If AI replaces musicians, does the entire plugin industry die with them?

[ghettosynth]

Found that two pages back:

Prince sang it using a Sennheiser MD441, a microphone I recommend for your voice.

Okay, I just looked up the mic. $1,249. LMAO. He is on drugs if he thinks I'm spending that much on a mic.

I should probably keep this to myself, but, I get a bit of a performance boner when I see those. I totally want one, it will be well suited to my voice, but I do not need one. It cannot be justified, like nearly all of the music gear that I buy. My god, don't ever let there be a stunning sale on one of these, it will come home with me and I might need a moment or two alone.

[BONES]

To say Prince made just mass market pap tells me you don't not know a lot about him, you just know his Top 20 songs (who cares about Top 20 songs, when his gems could be found on B-sides of Singles). I managed Prince's super hit song Kiss, do you really think it is my fav Prince song, it was cool and new and "overplayed" on radio and clubs. He made a lot of releases far from pop, like experimental, jazz, hard blues rock,latin and without his name on an album (pseodonym).

All of which is "pap" to my ears. It's rubbish, it doesn't connect with me in any way at all.

He is considered one of the best artists of all time, some call him Mozart of 20th century or genius like Miles Davis, Stevie Wonder, Lenny Kravitz, D'Angelo, Eddie Kramer many more, some even consider him one the best guitarists ever and some of the best musicians of music history joined him and went to cold Minneapolis like Larry Graham (Sly Sone), Maceo Parker (James Brown), Sonny Thompson (local idol in Mnpls), Sheila Escovedo (Stevie Wondee/mj), George Clinton (Parliament Funkadelic), MonoNeon (bassist), John Blackwell (drummer Cameo), Chaka Khan etc.

All people for whom I have little or no respect. They all make boring krap. They all go into the same basket as Madonna or Kylie Minogue - completely disposable.

The movie Purple Rain turned Eric Clapton's life from a hard addicted drug junkie into a musician with new energy and creativity etc etc etc

So what? Once a junkie, always a junkie and zero respect from me.

I think he might have saved my life, too, cuz when I heard Prince the first time, I was into guitar playing and learning a lot and my main inspiration was Hendrix and that was the worst inspiration one could have

How could you be so weak-minded as to let yourself be influenced by that? I listen to doom-laden music all the time but I'm the happiest idiot on the planet.

Let's not forget that two of the most successful producers Jimmy Jam and Terry Lewis

I can't say I've heard of either. I'm sure they're not Bob Clearmountain, Flood or Steve Lillywhite. They're probably not even John Fryer, Connie Plank, Mike Howlett or Martin Hannett.

You seem to feel that popularity is a measure of quality but the most popular things are rarely the best. As I said, it's about taste. A song either works for you or it doesn't and most popular music, be it Pop, Rock, Blues or Jazz, does not work for me at all, on any level.

[DCrown]

Found that two pages back:

Prince sang it using a Sennheiser MD441, a microphone I recommend for your voice.

Thanks Bert. Now I'd like to know why he recommends that mic. Unfortunately, only he can answer that.

It is called experience. I tested or could use a bit more than 20 mics in my life. I have listened to some of your tracks as u know and the voice is the main part that needs to be improved not just the singing quality. The importance of a mic is often underrated.
A dynamic mic is a better choice than a condenser for your voice timbre, singing style and quality imo
Now you could say that only condenser mics are/were used in prof studios.
Well, prof studios have a better room treatment than most homestudio, for a homestudio with no good acoustics a Neumann condenser could be the worst choice anyway and there are many examples of popular and good songs where a dynamic mic was used, MJ used dynamic mics on songs of Thriller album for example.

[DCrown]

"I can't say I've heard of either. I'm sure they're not Bob Clearmountain, Flood or Steve Lillywhite. They're probably not even John Fryer, Connie Plank, Mike Howlett or Martin Hannett.

You seem to feel that popularity is a measure of quality but the most popular things are rarely the best. As I said, it's about taste. A song either works for you or it doesn't and most popular music, be it Pop, Rock, Blues or Jazz, does not work for me at all, on any level."

I am not even surprised about your answer,
there's seems to be a kind of arrogance by considering Minneapolis somewhere close to end of the world or close to nowhere, whereas it has one of the best studios in world (Paisley Park, Flyte Time Studios, even a picky Michael Jackson went there) and has established as a hot spot for music productions and music scenes.

And you completely contradict yourself by saying
"You seem to feel that popularity is a measure of quality but the most popular things are rarely the best." and at the same time you post famous names like Bob Clearmountain, Flood or Steve Lillywhite. Since success in Western civilization is measeared by income/money I am pretty sure Jam and Lewis are way more successful than Steve Lillywhite with countless Top 10 releases in 1980s and 1990s. That's not what I consider success, though.
Your post really is embarrasing and you obviously show that you think LA, New York, Nashville, Detroit and Miami are the only good places were music was produced in USA.
Jam and Terry composed, wrote, arranged, played instruments, mixed and produced songs and you even dare to compare the one-sided mixing engineer talent of a Bob Clearmountain to their skills. You are a joke.

I wrote Prince's best songs are not his known pop songs, his B-sides and even releases without the name Prince on an album, even though a complete Prince album. You obviously also do not seem to know who Miles Davis was, he will make every name you post look very small.
Guess how much value your opinion on music and musicians has compared to Miles Davis' opinion!

Only some names who recorded at Paisley Park: Madonna (recorded "Like a Prayer" and "Justify My Love"), Stevie Wonder, R.E.M., Fine Young Cannibals, Lizzo, Barry Manilow, and Stone Temple Pilots, Cory Wong,
Prince Protegés/Collaborators: Sheila E., Tevin Campbell, Carmen Electra, Mavis Staples, George Clinton, Ingrid Chavez, and Mazarati.

Bob Mould, Soul Asylum, Warren Zevon, and Big Head Todd and the Monsters.
Rehearsals: The facility was also used for rehearsals by artists like The Bee Gees, Neil Young, and Kool and the Gang.

The list is way longer.

The MINNEAPOLIS Sound of 1980s is not just Prince influenced music, Minneapolis sound was/is also indie rock scene.
The Replacements, Hüsker Dü, Soul Asylum, Babes in Toyland, and the Suburbs who had influence on 90s alternative rock and grunge bands like Nirvana.

Some more infos about cold Minneapols:

Bruce Swedien, Lizzo, Bob Dylon, Sonny Thompson and Cory wong are from Minneapolis.
Lipps, Inc. ("Funkytown"), The Jayhawks, Atmosphere, Babes in Toyland, The Jets, too.

At a Minneapolis recording studio, Sound 80, everyone from Bob Dylan to Dave Brubeck to the St. Paul Chamber Orchestra recorded music during the 1970s. And it was at Sound 80 in 1978 where the SPCO made one of the first digital recordings to be released commercially.

Venues
First Avenue & 7th St Entry, Bunker's music bar.

If you want to know where Minneapolis is, search it on a map or use Google.
I give you a hint:
It is somewhere between New York and Seattle!


Is it possible you are influenced and fooled by
the Internet and AI, Bones? Is it possible? haha
Prince saw it all coming already in 1999, it's called wisdom.

And you really think your opinion on Prince matters? Despite of the fact that everyone can like or dislike whatever he wants.






Usually I don't give a d@@@ what majority thinks,
well, that performance obviously shows one of the greatest guitar performances of all time and I am obviously not the only one who agrees.
C'mon Bones, let us know again that you don't like that performance by Prince, let us know cuz your opinion is the most important one! haha
And Bones I know more guitarists than just Prince
like Ingwie Malmsteen, Van Halen, Al di Meola, Gerge Benson, Jimmy Page, Blackmore, Clapton, Joe Pass, Django, Sabicas, Adrian Smith, Rory Gallagher, Steve Vai, Steve Stevenson, Eddie Hazel, Jesse Johnson, BB King etc etc etc etc etc etc

"Use AI, but be aware of the fact that the prize is the soul!"

[enroe]

"Use AI, but be aware of the fact that the prize is the soul!"

Hayeah, I hadn't thought of that analogy before. But that's exactly it, if you
want to sum it up in a single sentence:

---> By using AI, you're selling your soul!

In Goethe's "Faust I," Faust makes a pact with the devil to gain magical
powers and finally achieve his long-sought goals. However, this pact with
the devil plunges him into a maelstrom of personal catastrophes and
completely dehumanizes him.

Look here.

"Faust I" is a great work of world literature. Since its creation in 1806, it
has stood for the idea that the use of flimsy, criminal means, corruption
and falsehoods (a pact with the devil!) ultimately leads to ruin.

And it's the same when you use AI for creative purposes, whether in art
or music: you sell your soul, you haven't actually created anything yourself,
you've just given in to your narcissistic ego - it leads to ruin.

[DCrown]

I read Faust, the first part is great.

I refered to Prince's speech in 1999 in the first video

[eassae]

Remember that time Bones, lead screamer of an AI cover band, gave his opinion on Prince, multi-instrumentalist, musical-genius, as if it mattered…that was funny.

[jancivil]

Here's the piece I referred to regarding Apple unable to get a real Large Reasoning model working (since I'm not good at technical writing):



averse to a YT link? Here's an extract from IMO the best overview on the page (in comments):

[Apple's] previous research shows that if you take the same problem from test datasets and essentially do logical find and replace on it, then run randomized examples against the LLM or LRM with the exact same structure of the problem but different numbers or details, then the LLM shows a drop in performance on that problem. Even though the structure of the problem and the thing you are asking the model to do are exactly the same, it gets the problem itself wrong. This drop in performance shows a) the test datasets somehow find their way into the model’s training datasets (like you said), and b) that these models don’t “reason” or “think”, but that they are (as other researchers point out) stochastic parrots.

[jancivil]

YEAH!
Trying to control a tool that is much smarter then you is not madness, sure!

Don't argue with fools, they'll drag you down to their level and beat you with their experience.

Acknowledging that conclusing sentence, I suppose there is no profit in showing you anything (since you enthusiastically leapt to a conclusion that absurd) but the fool is you.

[vurt]

"Use AI, but be aware of the fact that the prize is the soul!"

Hayeah, I hadn't thought of that analogy before. But that's exactly it, if you
want to sum it up in a single sentence:

---> By using AI, you're selling your soul!

In Goethe's "Faust I," Faust makes a pact with the devil to gain magical
powers and finally achieve his long-sought goals. However, this pact with
the devil plunges him into a maelstrom of personal catastrophes and
completely dehumanizes him.

Look here.

"Faust I" is a great work of world literature. Since its creation in 1806, it
has stood for the idea that the use of flimsy, criminal means, corruption
and falsehoods (a pact with the devil!) ultimately leads to ruin.

And it's the same when you use AI for creative purposes, whether in art
or music: you sell your soul, you haven't actually created anything yourself,
you've just given in to your narcissistic ego - it leads to ruin.

that would be an ecumenical matter.

[pugilistas]

Remember that time Bones, lead screamer of an AI cover band, gave his opinion on Prince, multi-instrumentalist, musical-genius, as if it mattered…that was funny.

The great thing about Bones is that he's such a self absorbed ijit, that all of his 'pronouncements' are easily ignored, or subjected to 'point and laugh'.

[BONES]

I am not even surprised about your answer, there's seems to be a kind of arrogance by considering Minneapolis somewhere close to end of the world or close to nowhere, whereas it has one of the best studios in world (Paisley Park, Flyte Time Studios, even a picky Michael Jackson went there) and has established as a hot spot for music productions and music scenes.

I don't see the relevance of the geographical location. To those of us in the outside world, it's just some city in America. Until you mentioned it, I had been totally unaware that Prince had worked from Minneapolis. I suppose I naturally assume that American music is made in NYC or LA but it's not something I put any thought or effort into. It's irrelevant.

And you completely contradict yourself by saying "You seem to feel that popularity is a measure of quality but the most popular things are rarely the best." and at the same time you post famous names like Bob Clearmountain, Flood or Steve Lillywhite.

Well, I'm not conflating famous with good, either, just pointing out that there are producers I've heard of and those guys aren't any of 'em. Again, that 's probably because they don't work on music that I have any interest in, because it's about taste.

Since success in Western civilization is measeared by income/money I am pretty sure Jam and Lewis are way more successful than Steve Lillywhite with countless Top 10 releases in 1980s and 1990s. That's not what I consider success, though.

Steve Lillywhite was a poor choice on your part. He's won more Grammys than your boys which is another measure of success. But it's not about that, it's who those guys have worked with that makes them known to me.

Your post really is embarrasing and you obviously show that you think LA, New York, Nashville, Detroit and Miami are the only good places were music was produced in USA.

Miami? I had no clue.

Jam and Terry composed, wrote, arranged, played instruments, mixed and produced songs and you even dare to compare the one-sided mixing engineer talent of a Bob Clearmountain to their skills.

The difference is that I have not knowingly heard anything they may have done, anything they've achieved is completely irrelevant and, therefore, easily dismissed. OTOH, I own literally dozens of records produced by Steve Lillywhite. And according to Sound on Sound magazine, Bob Clearmountain has "his name on more hit records than anyone else in the history of popular music". He also did the live mixing for Live Aid which probably makes him just slightly more important to the recent history of music than your guys. But, as I keep trying to get through your thick skull, it's about taste. You know the guys that make the music you like and I know the guys who make the music I like. That you don't seem to understand how much more the producers I've listed have achieved than your guys is more than a little disturbing, if I'm honest. Steve Lillywhite, for example, launched the careers of bands like U2, The Psychedelic Furs and XTC. Working with Hugh Padgham, he invented the gated reverb sound that dominated all the biggest hits of the 80s. That's an influence your guys can't hope to match.

You obviously also do not seem to know who Miles Davis was, he will make every name you post look very small.

Not even close. There is a whole wide world outside your bubble that, like me, couldn't even name a Miles Davis song. He was a trumpet player, wasn't he? (I just looked it up to be sure, he was.)

Guess how much value your opinion on music and musicians has compared to Miles Davis' opinion!

Guess how much I care.

[BONES]

And it's the same when you use AI for creative purposes, whether in art or music: you sell your soul, you haven't actually created anything yourself, you've just given in to your narcissistic ego - it leads to ruin.

So you sell your soul every time you play a song someone else wrote? You know, a cover version. If that's the case, then pretty much every performing artist I can think of has sold their soul at one time or another. Even The Beatles did covers and I can think of plenty of artists who are better known for their covers than their original songs. Is there no soul in the Michael Andrews and Gary Jules version of Tears For Fears' Mad World?



Is Soft Cell's version of Tainted Love completely soulless?




Clearly what you're trying to suggest is absolute nonsense.

[Apple's] previous research shows that if you take the same problem from test datasets and essentially do logical find and replace on it, then run randomized examples against the LLM or LRM with the exact same structure of the problem but different numbers or details, then the LLM shows a drop in performance on that problem. Even though the structure of the problem and the thing you are asking the model to do are exactly the same, it gets the problem itself wrong. This drop in performance shows a) the test datasets somehow find their way into the model’s training datasets (like you said), and b) that these models don’t “reason” or “think”, but that they are (as other researchers point out) stochastic parrots.

Can you explain how this is any different from being taught the wrong thing at school? Because that's exactly what it seems like to me - we all "learn" stuff which we later repeat parrot-like when it comes up in conversation. Only it's even worse with people because we now live in a world where even in the face of overwhelming evidence, people will not change their minds on something they have an emotional investment in. Little wonder that the minds we seek to create suffer from the same flaws, is it? And that's even before you remember that Apple's dalliance with AI has been far from successful, to the point that they've given up and paid Google to put Gemini into iPhones, so it's hard to put much stock into anything they have to say on the subject. It's just another failed appeal to authority.

[ghettosynth]

[Apple's] previous research shows that if you take the same problem from test datasets and essentially do logical find and replace on it, then run randomized examples against the LLM or LRM with the exact same structure of the problem but different numbers or details, then the LLM shows a drop in performance on that problem. Even though the structure of the problem and the thing you are asking the model to do are exactly the same, it gets the problem itself wrong. This drop in performance shows a) the test datasets somehow find their way into the model’s training datasets (like you said), and b) that these models don’t “reason” or “think”, but that they are (as other researchers point out) stochastic parrots.

Can you explain how this is any different from being taught the wrong thing at school? Because that's exactly what it seems like to me - we all "learn" stuff which we later repeat parrot-like when it comes up in conversation. Only it's even worse with people because we now live in a world where even in the face of overwhelming evidence, people will not change their minds on something they have an emotional investment in. Little wonder that the minds we seek to create suffer from the same flaws, is it? And that's even before you remember that Apple's dalliance with AI has been far from successful, to the point that they've given up and paid Google to put Gemini into iPhones, so it's hard to put much stock into anything they have to say on the subject. It's just another failed appeal to authority.

The GSM-Symbolic paper from Apple is genuinely useful work, but the conclusions being drawn from it in online circles dramatically overstate what it actually shows. The core contribution is a framework for generating parameterized variants of math problems, which is a real methodological advance for benchmark design. What it found is pretty much what you'd expect: models perform better on the original GSM8K problems because that data almost certainly leaked into training sets, and performance drops when you substitute different names and numbers. Surprising to no one.

When you control for the contamination by looking only at the generated variants, reasoning models consistently outperform non-reasoning models on both mean accuracy and variance across every difficulty level tested. The paper never shows non-reasoning models matching or beating reasoning models on any condition. So the popular take that this paper somehow demonstrates reasoning models aren't useful is directly contradicted by the data. The NoOp result, where adding irrelevant clauses like "five were smaller than average" causes performance drops, isn't surprising either. However, even there reasoning models hold up better than non-reasoning models, and no model drops to zero. Models are less accurate when you add distracting information. Again, not shocking.

The "can LLMs really understand mathematics" framing in the paper is where things go off the rails. Understanding is never defined, the comparison to a grade school student is analytically meaningless since they're not performing the same operation, and the alarm expressed about variance in a stochastic system is misplaced. The data is solid. The interpretation layered on top of it reaches well beyond what the methodology supports.

In short, researchers don't think that models think. The term "reasoning" is overloaded here. It was never defined as, nor intended to mean, formal mathematical reasoning. In practice, reasoning models explore the problem space statistically. This process is intrinsically stochastic and empirically useful. The industry perceives reasoning as an engineering improvement that demonstrably improves performance on hard problems, not a claim about machine cognition. Empirically testing whether that mechanism produces formal mathematical reasoning is testing for something nobody credible claimed it was.

Again, George Box is our friend. "All models are wrong, some models are useful."

[enroe]

And it's the same when you use AI for creative purposes, whether in art or music: you sell your soul, you haven't actually created anything yourself, you've just given in to your narcissistic ego - it leads to ruin.

So you sell your soul every time you play a song someone else wrote? You know, a cover version. If that's the case, then pretty much every performing artist I can think of has sold their soul at one time or another. Even The Beatles did covers and I can think of plenty of artists who are better known for their covers than their original songs. Is there no soul in the Michael Andrews and Gary Jules version of Tears For Fears' Mad World?

This is NOT about cover versions! It has absolutely nothing to do with that.

But by using an AI, which has no (human) feelings, and referring to it, by
playing such songs, you automatically deprive yourself of this basic
motivation to express something of your own, to perform something you
have created yourself as a work of art or piece of music.

You're naively clinging to a kind of neurally learned, artificial feeling. That
might sound good, but it's no longer human nature at all.

And that's exactly what I mean: By sliding into the hell of AI, you're making
a Faustian pact with the devil.