Flood of plugins ?!?

[seablade]

@x42

Not necessarily. Claude is expensive. I know people here in the Berlin audio community who easily spend 1k€ on it every month. And unless you know what you want and have at least a basic understanding of Sound, DSP and ears to check, the result is crap.

And even then it isn’t always perfect. I did create a basic ear training app for my students, and there is still a particular issue that pops up when one of my student starts a quiz of 5 quesitons, the first one seems to misbehave, but I haven’t yet been able to track it down for instance. From a vibe code perspective, this requires an understanding of not only he results, but also how to get the results and valid tests to confirm the results, and a knowledge of how to describe those so that an LLM gets appropriate instruction.

So there is definitely still knowledge needed. As I have mentioned elsewhere, LLMs can be good assistants if you are already knowledgeable in the field in question, and this includes programming. Just because they can do something doesn’t mean they can replace the experts at this time, the experts are the ones that get the best results from it I believe.

Seablade

[seablade]

@Michael_Willis

My skill set will be more valuable than ever because either the AI bubble will burst or Jevon’s paradox will reward those of us that understand how computational systems actually work

KInd of like how the marketplace changed with the advent of the personal computer and those that understood how to use it vs those that did not?

Not a perfect example, but I do believe that this is turning into a major disruptive factor that will CHANGE how jobs, not just programming, but many jobs, will move forward? I have seen impacts of this beyond just programming, for instance my work as a production manager at an entertainment venue, I have been able to use it to create some documentation I have been meaning to create for years that requires a lot of gathering of information, and compiling it into spreadsheets/documents. MUCH quicker when using AI appropriately, but I sitll ahve ot go throguh and verify, so not a perfect panacea either but neither am I looking for one.

Seablade

[Majik (Keith Milner)]

@seablade

And even then it isn’t always perfect.

the word “always” here is bearing a lot of weight. In my experience, on coding projects of any significant size, it’s never perfect. Sometimes it’s good enough.

And it requires constant supervision else it goes off on it’s own track.

Cheers,

Keith

[seablade]

@Majik

the word “always” here is bearing a lot of weight. In my experience, on coding projects of any significant size, it’s never perfect. Sometimes it’s good enough

Well I can’t really argue that, but I was more referring to, it sometimes misses the point entirely and you have to rope it back in:)

   Seablade

[mike-acmtaudio]

@Majik

And it requires constant supervision else it goes off on it’s own track.

Almost always the time you save by getting a computer to do something for you is spent watching it to make sure it doesn’t do it wrong (That was as true in the case of “Why did it suddenly decided to print out all the pages wrong” as it is now in (detailed) review of AI generated code).

@seablade

And even then it isn’t always perfect.

I’ve been trying out a few AI generated DAWs, and they look impressive, and at some level they are, but when you get into actually using them, it becomes apparent that the deeper architecture and behind the scenes ‘plumbing’ of a complex realtime application doesn’t always translate reliably. I’ve had a few weird ‘I don’t know why’ crashes - which is, as we say in the audio industry, bad. (That sort of thing wouldn’t go down well in a professional environment. No software is perfect by any means, but I’ve been in situations where you get one shot at it, and if something blows up you’re not getting invited back into the room, or possibly even the building)

[seablade]

@mike-acmtaudio

I’ve been trying out a few AI generated DAWs, and they look impressive, and at some level they are, but when you get into actually using them, it becomes apparent that the deeper architecture and behind the scenes ‘plumbing’ of a complex realtime application doesn’t always translate reliably. I’ve had a few weird ‘I don’t know why’ crashes - which is, as we say in the audio industry, bad. (That sort of thing wouldn’t go down well in a professional environment. No software is perfect by any means, but I’ve been in situations where you get one shot at it, and if something blows up you’re not getting invited back into the room, or possibly even the building)

Yea that is kind of my point though about AI/LLMs being an assistant. You really do truly need ot understand the topic before it can be useful already, as opposed to just telling the AI to ‘do it’. That is what I find is commonly misunderstood, AI doesn’t replace the need to be able to do it yourself, it just may take some of the grunt work from you.

I can’t say for certain that this is ‘right’ or not in using AI, but what I am finding myself doing when I deal with this, is to try to force it as much as possible to either use technology I am already familiar with, aka languages, toolsets, libraries, etc. so that I can read the output and be fairly confident in my ability, or to use those that I know others have made to solve the problems I am having. Ie. I am looking into Faust right now for DSP, because I trust those that have written Faust to know more about realtime programming than I understand, or I am forcing it to use existing authentication methods so that I don’t screw up the implementation of them myself. But even in all that, it does still make mistakes, and I need to understand enough of the topic ot catch them, and understand why they are mistakes in order to fix them.

But all that being said, your line about

@mike-acmtaudio

and they look impressive, and at some level they are, but when you get into actually using them, it becomes apparent that the deeper architecture and behind the scenes ‘plumbing’ of a complex realtime application doesn’t always translate reliably

Is exactly why we are likely to get so much slop as well. You already see evidence of this with videos (And other content) online, just because you can use AI to create, doesn’t make it good, and in order to make it good, you have to ALREADY understand what defines the ‘good’ aspect of things. So really the best people to use it are the ones that are already competent and able to do it all themselves anyways.

Hopefully that brain ramble made some sense anyways.

  Seablade

[amte]

IMO, AI is the illusion for people who need the imagination they know/can every thing! It will not be able to replace professional knowledge, which is not merely a goal-directed project-based quick knowledge, but a lot of deeper learning around the subject by doing and failing. This is what makes a human-professional what he is.

If I may do a little foretelling: I think the flood of fast-food-like superficial digital content (including software, even if they happen to seem to work for the current hour) will lead to a loss of interest/trust in the digital life as took place up to now (see e.g. emerging theories like the Dead Internet theory).

So just maybe, SI (superficial intelligence) is a more accurate acronym?!

Update 21.10.26 20:00
And as the PHP-incident today showed, it is also good in helping eliminating problems which didn’t exist!

[vdh2137]

I just tried SoundCraft and Faderframe.

1. SoundCraft - no plugin GUIs at all
2. Faderframe - after loading Vital i had to power-cycle my laptop because of massive memory leak (32GB in 10 seconds)

Ardour still works fine

[GMaq (Glen MacArthur)]

We’ll see how much of it is maintained, that’s when the professionals show up…

[x42 (Robin Gareus)]

I would assume that those crashes and leaks will can eventually be fixed by LLMs. Faderframe’s code is also incredibly well structured and could potentially allow human contributions. Also just reading the code, many concepts have been copied from Ardour and Tracktion.

@GMaq

We’ll see how much of it is maintained, that’s when the professionals show up…

More likely this appeals more to singer/songwriters and casual musicians or bands. Not professional recording studios.

–

A major aspect that Ardour still has an edge on is this amazing community.
You cannot vibe code that

[moonshaadow]

I don’t want to add my voice to the pros or cons of AI.
I share @paul 's opinion on many aspects on the impacts of AI, and yet I myself practice vibe-coding, undoubtedly in an egoist impulse to manage to do things unthinkable for me only 3 months ago.

We can discuss the good or the bad for hours, but the fact that AI is here, that it transforms many professions, and that before the end of this discussion the relevance of the proposed code will have already increased considerably.

Without a very strong political intention, it will not be possible to stop this, and the countries are so economically interdependent on each other, that we can already bet that nothing will change at all.

@Ljuba , to answer the initial problem, namely how to know if it’s vibe-coded or not: unless it’s explicitly said, you won’t be able to know, especially if there was a desire to camouflage it.
Are there any good ones, is it safe? I sincerely believe that AI or not, the risk was already present: among the tens (hundreds?) of the thousands of plugins that already exist, some are bad, some are excellent, some are buggy or very poorly coded: for a typical user who doesn’t know where to look, it’s the lottery. In 25 years of “testing” plugins, I have had a lot of disappointments, but also a lot of favorites.

The only valid questions are: what do I want to accomplish, do I really need this type of plugin to achieve my goals?
Where you place your ethics cursor is important, but it will be very difficult now not to use AI tools in spite of yourself even without being aware of it.
All we want in the end is: to mix or produce music, sound design, podcasts, sound effects, voices… . And people have known how to do this for a long time, so it’s doable!

[Piergi]

@moonshaadow

All we want in the end is: to mix or produce music, sound design, podcasts, sound effects, voices…

In my priorities avoiding the destruction of the only environment where humans can live is also pretty high. And so is living in a nazi-free world. YMMV.

[moonshaadow]

Indeed. That is really the most important point.
But IMO the AI is just a very accelerated continuation of what has already been in place for a long time. Our behavior was already problematic, and there is no political will to change anything.
Initiatives to reverse the trend are legion, there are more and more people involved in eco-citizenship, and yet their actions are a drop in the bucket in front of the magnitude of capitalist mechanisms to increase yields.

You probably understood very well that I was talking here about the use of software, and therefore obviously, electronic equipment whose social and environmental cost is already enormous and well documented. So yes, obviously we have to be even more vigilant about it and look at AI for what it is: a new tool of mass destruction. This is not new: humans are omnivorous AND autophagous.

[moonshaadow]

Oh and just for example (but that doesn’t define the truth, it’s still the lottery) :
First plugin tested on the site linuxdaw.org : very nice plugin, with very promising features. Not to mention the dev’s website which is a visual work of art, and offers the plugin as being “the brain of your studio”. An extremely low CPU consumption, in short everything to please. But very quickly I was dealing with undesirable results, and three parameters later, a Carla crash.
So for the joke: replacing your brain is dangerous!