AI can now generate music. from text........

J

Jan Panteltje

Guest
AI can now generate CD-quality music from text, and it\'s only getting better
https://arstechnica.com/information-technology/2023/09/ai-can-now-generate-cd-quality-music-from-text-and-its-only-getting-better/
Musicians: Speak now or forever hold your beats.

Direct link with examples:
https://stability.ai/research/stable-audio-efficient-timing-latent-diffusion

for what it is worth ;-(
 
On Thursday, 14 September 2023 at 07:17:34 UTC+2, Jan Panteltje wrote:
AI can now generate CD-quality music from text, and it\'s only getting better
https://arstechnica.com/information-technology/2023/09/ai-can-now-generate-cd-quality-music-from-text-and-its-only-getting-better/
Musicians: Speak now or forever hold your beats.

Direct link with examples:
https://stability.ai/research/stable-audio-efficient-timing-latent-diffusion

for what it is worth ;-(
Click to expand...

AI is marketing fake to attract poor guys into delusional activities
 
On Thursday, September 14, 2023 at 11:19:11 PM UTC+10, a a wrote:
On Thursday, 14 September 2023 at 07:17:34 UTC+2, Jan Panteltje wrote:
AI can now generate CD-quality music from text, and it\'s only getting better
https://arstechnica.com/information-technology/2023/09/ai-can-now-generate-cd-quality-music-from-text-and-its-only-getting-better/
Musicians: Speak now or forever hold your beats.

Direct link with examples:
https://stability.ai/research/stable-audio-efficient-timing-latent-diffusion

for what it is worth ;-(

AI is marketing fake to attract poor guys into delusional activities
Click to expand...

Christopher Longuett-Higgins worked out the rules back in the 1970s.

https://en.wikipedia.org/wiki/Christopher_Longuet-Higgins

The wikipedia biography doesn\'t bring out that side of his story too well, but in the late 1970\'s I was working at EMI Central Research which had recently published a paper saying that it was impossible to go from frequency information to musical notation, and the researchers were a bit peeved when he published a paper that spelled out the rules that let you do it. A few years later he published similar rules for extracting musical timing. The intelligence he applied was entirely natural, but he\'d moved from theoretical chemistry to cognitive science by then, so I suppose he was part of the artificial intelligence community. He was a colleague of my wife, at the time, and I did get to meet him from tine to time.

--
Bill Sloman, Sydney
 
My test for AI is to do my taxes. Give it some 1099s, W2s and other information and have it go to the IRS site read the instructions, read the tax code, find the correct forms and fill them out correctly. Not run a program written by a human who understood the tax code and read the instructions, but figure it out for itself.

Right now I think you could ask the AI a question about the tax code and it would regurgitate incomplete information from the tax instructions and tax code and give you answer that sounds good but might be completely wrong. ChatGPT is a glorified Eliza.
 
On Friday, September 15, 2023 at 12:35:05 AM UTC+10, Wanderer wrote:
My test for AI is to do my taxes. Give it some 1099s, W2s and other information and have it go to the IRS site read the instructions, read the tax code, find the correct forms and fill them out correctly. Not run a program written by a human who understood the tax code and read the instructions, but figure it out for itself.

Right now I think you could ask the AI a question about the tax code and it would regurgitate incomplete information from the tax instructions and tax code and give you answer that sounds good but might be completely wrong. ChatGPT is a glorified Eliza.
Click to expand...

But we weren\'t talking about ChatGPT, but a rather more specific problem centred on musical notation.

Idle people want artificial intelligence to do their thinking for them, while it takes real intelligence to take a problem apart in a way that admits any solution.
For tricky problems, you can patent the solution.

ChatGPT doesn\'t do that, but rather searches a huge volume of text of uncertain accuracy to find strings of text that have been served up as solutions in the past.

The text searched presumably includes all of Donald Trump\'s public statements.

--
Bill Sloman, Sydney
 
On September 14, Anthony William Sloman wrote:
Idle people want artificial intelligence to do their thinking for them, while it takes real intelligence
to take a problem apart in a way that admits any solution.
For tricky problems, you can patent the solution.
Click to expand...

ChatGirl has not yet exhibited ingenuity.

ChatGPT doesn\'t do that, but rather searches a huge volume of text of uncertain accuracy
to find strings of text that have been served up as solutions in the past.
Click to expand...

The most interesting question is whether this characterizes human
so-called intelligence, especially infant learning. How do babies
learn? Not by listening to lectures. They absorb continuous streams
of information, and somehow make sense of it. How? Probably by
encoding correlations and repeatable time sequences. Isn\'t that
what the artificial neural nets do?

So it\'s possible these mundane statistical algorithms might
produce the Ultimate Brain -

--
Rich
 
Jan Panteltje wrote:
> CD-quality music from text,

midi2txt and txt2midi are very old software. If the old Soundfont
formats were \"CD quality\" then that is old, too. Subsequent
improvements on Soundfont emerged.

--
Les Cargill
 
RichD wrote:
On September 14, Anthony William Sloman wrote:
Idle people want artificial intelligence to do their thinking for them, while it takes real intelligence
to take a problem apart in a way that admits any solution.
For tricky problems, you can patent the solution.

ChatGirl has not yet exhibited ingenuity.

ChatGPT doesn\'t do that, but rather searches a huge volume of text of uncertain accuracy
to find strings of text that have been served up as solutions in the past.

The most interesting question is whether this characterizes human
so-called intelligence, especially infant learning. How do babies
learn? Not by listening to lectures. They absorb continuous streams
of information, and somehow make sense of it. How? Probably by
encoding correlations and repeatable time sequences. Isn\'t that
what the artificial neural nets do?

So it\'s possible these mundane statistical algorithms might
produce the Ultimate Brain -

--
Rich
Click to expand...

John Searle said pile of gear lacks the ability to be a \"philosophical
subject\"; a subject as defined by philosophy.

Nothing can be subjective to an AI.

I like the way Adam Savage from Mythbusters put it better - AI has
no point of view.

--
Les Cargill
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top