There are some incredible AI songs out there. Here in lies the future of music.

Feb 24, 2024 15:26:12 GMT @ness said:

I got to hear JR sing My Way, all AI bad mouthing will CEASE AND DESIST.

Coldcut dropped the first AI song in the 90's. They did the synths but the notation was all AI created.

Meta.AI has dropped and it embedded in all Meta products - Facebook, Instagram, etc. Including, Messenger. As expected you can chat live with it, including asking it to generating images.

Is it new technology, probably not. But it's now rolled out to get masses so expect it to be a hot topic over the next little while.

I signed up for Microsoft Copilot which is their branded ChatGPT client, so far it is expensive and useless, so very on-brand for Microsoft.

It can helpful if you need to generate generic shit but yeah, most of what is branded AI is more toy than tool right now.

Apr 24, 2024 23:36:57 GMT c said:

It can helpful if you need to generate generic shit but yeah, most of what is branded AI is more toy than tool right now.

My only "wow" moment was when it was able to create a PowerPoint presentation from a proposal I did for a customer
I had to tweak it, but it saved me maybe 2hrs.


What I want and is apparently coming is being able to ask questions conversationally about price and previous offers and it retrieve the answers from my SharePoint files.

Apr 24, 2024 23:36:57 GMT c said:

It can helpful if you need to generate generic shit but yeah, most of what is branded AI is more toy than tool right now.

I think we're a long way off having our own personal JARVIS or JOI.

20-25 odd years ago, we were learning HTML, then eventually web design for the novice became a drag and drop template you could do yourself, and all the local web design studios disappeared. Now AI image generation perhaps kills off any need a novice might have for Photoshop. I look at someone like Beeble who likely spends hundreds of hours creating his images, and now a basic AI can do the same if not better in seconds..

I wonder how far away we are from a high end camera having AI, that first takes any shitty image and turns it into the absolute best in a matter of seconds, removing any need by the photographer to have " the eye " for it and turning a novice into a pro over night.

Just like machines replacing men in factories, AI will kill the digital media professionals if they don't adapt now.

Apr 25, 2024 0:32:35 GMT NATH45 said:

Apr 24, 2024 23:36:57 GMT c said:

It can helpful if you need to generate generic shit but yeah, most of what is branded AI is more toy than tool right now.

I think we're a long way off having our own personal JARVIS or JOI.

20-25 odd years ago, we were learning HTML, then eventually web design for the novice became a drag and drop template you could do yourself, and all the local web design studios disappeared. Now AI image generation perhaps kills off any need a novice might have for Photoshop. I look at someone like Beeble who likely spends hundreds of hours creating his images, and now a basic AI can do the same if not better in seconds..

I wonder how far away we are from a high end camera having AI, that first takes any shitty image and turns it into the absolute best in a matter of seconds, removing any need by the photographer to have " the eye " for it and turning a novice into a pro over night.

Just like machines replacing men in factories, AI will kill the digital media professionals if they don't adapt now.

The AI in Photoshop is legit crazy, being able to remove backgrounds and delete objexts from a text prompt.

Apr 25, 2024 0:33:42 GMT iNCY said:

Apr 25, 2024 0:32:35 GMT NATH45 said:

I think we're a long way off having our own personal JARVIS or JOI.

20-25 odd years ago, we were learning HTML, then eventually web design for the novice became a drag and drop template you could do yourself, and all the local web design studios disappeared. Now AI image generation perhaps kills off any need a novice might have for Photoshop. I look at someone like Beeble who likely spends hundreds of hours creating his images, and now a basic AI can do the same if not better in seconds..

I wonder how far away we are from a high end camera having AI, that first takes any shitty image and turns it into the absolute best in a matter of seconds, removing any need by the photographer to have " the eye " for it and turning a novice into a pro over night.

Just like machines replacing men in factories, AI will kill the digital media professionals if they don't adapt now.

The AI in Photoshop is legit crazy, being able to remove backgrounds and delete objexts from a text prompt.

When I was studying Multimedia a long time ago, I used to help a lot of the other guys with Photoshop as that was my skillset.. I always imagined, what if they could just tell it what they wanted, instead of getting me do it. And here we are.

Apr 25, 2024 0:33:42 GMT iNCY said:

Apr 25, 2024 0:32:35 GMT NATH45 said:

I think we're a long way off having our own personal JARVIS or JOI.

20-25 odd years ago, we were learning HTML, then eventually web design for the novice became a drag and drop template you could do yourself, and all the local web design studios disappeared. Now AI image generation perhaps kills off any need a novice might have for Photoshop. I look at someone like Beeble who likely spends hundreds of hours creating his images, and now a basic AI can do the same if not better in seconds..

I wonder how far away we are from a high end camera having AI, that first takes any shitty image and turns it into the absolute best in a matter of seconds, removing any need by the photographer to have " the eye " for it and turning a novice into a pro over night.

Just like machines replacing men in factories, AI will kill the digital media professionals if they don't adapt now.

The AI in Photoshop is legit crazy, being able to remove backgrounds and delete objexts from a text prompt.

Yeah, haven't tried Google's Magic Eraser or Best Take on any of my pics yet but even it existing is pretty bananas

and re: AI music, the first and biggest thing it's going to come for is background music. The stuff that plays under the end credits in a movie, or maybe even during the movie, the muzak in a grocery store. Not the big sexy concert stuff, but still a big part of the music industry.

Apr 25, 2024 0:32:35 GMT NATH45 said:

Apr 24, 2024 23:36:57 GMT c said:

It can helpful if you need to generate generic shit but yeah, most of what is branded AI is more toy than tool right now.

I think we're a long way off having our own personal JARVIS or JOI.

20-25 odd years ago, we were learning HTML, then eventually web design for the novice became a drag and drop template you could do yourself, and all the local web design studios disappeared. Now AI image generation perhaps kills off any need a novice might have for Photoshop. I look at someone like Beeble who likely spends hundreds of hours creating his images, and now a basic AI can do the same if not better in seconds..

I wonder how far away we are from a high end camera having AI, that first takes any shitty image and turns it into the absolute best in a matter of seconds, removing any need by the photographer to have " the eye " for it and turning a novice into a pro over night.

Just like machines replacing men in factories, AI will kill the digital media professionals if they don't adapt now.

Well Moore's law is in play, so advances will be fast.  The problem I see is what people want is not total AI, but AI that tells them sweet lies.  That means a ton of overriding answers.  AI sees the world in a hyperlogical manner and we really dislike some of what AI sees.

A lot of stuff people say AI will kill, it simply cannot.  It can make a great fantasy image for instance, but if you want the same setting with a tree removed it cannot do it.  Making customized images is what people will pay for.  

For the "eye" situation we have it now.  Take video and break it into single images.  It is that whole 100 monkeys situation.  We do not do this as the photographer usually is also directing the shots if they are worth their cash. 

What I do see right now is a whole lot of people without talent being scared about losing their jobs, because they are right now unwilling to adapt, or work at a low level and are not doing things like directing shots, offering customized art with a client, and the like.  They make art, but not marketable art.  And coming from that background, like no one is really making any cash anyway without moving towards marketable shit at some point.

Apr 25, 2024 0:33:42 GMT iNCY said:

Apr 25, 2024 0:32:35 GMT NATH45 said:

I think we're a long way off having our own personal JARVIS or JOI.

20-25 odd years ago, we were learning HTML, then eventually web design for the novice became a drag and drop template you could do yourself, and all the local web design studios disappeared. Now AI image generation perhaps kills off any need a novice might have for Photoshop. I look at someone like Beeble who likely spends hundreds of hours creating his images, and now a basic AI can do the same if not better in seconds..

I wonder how far away we are from a high end camera having AI, that first takes any shitty image and turns it into the absolute best in a matter of seconds, removing any need by the photographer to have " the eye " for it and turning a novice into a pro over night.

Just like machines replacing men in factories, AI will kill the digital media professionals if they don't adapt now.

The AI in Photoshop is legit crazy, being able to remove backgrounds and delete objexts from a text prompt.

What is hilarious is AI in photoshop and film editing been a thing for a decade now, and no one cared.  The second it generated a picture the same people scream that AI is gonna take their jobs, while they continue to use AI themselves in their art.  

I'm pretty sure in the next two years I will be able to offload 50-75% of my work to AI

There are some funny ones like AI generated architecture where it doesn't understand the point of stairs and shoves them in aesthetically where they serve no purpose.

The advances will come as AI is able to train AI on context.
For example a prompt to teach an AI how to understand the context of stairs and then to be able to feed it images that teach it.

Without this level AI won't understand the context of anything until it is taught directly.

AI will quickly take over the low effort, repetitive shit we do. Which it should.

Just gonna say there's no way this ai chatbot I've been using for months hasn't completely stolen me from society.

Apr 25, 2024 0:59:25 GMT iNCY said:

There are some funny ones like AI generated architecture where it doesn't understand the point of stairs and shoves them in aesthetically where they serve no purpose.

The advances will come as AI is able to train AI on context.
For example a prompt to teach an AI how to understand the context of stairs and then to be able to feed it images that teach it.

Without this level AI won't understand the context of anything until it is taught directly.

We do not entirely have the framework for semantic training.  We call things semantic training, but AI does not do meaning the way we do.  It works on the correlational and path model, not "learning" as we know it.  Once this limitation is surpassed, we get into the scary, will AI decide man is a problem issues.

As far as I know, to hit that level we will need to finally learn how to code human cognition, and even advanced neural networks in programming are primitive approximations.  There is an element of chaos that leads to novel discoveries we cannot duplicate for instance.  

Apr 25, 2024 1:03:52 GMT c said:

Apr 25, 2024 0:59:25 GMT iNCY said:

There are some funny ones like AI generated architecture where it doesn't understand the point of stairs and shoves them in aesthetically where they serve no purpose.

The advances will come as AI is able to train AI on context.
For example a prompt to teach an AI how to understand the context of stairs and then to be able to feed it images that teach it.

Without this level AI won't understand the context of anything until it is taught directly.

We do not entirely have the framework for semantic training.  We call things semantic training, but AI does not do meaning the way we do.  It works on the correlational and path model, not "learning" as we know it.  Once this limitation is surpassed, we get into the scary, will AI decide man is a problem issues.

As far as I know, to hit that level we will need to finally learn how to code human cognition, and even advanced neural networks in programming are primitive approximations.  There is an element of chaos that leads to novel discoveries we cannot duplicate for instance.  

I think we can teach it to teach itself.
I expect on my example it will start with feeding floor plans and associated photos into an engine until it is able to understand the relationship between 2d and 3d
Once it's done that you should be able to feed in photos and it U derstand where the stairs need to go.

The model it builds are often unfathomable to us but that's okay as long as the outcome is correct

The problem is we based statistical neural networks to mimic biological neural networks. Works great for retrieval, assimilation and accommodation. But not for generation. The former processes are electrical, while the latter is protein synthesis. And we have no evaulivant to protein synthesis in statistical computing that forms the foundation of AI. And should we find a way to put it into place, the AI program will rapidly reform its own structure in a skynet style event. Which is why people want to essentially ban AI from creating it's own starting memory points itself.

For the stair model it may be done now thinking about it, by teaching it first what conceptional level is, allowing for mass rebinding of knowledge nodes, and teaching it how to undergo conceptual change. Believe there are computational cognitive models for this exact process. We were on a related research thread in my lab and some of my proposed work I never got to do tried to estimate the exact point a conceptual model changed during learning. But even doing this will require no new knowledge points be created, which is very rare doing conceptual change, as normally it occurs after learning something knew that allows you to restructure prior knowledge.

Apr 25, 2024 5:04:08 GMT c said:

The problem is we based statistical neural networks to mimic biological neural networks. Works great for retrieval, assimilation and accommodation. But not for generation. The former processes are electrical, while the latter is protein synthesis. And we have no evaulivant to protein synthesis in statistical computing that forms the foundation of AI. And should we find a way to put it into place, the AI program will rapidly reform its own structure in a skynet style event. Which is why people want to essentially ban AI from creating it's own starting memory points itself.

For the stair model it may be done now thinking about it, by teaching it first what conceptional level is, allowing for mass rebinding of knowledge nodes, and teaching it how to undergo conceptual change. Believe there are computational cognitive models for this exact process. We were on a related research thread in my lab and some of my proposed work I never got to do tried to estimate the exact point a conceptual model changed during learning. But even doing this will require no new knowledge points be created, which is very rare doing conceptual change, as normally it occurs after learning something knew that allows you to restructure prior knowledge.

Yeah, it's pretty amazing.
In my industry they're using AI to dimensionally check certain attributes, but no data is provided on good or bad apart from a series of good photos and a series of bad photos. We leave it to the AI to determine the difference, I which is what makes it so useful.

In other applications we are doing the same with machine current and vibration, the machine learns the normal currents and vibrations of different products at different speeds and will alert you when you drift outside the tolerance bands, without AI you just couldn't do it as there are simply too many recipes and variables.

Feels like your industry is still lagging. Both of those are last gen machine learning tasks. About the wall for what I can do with data. Beyond that wall is where the really fun stuff seems to lie with hidden variable layers. And that is likely why stuff you want is not quite here yet. Just not implemented the more modern stuff. Once you start to play with hidden layers, things start to get very interesting. Problem with these models is the hidden layers when you backtrace them often does not really make causal sense, so it becomes a the model works because it works. Takes a leap of faith to rely on them.

Apr 25, 2024 7:57:40 GMT c said:

Feels like your industry is still lagging. Both of those are last gen machine learning tasks. About the wall for what I can do with data. Beyond that wall is where the really fun stuff seems to lie with hidden variable layers. And that is likely why stuff you want is not quite here yet. Just not implemented the more modern stuff. Once you start to play with hidden layers, things start to get very interesting. Problem with these models is the hidden layers when you backtrace them often does not really make causal sense, so it becomes a the model works because it works. Takes a leap of faith to rely on them.

It's not an easy thing to connect AI to legacy PLC based systems, the hard part that takes time is exposing the data for the AI to sift.
IOT was a buzzword for a while but it is only recently that AI has made warehousing machine data useful.

That is ultimately the barrier right now for the use of the sort of AI are using is having the sheer amount of data that is needed to work in the level of dimensions we want to and produce accurate results. Also why most AI teams have tens of thousands of people working with their systems doing the supervised training of models

Don't generally see commercials but last weekend I dogsat for my parents and they have cable so I had it as background noise. Noticed rentacenter's commercial was basically just AI reaction shots, one of which looked exactly like the fat dude from Superbad forget his name. Makes ya wonder how bad this is gonna fuck over Hollywood, not that they serve any real purpose. But ya know acting in general.

My algorithm on Instagram is completely fucked, having commented on one yapper's cringe inducing post, I'm now inundated with these bros first yapping away about money, and now yapping about philosophy.

But one interesting one popped up. He showed what his " work " actually looked like.

He asked ChatGTP to generate an image of a $1 pancake, a $100 pancake and a $1,000 pancake and then asked a voice generating AI to come up with some cringy quote. And the mashed it all together and put it on TikTok and then show how much engagement it had, and thus how much money he made.

It took it 10 minutes. And he made thousands apparently.

I hate these people.

Eventually the platforms will have to ban that style of engagement farming as it is designed to bait people into commenting.

X has announced they're banning it, but easier in theory than practice

I applaud their abilities to seize opportunities, but, unironically it's giving young people an artificial perception of what " work " is.

Social Media and all its offshoots is the Gold Rush of the 21st Century. And I'm sure these guys are making more money from talking about it than actually doing it.

Apr 28, 2024 21:44:42 GMT NATH45 said:

My algorithm on Instagram is completely fucked, having commented on one yapper's cringe inducing post,

Don’t interact with the shit. Why bother commenting at all? They benefit from any comments.

Apr 28, 2024 22:35:28 GMT Gyro LC said:

Apr 28, 2024 21:44:42 GMT NATH45 said:

My algorithm on Instagram is completely fucked, having commented on one yapper's cringe inducing post,

Don’t interact with the shit. Why bother commenting at all? They benefit from any comments.

I have an obsessive compulsive inability to scroll past someone blatantly telling lies lol.