Browser ports
category: residue [glöplog]
Curious about what Preacher meant because I don't think this is what he meant and now I am wondering if it's what he meant and I don't want to extrude what he said.
Quote:
[...]how would you feel about the same result (one click on a link to watch an old release in a browser) if it was done with an emulator running in a browser?[...]
Personally I don't see much of a difference
If you only care about the end-result. I think a good analogy would be seeing an amazing painting in real-life (original) vs seeing a step-by-step accurate recreation using the tools and process of the time (emulator) vs seeing a recreation using modern technology (for the sake of this analogy a high-resolution uv printer recreating the texture of the brush strokes). The emulator (ideally) runs the original code while the other only recreates the output. In my opinion these ports are equal to a lossless capture, but inferior to an emulator running the original code. An original painting is worth more than a manual recreation which is worth more than a high-quality print.
Furthermore I find feeding AI models (demoscene) code and have/help it figure it out is ethically questionable considering it's all fed back into the training-set/weights. But I can already hear lamers saying "its inevitable", "everything is a remix" and "no code is that original" 🙄
Quote:
I guess this is sarcasm? :9 In any case, I asume that none of the demoports in this thread are human made.
Not entirely. :9
And, in any case, it is far closer to not being sarcasm at all than to being only half-sarcastic. ;]
I am a very straightforward person, and when I say I take my hat off to you, I mean it sincerely. And while I understand that, given the current climate, you might not want to admit to creating or being involved in the development of these programmed artifacts, there is no need to pretend to me that your work was trivial or that publishing these little things is nothing more than a mere curiosity.
It is a shame that one can no longer speak freely here on pouët, due to the irrationality and bad faith of certain individuals who poison every thread and twist every word that is said.
¯\_(ツ)_/¯
Quote:
Furthermore I find feeding AI models (demoscene) code and have/help it figure it out is ethically questionable considering it's all fed back into the training-set/weights.
I largely agree.
However, for me, the key lies in maintaining control over the tools we use and not depending on anything in the cloud.
That’s why we should use only local AI and even train it ourselves.
…yet you maniacally applaud literally every half-assed brainfart that used any sort of AI…
Yeah, it’s all those evil “biocentric haters”. Luckily we have champions of free speech like you, Tomkh, Elon Musk, etc.
Quote:
It is a shame that one can no longer speak freely here on pouët, due to the irrationality and bad faith of certain individuals who poison every thread and twist every word that is said.
Yeah, it’s all those evil “biocentric haters”. Luckily we have champions of free speech like you, Tomkh, Elon Musk, etc.
Quote:
Not entirely. :9
Well, thanks then :). But I would indeed, not view this as super involved or skilled work. It's usually an idle side task while I am working on something else.
Discourse regarding AI is indeed difficult here - relatable though.
Quote:
ethically questionable considering it's all fed back into the training-set/weights.
I can be the lamer to say that the current models are getting incredibly good in creating original demos. Porting rather holds them back.
The new model released by ant yesterday really shocked me by how much better it became in creating demo for the RP2350 (a microcontroller!). It created an 8k intro and a demo from scratch and with little more guidance other than "beat the other demos in the repo".
What to make of this is an uncomfortable question. But it is there...
Quote:
In my opinion these ports are equal to a lossless capture, but inferior to an emulator running the original code.
This.
Quote:
I can be the lamer to say that the current models are getting incredibly good in creating original demos.
Even if it's true, building demos is a sport. Using AI feels like doping to me.
the less layers the code has to go through the better
Quote:
An original painting is worth more than a manual recreation which is worth more than a high-quality print.
Fully with you here, but there's the part where if I can't fly to the Louvre (don't have an amiga) and don't want to look at a shitty photograph (youtube) I'm personally ok with a perfect recreation in either form in the middle but for example do prefer a standalone sub 64k binary for a 64k intro on windows over an emulation.
Quote:
Furthermore I find feeding AI models (demoscene) code and have/help it figure it out is ethically questionable considering it's all fed back into the training-set/weights. But I can already hear lamers saying "its inevitable", "everything is a remix" and "no code is that original" 🙄
Here I need to go back to my original assumption: there's no such thing as closed-source software anymore and now that you say it I wouldn't put it past the greedy ai companies to suddenly "find" new training data in the form of binary software. As a matter of fact Elon is already on a similar train of thought.
Wait, there’s people here in 2026 who fondly call that ziegheiling fascist clown piece of shit by his name “Elon” and link to his verbal diarrheas? Well fuck me sideways.
Quote:
What to make of this is an uncomfortable question. But it is there...
What would that question be?
For me demoscene productions are about flexing skills (coding, artistic, scientific), LLMs have no place here, just like a chess computer has no place in a chess tournament, a calculator has no place in a mental calculation contest and so on. One might use those tools to aid acquiring said skills / knowledge outside the competitions but that's an entirely different process from prompting an LLM to create a prod, and even then, experience and the ability to connect and willingness to challenge concepts is what yields actual insights, skill development and eventually/potentially discoveries.
Thus in my eyes LLM prods are worthless and I'd vote a "grey screen with no music" over a category defying LLM prod. Except for fast compos and live coding there's no way to stop people from cheating (using LLMs or other means), all we can rely on is having a code of honor.
Quote:
Wait, there’s people here in 2026 who fondly call that ziegheiling fascist clown piece of shit by his name “Elon” and link to his verbal diarrheas? Well fuck me sideways.
Musk is taken
Quote:
prefer a standalone sub 64k binary for a 64k intro on windows over an emulation.
I understand and am with you most of the time, but for some prods I do derive additional satisfaction from actually seeing it run, rather than just seeing its output even if the output is 100% identical.
Quote:
Here I need to go back to my original assumption: there's no such thing as closed-source software anymore and now that you say it I wouldn't put it past the greedy ai companies to suddenly "find" new training data in the form of binary software.
Well yeah because somehow "the world" (aka people with financial leverage and interest in the matter) decided to not give a fuck about the whole concept of copyright anymore, which imho is a terrible crime against human culture and society.
And my point was that I wouldn't put it past these people to mine binaries for content next, especially since they're already talking about shit like this.
Neither would I, but that doesn't negate the ethical dilemma of aiding that process :)
Most of the current progress in math and code is based on reinforcement learning, basically trial and error vs. a verifier that tests the quality of the output. That is f.e. how the models learn to create perfect 3d models only in text.
No "stolen data" needed for that. I think that angle is a bit of a stale discussion point by now, especially for text and code.
No "stolen data" needed for that. I think that angle is a bit of a stale discussion point by now, especially for text and code.
But without doubt, they learn from your interactions with their models. And with reinforcement learning its not even necessary to get clear text session feedback, just normal session quality telemetry will probably already provide a lot of useful information to them.
I am suspecting that even "legitimate interest" data as defined in GDPR provides a lot of entropy for model improvement. (I have been considering to build a demonstration for that, but oh so many side projects...)
I am suspecting that even "legitimate interest" data as defined in GDPR provides a lot of entropy for model improvement. (I have been considering to build a demonstration for that, but oh so many side projects...)
Kindly asking to stick to topic.
Quote:
Curious: how would you feel about the same result (one click on a link to watch an old release in a browser) if it was done with an emulator running in a browser? (Random example although a game)
That's how I watch most of my oldschool demos (in an emulator), except the C64 stuff that I watch on my real machine. I don't care if the emulator runs in a browser or as a normal binary.
On a logical level, I understand that there's not much of a distinction. I can try to convince my brain that this is how it would look like on an Amiga or whatever, but on an emotional level it still feels like a facsimile to me. A fake copy. To use a guitar analogy, there are brand new "reliced" instruments that are made to look old and worn. And I get it that one of those guitars will play just fine and feel nice (have tried out some that were excellent), but I prefer my things to be real for some definition of real that I cannot define exactly. The knowledge that the emulator is running the original binary is meaningful to me even if the output is exactly the same. The knowledge that an actual person is singing the vocals in a song is meaningful to me even if Suno would sound exactly the same. The knowledge that an actual person wrote the article or the book. And so on. And I want my demo experiences to be meaninful.
genuine
Quote:
Kindly asking to stick to topic.
Oh shit sorry, someone use the lie machine to make a shit JS port of Xorgasm.
Lol fuck AI.
Less than zero respect to all the meat proxies vomiting their slop here. These JS ports are nothing but a fuckin’ insult to the scene.
Less than zero respect to all the meat proxies vomiting their slop here. These JS ports are nothing but a fuckin’ insult to the scene.
And here I was wondering what uncle-x thinks of all this, because I've never heard his opinion about AI
Waiting for a demo called „Lol fuck AI“. ;o)
