The “it” that is being described and the inevitability at some point becomes theological AGI claptrap alongside the general public’s conflation of LLMs with NLP with broad ML, alongside the awful billionaires pushing dogshit alongside the world that is pretty crappy for most people and pushes them to daydream that makes these conversations so contentious.
Are SAAS subscription services repackaging chatGPT into audio programs? Of course.
Are audio developers that always used ML calling themselves “AI” to get funding and bells and whistles? Sure
Will ChatGPT generate me an on the fly convolution response? No.
Will ChatGPT make that any easier than it was before? Not really.
Will ChatGPT let some people modify an audio plugin template and customize? Sure, why not!
Some stuff to me isn’t worth thinking about because you still will need to put hard work into it, and if a person doesn’t want to understand the underlying concepts or become an autodidact, GPT will not get you there.
If you do, you will understand the limits of stochastic parrots at some point and the nuances will come out, a person will not be able to maintain the promise unless they abandon any homework outside of what they’re told is real.
A small stream then. Maybe I just meant the inevitable “homemade fx plugins” thread on here.
It’s creates a new branch of pseudo developers reliant on LLMs. Fake ass devs you could say.
i mean most of the AI users are just casuals
I taught at an iOS/objective-c bootcamp for like a year
most people end up just “borrowing” boiler plate code and forking other people’s github repos and add a lil spice to make it their own
no harm in that, aslong as it’s cited correctly. 90% of stuff was pretty much just an “exercise” in learning and then once they found an area they loved within that they delved far deeper, and followed more traditional learning practices. I don’t see it as any different.
Wield AI’s magic, and use it to your advantage but don’t rely on it as a crutch. You still need to dig into topics that are barely spoken about or not at all online to really push forward
And again, very much not speaking of anyone here but most of the pushiest AI peddlers were NFT tokenbros a year or two back. I was going to “have fun staying poor” and they’re in fits of ecstasy over Grok and how it can write audiobooks to sell on the Kindle store or whatever Gary Vee hustle culture du jour.
Exactly. As an old they taught us to “learn to code” with Basic, with a turtle writing a vector, with basic loops and you could murderize persons with the print manuals which were outdated pretty regularly.
Tools are tools but the more promise you see outside yourself, the harder it will be to reconcile the work it takes to make something. Not unique to AI, but offloading deep thinking for projects (over specific tasks) comes with costs that you still need to pay to have something that’s useful in the real world.
I saw a Reddit post where someone kid defaulted to chat gpt for the question: how many hours have passed since yesterday when you wake up 7am?
Didn’t even attempt 24+7… it’s game over man.
Sounds like me using Reaktor and taking smarter people’s macros to make my own instruments. But at least I’m learning about signal flow, it’s more fun than being a prompt jockey.
Yeah! And templates are useful for learning Max, Supercollider and certainly DSP, but what are you going to do?
If you’re going to take a basic template and release it as your own, you could already do this.
Oddly enough that works better in hardware packaging, how many guitar pedals use the default demo algos for the Spin Semiconductor FV -1 or most basic implementation of the PT2399 delay chip?
At least with cloning there, you learn aspects of integration and have the funds to tweak shit.
And at some point if the computer does it all, soup to nuts, why is your prompt even necessary? What is your brain adding or connecting with if the connections are already made for you?
Anyhoo, I’ll leave some oxygen for discussion of commercial software
I think a lot of that comes down to the disparity between “generalized AI” = everything will do everything and nobody will ever hurt and also it will think and feel and become my waifu- versus targeted results of “machine learning” which I’ve been for all of my adult life in a myriad of useful, tangible forms, like autocorrect.
AGI is more where things become an unironic, terrible-ecstatic doomcult run by billionaires who drink giardia-tainted water and truly view themselves “techno-shamans” while puppeting around people over the internet into sunning themselves on a nonexistent future.
…Sonic Charge’s patternarium is fun as heck! And doesn’t need to gaslight me!
There are some incredible takes in the comments to Benn Jordan’s latest video. Not only admitting to spamming streaming with AI slop and thinking it’s gonna make them rich, but actually trying to argue that the stuff is legitimately great music - I nearly jumped out a window when some guy posted AI generated Christian rock and proclaimed that anyone who doesn’t like it has no soul. Other classics - “have fun learning stupid instruments”, “You don’t like Suno slop? Skill issue, you just need better prompts.”
Yep, exactly the same register. And I feel like the bubble is going to burst in a very similar way once it becomes clear that the models aren’t getting any better and they don’t actually produce anything of legitimate value. Unless either you’re going to totally rework the output, or you don’t know the difference between good and bad or understand why it matters. Think this piece really nailed the mindset.
As a tool for writing, it’s worse than useless, and anyone with even a little experience of making their living from their ideas and their writing knows that. The fact that Altman doesn’t thus tells us something very important: the guy has never meaningfully interacted with any kind of worthwhile literature in a serious way. He thinks about literature in the same kind of way that a bourgeois family thinks about a Thomas Kinkade painting: it’s something to tie the room together. And when you think of art in that way, automating it is natural: vaguely pretty artistic slop is, after all, just a commodity in this worldview.
software is not dead. The biggest names have been producing awful work but to me, that just means it’s the perfect time for indie developers to start with fresh angles.
Analog modelling is always getting better each year with new research and so of course we’ll see better compressors and distortions over time, but in terms of ideas I’ve got an endless list of new processors I’m working on. It takes years to develop plugins though so it’s slow and will be a while until the market sees any of my new stuff. The awful state of the big companies is a godsend though. YouTubers are starved for interesting tech to get excited about.
(My work is focused mostly around software for sound design, with generative features)
YES.
at least i need several classic machines emulations – TB-303 / SH-101 / Pro One / MS-20 / Odyssey / DX7, to name a few.
same for drum machines — i prefer some working LM-1 / LinnDrum / SDS-V emulations, whatever they are under the hood, to spending hours choosing right samples.
LLMs can not hear and would not be able to hear on a good enough level for a long time. Pro audio grade quality is far from being a priority for the LLM world.
It’s ok I just asked chatgpt to 3D print some ears and stuck them on the side of the screen. Then I prompted to “ensure a good taste in sound in the output of the plug-in.”
People have got it really twisted.
We can create software without any coding knowledge is my point. I never said it would be any good. I do think it will get better. I’m not clear how I’ve been duped yet.
