after 25 years wikipedia showed what it truly was created for, by selling the content for training. otherwise - okay, this was a cool project, perhaps we need better. like federated, crypto-signed articles that once collected together, @atproto style, produce the article with notable changes to it.
Their enterprise offering is more for fresh retrieval than training. For training, you can just download the free database dump — one you would inadvertently end up recreating if you were to use their enterprise APIs in a (pre-)training pipeline.
I’m using a harsh allegory to express massive discontent by the fact that someone was catering to user content for 25 years only for this content to become training corpus.
It is perhaps not that Wikipedia in particular been created for this, that much we hope for, but nowadays it seems such public services are best monetised in this way. I have an actual memory from when Wikipedia started and the enthusiasm of millions of people for it.
And no, I’m not alright with the fact so many people contributed effort AND money to this project only for Jimmy to figure how to sell it better to big corpo.
Seems unfair, as it seems unfair to get these downvotes. Like nobody liked the fact MS bought and used all of GitHub to create copilot, so how is this different?
Nobody cares until is too late. And it is also very hard to get it right, given most p2ps eventually become centralized, or depend on a centralized mesh of hosts. Otherwise I totally agree with the statement, not sure whether is practically possible.
What applications are base on this? I mean it sounds super charming and nostalgic to drop a line or two which runs on WinXP, but is this actually useful?
Mostly legacy industrial machines that need some additional software for telemetry, scheduling, automation etc.
These machines are likely to live at least another 10-15 years and even the brand new ones being sold today uses Windows 7.
Modern languages and frameworks proceed and leave these old systems behind, but everything from our infrastructure to manufacturing capacity that exists runs on legacy systems, not modern computers. The cost of replacing the computers is usually more than the machine itself.
Precisely - every reverb is impulse response, lots of other effects are effectively some sort of convolution with neural networks that we otherwise call AI. Arpegiattors are AI and the random jumps between patterns in Ableton are a Markov Chain.
What does Bandcamp really mean? Perhaps sampling others voices and music is barred, not these mini-AIs that are everywhere ?
Yeah, and oscillators ringing together in an FFT choir based on notes from a diffused image is absolutely, totally not an AI, just algorithms. Really, why be so rude, given you understand the math behind it? Obtuse is not a nice word, not something I would say to people at random. Because, you see, back in the day generative grammars were called AI, so were so many other discreet structures which are employed in music generation, sorry production, on an everyday basis.
Algorithmic progression generation IS IN USE for years, sorry you didn't mention, or perhaps you don't listen that much to everyday radio. Markov chains, constraint solvers, and rule-based harmony live in many VSTs... the fact there are so many "experimentors" out dare winding knobs to match a pleasurable pattern, does not change the fact they be 100% ignorant about the 'deux ex machina'.
I'm surrounded by producers having absolutely no clue about the vast amount of actual AI and actual probabilistic algorithms that make their "unique" sounds possible. And all of them are 100% ignorant of what AI means when they say it, because they don't mean a specific thing.
How is this not AI? Or one needs an transformer-based model to call it AI? This whole story did not start an year or two ago, you may be late for history class though. The fact there's been this moving marketing concept of what "AI" actually is, does not change the reality of most modern music (including acoustic) at some point of the production process getting artificially enhanced by honestly super-complex systems that are intelligent enough to do what otherwise would take 20x more effort to get right.
Author fails to recognize the fact that CLI agents make all kind of hoisting easier and fun. Like publishing to CloudFlare Pages which costs close to nothing and now takes seconds, while previously could taker days.
Claude Code does a very decent UI, somehow the text mode is much more attractive. As if there is once in a lifetime opportunity to make the console great again.
Just use blend2d - it is CPU only but it is plenty fast enough. Cache the rasterization to a texture if needed. Alternatively, see blaze by the same author as this article: https://gasiulis.name/parallel-rasterization-on-cpu/
Blend2D is a CPU-only rendering engine, so I don't think it's a fair comparison to ThorVG. If we're talking about CPU rendering, ThorVG is faster than Skia. (no idea about Blend2d) But at high resolutions, CPU rendering has serious limitations anyway. Blend2D is still more of an experimental project that JIT kills the compatiblity and Vello is not yet production-ready and webgpu only. No point of arguing fast today if it's not usable in real-world scenarios.
How JIT kills compatibility if it's only enabled on x86 and aaarch64? You can compile Blend2D without it and it would just work.
So no, it doesn't kill any compatibility - it only shows a different approach.
BTW GPU-only renderers suck, and many renderers that have GPU and CPU engines suck when GPU is not available or have bugs. Strong CPU rendering performance is just necessary for any kind of library if you want true compatibility across various platforms.
I have seen many many times broken rendering on GPU without any ability to switch to CPU. And the biggest problem is that more exotic HW you run it on, less chance that somebody would be able to fix it (talking about GPUs).
reply