The past, present and uncertain future of scene-run infrastructure
category: general [glöplog]
I don't know why everyone has to **insist** on having to train AI on every data that is available? Just cut the invasive and unsolicited shit already?
NR4: Did you read any of the above before posting this?
Quote:
There was the plan to create a scene-internal LLM AI that is fed with all scene knowledge there is
Im referencing this exact quote
I am not producing for your LLM. Neither my code nor my resulting products are intended for this. The fact that I have to read the whole earlier essay before even knowing that this was intended to be done, annoys me more than only a bit.
I WILL NOT BUILD A DEMOSCENE LLM. NONE OF THE SCENECITY DEVELOPERS ARE WORKING ON ANY OF THIS. IT WAS AN IDEA, AND A TECHNICAL CONCEPT ON HOW IT COULD BE DONE. NO, NOT EVERY SCENER NOW HAS TO POST THAT HE DOES NOT WANT HIS ART TO BE USED IN AN AI THAT DOES NOT EXIST AND WILL NOT EXIST.
Let's please have "all the hypothetical and imaginary things on this planet that I am absolutely against!!!" lists in a different thread than this :)
Let's please have "all the hypothetical and imaginary things on this planet that I am absolutely against!!!" lists in a different thread than this :)
nr4: As mentioned a couple of times above I should not have used the word "plan". It was an idea, and a technical plan/concept on how that could be implemented. Not "plan" in the meaning of "we have decided to this", a coding plan. Misunderstanding solved?
It is, in general, difficult to get others to buy into your vision.
Even more so when you keep getting in fights with them.
Even more so when you keep getting in fights with them.
thanks for the clarification (in your bold and uppercase statement). The idea is unethical and does not provide a suitable, effective or appropriate solution for any of the (existent) problems your essay lists.
HEY SCAMP YOU'RE WELCOME TO USE ALL THE CRACKTROS I'VE UPLOADED IN YOUR SCENECITY AI
From my perspective the demoscene seems to have way too many sites and services, not too few. At this point building a new service that overlaps with existing functions is like https://xkcd.com/927/ - no harm in trying but be ready for a disappointment if you're hoping to gather all the demosceners there.
Not everything has to be in the same place, but we really don't need 20 different, partially overlapping services on which we "should" upload/discuss content. It just makes the scene seem more quiet than it actually is.
Not everything has to be in the same place, but we really don't need 20 different, partially overlapping services on which we "should" upload/discuss content. It just makes the scene seem more quiet than it actually is.
as someone who is actually a professionally trained historian, i can only say that preserving history by storing it into a LLM is a completely twisted idea. it might work in some sort of science fiction universe, but will be absolutely counterproductive in real use. not only because it might hallucinate, but also because it's a black box. historians don't work with black boxes, they work with primary sources. one of the crucial premises of historical work is confirmability: you see a quotation, you look at the footnote, and can theoretically visit the named archive yourself to check what the primary source really says. this is absolutely impossible if historical knowledge is stored only in an LLM. so, there is absolutely no alternative to actual preservation of digital artefacts, and there are lots of promising and working solutions - not only in the scene, but also in professional archive/museum contexts, to achieve this.
as for the infrastructure, i can share your worries, but i see the solution not in building yet another infrastructure, but in supporting the most open and reliable ones that are around - and these, for me, are demozoo and scene.org. especially the latter is probably the most solid and ulikely-to-fail scene infrastructure, as it's hosted by a university and mirrored by a dozen of servers worldwide - so it should take a global catastrophy to bring scene.org down (and in that case, there will be more important things to carry about than the scene, anyway).
as for the infrastructure, i can share your worries, but i see the solution not in building yet another infrastructure, but in supporting the most open and reliable ones that are around - and these, for me, are demozoo and scene.org. especially the latter is probably the most solid and ulikely-to-fail scene infrastructure, as it's hosted by a university and mirrored by a dozen of servers worldwide - so it should take a global catastrophy to bring scene.org down (and in that case, there will be more important things to carry about than the scene, anyway).
Quote:
2.) Preserve all knowledge that might get deleted in an LLM. In 10 years from now, at least you have a central knowledge backup
3.) Accept that in 10 years time a scene historian will hallucinate about what the scene was like.
4.) Accept that most of our knowledge and heritage will be forgotten
all 3 things can be true at once. LLMs are incredibly good at hallucinating too. There's a lot of scene history/stories that I have only ever been told by a drunk sceners at parties that you wouldn't find online either.
There's plenty of info out there already, and I agree we gotta consolidate it somewhere, but it has to be in searchable and human readable text format, rather than abstracting it away behind a glorified marokv chain bot.
You can always have something fancy like what Kagi Search uses to help summarize data.
Not a perfect solution, mind you, but a lot better than shoving it in a black box and hoping for the best.
dipswitch: I am not sure how many admins scene.org still has. For the last 10 years or so I only had a single contact there - Gargaj. With him clearly being in permanent ultra-destructive mode now, I won't ask.
A lot of sites where taken for granted and regarded "most solid and unlikely-to-fail". Until they disappeared. It's ironic, but it looks like even the "old people" appear to have forgotten how much history has already been... forgotten.
And I don't have any idea on how I could preserve untergrund.net, especially in this general hostile climate the scene is in. It's a huge technical challenge. And maybe I simply should stop worrying about the scene's legacy. Maybe it's just not my job to worry about it. I'll have my own memories about how things really were, maybe I just should not give a damn about what is preserved for others.
And for the few remaining friends I still care about in the scene, I probably can just go to slengpung NOW to download their images should I once again need some for a funeral. Then I no longer have to care about that part of scene history also getting deleted.
The pouet admin trolling in a thread about preserving his legacy?
I guess I need to stop giving a fuck, too.
A lot of sites where taken for granted and regarded "most solid and unlikely-to-fail". Until they disappeared. It's ironic, but it looks like even the "old people" appear to have forgotten how much history has already been... forgotten.
And I don't have any idea on how I could preserve untergrund.net, especially in this general hostile climate the scene is in. It's a huge technical challenge. And maybe I simply should stop worrying about the scene's legacy. Maybe it's just not my job to worry about it. I'll have my own memories about how things really were, maybe I just should not give a damn about what is preserved for others.
And for the few remaining friends I still care about in the scene, I probably can just go to slengpung NOW to download their images should I once again need some for a funeral. Then I no longer have to care about that part of scene history also getting deleted.
The pouet admin trolling in a thread about preserving his legacy?
I guess I need to stop giving a fuck, too.
gee, talk about an uncalled for public witch hunt. This is disgusting, folks.
This is not about scamp, but discussing the options for long-term preservation of demoscene heritage, should decade-long established sites cease to function. Regarding this, scamp's made some valid points worth thinking about.
Makes me wonder, if the backlash would've been the same if OP's text had originated from another handle?
Just my IMHO.
This is not about scamp, but discussing the options for long-term preservation of demoscene heritage, should decade-long established sites cease to function. Regarding this, scamp's made some valid points worth thinking about.
Makes me wonder, if the backlash would've been the same if OP's text had originated from another handle?
Just my IMHO.
Quote:
Makes me wonder, if the backlash would've been the same if OP's text had originated from another handle?
Quote:
uncalled for public witch hunt.
To be honest, the idea with the AI dragged down the whole thread. I agree there were valid points in the original post. My reaction has nothing to do with scamp as OP. My reaction is purely anti-LLM motivated.
@NR4: Understood. Thanks for clearing this up <3
I need to address the Slengpung situation. It's been pretty much dead in the water for quite some time now. I remade it from scratch with current tech and let peeps take a peek of what it could be.
Slengpung is problematic in terms of privacy, but I still think it cloud be done in a way that will allow it to flourish and above all, be available.
The feedback about it was very positive and encouraging. But then there were the comments along the lines of "let it die", "I don't care" etc. While I was not deterred by those comments, they did raise my eyebrows a bit. Those comments were few and far between, but showcased that some people are fine in letting that much history rot and eventually wither away and be lost forever.
The development is paused currently, but I might kick it back into gear after I've had a talk with one of it's founders. Whom which I'm finding it quite difficult to get a hold of.
Slengpung is problematic in terms of privacy, but I still think it cloud be done in a way that will allow it to flourish and above all, be available.
The feedback about it was very positive and encouraging. But then there were the comments along the lines of "let it die", "I don't care" etc. While I was not deterred by those comments, they did raise my eyebrows a bit. Those comments were few and far between, but showcased that some people are fine in letting that much history rot and eventually wither away and be lost forever.
The development is paused currently, but I might kick it back into gear after I've had a talk with one of it's founders. Whom which I'm finding it quite difficult to get a hold of.
@t-101: as i said, please do get in touch (on discord, for example). i'm not a slengpung founder, but a staff member from back in 2002 (and one of the few who supplied/curated content until the very end), so i am very interested in your efforts and would support them in any way possible, incl. reaching out to other staff members.
@t-101 There is a lot of old memories on slengpung. I visit it every now and then and riding the nostalgia wave. Hope it stays online. Or that we at least get a warning before it goes away to save the memories for ourself.
Slengpung is hosted by SceneSat in its current form and is not going anywhere without any notification. I can vouch for that.
I can only second sunspires initial post, maybe focus on the title of the thread rather than some arbitrary example / idea deep down at the end of the original post?
I started writing a more elaborate post pertaining to the original topic a while ago but I find it hard to gauge the need of the many vs personal ideas of independence, the convenience that big corp platforms offer is hard to beat and in a world where there's 1e7 ways to spend every waking second of your existence and the slowly changing yet very heterogeneous demographic in the scene it's hard to argue with people to spend time on navigating the woes of niche, partially archaic platforms.
Btw. fyi. I am working on a sleek, modern platform to consume and comment on demos built on top of the demozoo database, if it'll ever come to fruition / a public release is still uncertain though.
I started writing a more elaborate post pertaining to the original topic a while ago but I find it hard to gauge the need of the many vs personal ideas of independence, the convenience that big corp platforms offer is hard to beat and in a world where there's 1e7 ways to spend every waking second of your existence and the slowly changing yet very heterogeneous demographic in the scene it's hard to argue with people to spend time on navigating the woes of niche, partially archaic platforms.
Btw. fyi. I am working on a sleek, modern platform to consume and comment on demos built on top of the demozoo database, if it'll ever come to fruition / a public release is still uncertain though.
The scene is dead.
Long live the scene!
Quote:
The scene is dead.
The scene has been dying every year since 1988 and has been reborn every year since then.
