I find Ardour already a very mature and feature-rich DAW, I am 100% for a slower, but carefully thought and assessed human-development. Also something you already pointed out, and mostly comes very short in such discussions, is the ethical question of becoming part of a not-really-totally-positive technology which, as you said, so far shows mostly signs of a rather worse future for humanity (it’s costs, huge water-consumption where even in Germany water is becoming shorter and shorter due to climate-change, speeding-up the climate crisis, and supporting the big-tech monopoly of non-ethical companies).
Rhett Shull just published a very insightful and interesting video about AI. His thoughts and comments are very relevant to this discussion.
I had a couple of thoughts about the energy/environment impact.
-
If it takes 2 developers working for 1 year to produce the next version of a piece of software, or with AI it takes 1 developer 1 month to do the same thing, which uses less energy?
-
Inefficient software wastes CPU cycles which costs energy. If an AI tool finds and solves that inefficiency the saving of CPU cycles multiplied by X number of users over X number of years could save more energy than it took to make the fix.
Do you really think the unemployed developer in that scenario will use less energy ?
Will it ever be worth the training cost (not only environmental) of the LLM though ? Besides this question would only make sense if LLMs were used solely to save CPU cycles, which they are clearly not.
Good point. It takes a lot of energy to keep your computer on for a year.
That’s interesting!
I very much liked the the aspect of keeping old hardware but updating the firmware/software instead.
What probably will happen that creative people will stop producing commercial software. No more progress here.
So, could this boost open software plus associated business models ?
Software business model used to be protected by the fact that it is expensive and slow to develop software. It used to take real effort, resources and money to be competitive in that marketplace.
Now this barrier is gone and there will be wild competition in every software category and it will drive prices down near zero. The only model that probably survives is where the software runs on a cloud server and you get access to its GUI with a browser only if you pay for a subscription. Companies that make hardware will be safe.
And there will be a flood of open source software.
Here is a news story that tells that Microsoft is planning to lay off a lot of people: Microsoft’s 60,000-person buyout offer hints at a future with fewer workers and more equipment | Fortune
Why are you so willfully blind to the fact that the only reason it is cost effective to use generative AI is because we are in the early stages (1 and / or 2) of the inevitable enhsittification cycle, where the cost to the users (and latterly businesses) is heavily subsidised. Eventually when we get to stage 3, the investors will want to see a return on their investment, the enshittifactory business pressures will kick in and the cost will go up. At that point giving software away for free becomes unsustainable.
I didn’t mean to tick anybody off with my view of the future, sorry 'bout that, Sure big AI companies are not profitable now, but they are not the only ones giving access to AI - models. The open source models are also quite good at developing software and we will still have access to them even if some of the big companies won’t survive.
Also hardware and algorithmic evolution will not stop and they will drive costs and energy consumption down. This has happened many times before in computer history and I cannot see why it would not happens this time also.
I believe we will mostly use AI - models on our own hardware in a couple of years time. One does not need the latest and greatest thing to get things done, smaller models will do just fine.
I think that’s a bit naive.
-
If you can actually get the hardware (I’m being facetious but actually, component scarcity is a problem, and to some extent plays into the hands of the very companies creating the scarcity because we’ll be pushed more towards (their) cloud compute model rather than a local compute model. I’m not going to spend a fortune on PC parts just so I can keep an obsolete pedal that cost a fraction of that price working)
-
You don’t need the latest greatest models, but if you’re running on your own hardware at a couple of tokens a second, then i guess whether it’s practical just depends how much time you have.
While I’m not disputing that open source can be a good thing, I also worry about reverse engineer anything - if you modify a guitar pedal as in Rhett’s example and it goes badly, that’s of relatively little consequence, and you’ve just learned a valuable life lesson about titting around with something you don’t understand. But, I wouldn’t want someone to reverse engineer a safety critical software component of their car for example.
The barrier you describe is only gone for producing low quality software. For the rest AI only really fills the implementation gap (writing the code), you still need a competent program who understands how to create, plan, and manage a large codebase.
It seems I hit a nerve here, so I will back off.
I’m currently running Qwen 3.8 Next Flash on my own hardware (RTX 3090) for about 90 tokens a second. There is nothing naive about that, it’s just a fact.
You are right that you need a software engineer to get AI to create good software. As I’ve written in another topic my brother is a senior software engineer (20 years or more coding experience). And the company he is working in has endorsed AI. My brother tells me that he is now able to create software much much faster and be more productive. Even so that they did once run out of features to implement as another team is responsible for coming up with new ones. His jobs is to guide and prompt AI to get things done according to company standards.
So I’m speaking from experience, AI is not just for creating subpar / toy things. As with any instrument it depends of the persons knowledge and talent “playing” it if you get good or bad results out. Humans will still design and architect software and AI will many times be used to create the code according to strict rules so that it is up to the standard,
Anyway this is my final comment on this topic, 'cause I don’t want to heat things up here. I guess this thread has served its purpose and all viewpoint has been made. I respect opposite views also, just wanted to throw mine into the discussion getting a bit gloomy at times ![]()
But that’s not the only energy or environmental cost of AI. This is a little reminiscent of the ‘adjusted profits before EBITDA’ used to justify that AI is profitable (but only if you exclude the cost of things you have to pay for). It’s environmentally better, but only if you exclude the things that make it environmentally worse.
I’ve written a book and a sophisticate book promo website using Claude over the last 18 months. The productivity boost is sensational. I’d encourage trying to figure out the ethical and legal concerns, because I can’t imagine going back to an unassisted workflow.
I’m not sure what your point is. Best seller https://www.simonandschuster.com/books/Incorruptible/Eric-Ries/9798893311860 was also written with AI assistance. That doesn’t mean the author did nothing. It means the author was able to tell their story with remarkable eloquence.
An MCP server is still AI integration. Is is what it is. That project doesn’t attack people, in fact they go out of their way to ban anyone caught harassing devs instead of simply being an advocate for keeping AI out of FOSS. I imagine LMMS is listed as an alternative because AI has infested so many corners of the tech world it’s near impossible to find a software free of it anymore. Hence, the reason that project exists.
hmmm…
(I don’t mind being a little rude when it comes to these subjects)
You know exactly what’s his point.
You just like to think you archived something…guess what, you didn’t, Claude did.
An acquaintance of mine thinks he has started to create cool music…guess what, he hasn’t, Suno did.
I’m kind of allergic to this fake authorship (and we’re in the middle of its epidemic) as you can all tell…brings up the worst in me
.
In the case of that bestseller, who exactly had that profound style, who exactly was such an elegant storyteller?
Call me rude, call me primitive, i don’t mind, i’m still wholeheartedly against it.
When I submitted a patch correcting the record on Ardour, the project maintainer argued about MCP till he (or she) was blue in the face, closed the merge request, and then proceeded to thrash Ardour devs, knowing that I wouldn’t be able to respond to comments in a closed ticket.
Good luck finding LLM-generated code in Qtractor ![]()
I’m sorry, but I’ve been around long enough to recognize the lazy approach to research. It’s just like with all those sloppy “10 audio editors for Linux” posts that somehow end up featuring LMMS (which doesn’t even record or edit audio) and Rosegarden.