit each note is phenomenal.

[bermudagold]

1. Will local LLMs be sufficient for most consumer use cases?
2. Will most consumers decide to be tech savvy enough to install, configure, and maintain local LLMs?
3. If the answer to 1&2 is yes, will acceptance/adoption of local LLMs pop the AI bubble...doing to datacenters something similar as what was done to recording studios as posited below?


the only music focused AI I've seen with a localized LLM option is resing...are there any others?

ReSing isn't an LLM. It's a Vocal Singing Model. I don't know if something similar, like Synthesizer V, is processing locally. I should turn off my ethernet and find out. Whatever it's doing, it seems to happen pretty fast, but not instantaneously. If I type in a word, there's like a second where you can see it redraw the audio.

whatever the semantics...its using AI that you can train locally to create your own instruments
"Unlike cloud-based tools, ReSing works directly on your computer, eliminating the need for data transmission and queuing. Additionally, ReSing is the first platform of its kind that allows you to model your voice or instrument on your device for personal use or lease it for profit."...Also allows you to "Import voice RVC** models from popular open communities"

I guess Synthesizer V is the same, though it's not transforming a recording, but instead using text on a piano roll. I personally find it sounds significantly better than ReSing, and while it's more of a hassle to get your vocal set up, the ability to really edit each note is phenomenal.

cool...always useful to get data points from people who have actually tried/used these things...i haven't seen any demos of resing's user trained models...only the typical sponsored youtube fluff...BT's product at least runs offline but I dont think you can train local...and it seems the results from ones that run local arent as good as suno or ace studio...but this market is crowded already...i've been keeping a list of musical AI apps and its over like 30 already with tons of overlap

[ghettosynth]

that's the question...who has quantified how big the gap is between enterprise, small business, and consumer use cases?...yeah the analogy is tenuous in its oversimplification...several of the folks in the comments agree with you and point out the holes...he's a sensationalist for engagement...but that's the majority of what social media has become at the end of the day...lotsa folk look to him however on the intersection of music and AI topically

PRONON expresses their thinking process run on a local model for a typical consumer task.

https://www.youtube.com/playlist?list=P ... ezaETHbNvq

[pekbro]

I was just thinking an AI online shopper would be handy, tell it what you want and it will find you the absolute best deal possible, wherever that may be.