recompiled Unreal / Second Reality / Crystal Dream II / Stars
category: residue [glöplog]
Quote:
It runs smooth on my 2019 MacBook Pro (RX5500M). In fact it runs a lot better than on the M2, because it doesn't have to go through Rosetta... So...
I'm happy for you!
Quote:
If you still think it's WINE's fault, how about getting that fixed instead?
Yes, I'll probably do it someday. When I retire and don't know what else to do with my time.
The rules linked by bsp in the onliners look interesting, maybe they would also make sense for the demoscene:
https://eab.abime.net/showthread.php?t=124013
with https://forums.nesdev.org/viewtopic.php?t=26884 being the origin.
https://eab.abime.net/showthread.php?t=124013
with https://forums.nesdev.org/viewtopic.php?t=26884 being the origin.
I don't know which one of the "AI threads" this is, but I'll write something.
So. Like the climate change, this is now happening and we just need to adapt as a society.
(not to shitty companies, but to the fact that these rather capable models exist now and can be executed to achieve pretty complicated things and their HW requirements are basically same as with latest video game entertainment products)
So - in that vein - how do you people think this changes software, making of software and the whole software business, in the immediate / long term future?
Only honest opinions/predictions please.
So. Like the climate change, this is now happening and we just need to adapt as a society.
(not to shitty companies, but to the fact that these rather capable models exist now and can be executed to achieve pretty complicated things and their HW requirements are basically same as with latest video game entertainment products)
So - in that vein - how do you people think this changes software, making of software and the whole software business, in the immediate / long term future?
Only honest opinions/predictions please.
I would be surprised if this turns out to be sustainable. As we all know, it's a heavily subsidized money pit and may crash the economy once the bubble bursts. The tech is fascinating, yes, but the socio-economic and environmental cost is simply too high.
AI will certainly not disappear but it's too early to try to roll it out at this scale, i.e. I think that (for now) it does not belong in the hands of everyone (local LLMs and search engines aside). For it to be useful (fast enough to run the trillion+ parameter top models) you need those data centers and heavy duty equipment (A100 etc), a gaming video card won't do. But the tech isn't there, yet (i.e. not efficient enough), and we also don't have unlimited-energy fusion reactors, yet (maybe in 20 years, har).
Another concern is that we are heading straight into an idiocrazy if everyone just "vibes" and never learns how things actually work.
It somehow works at the moment since the current seniors grew up without AI, gathered knowledge and know exactly what to prompt but the next generation already struggles to find employment because (some) companies don't see worth in them anymore ("they can easily be replaced by an AI" was already said two years ago). In every society there's this "social contract" which does not just mean that the young are simply paying our rent but that the old are also giving the young a fair chance to learn and evolve, and eventually become seniors themselves. Trying to cut them out for short term profits is a major no-go.
With tools like REA (reverse-engineer-all), and the general "everything's up for grabs" attitude that the legally questionable model training on copyrighted material made hugely popular, there are some tough times ahead for many software companies.
If what I heard is correct, some of the big players now offer AI subscriptions that protect their customers from lawsuits.
On the other hand, (some) companies require their employees to sign a document that shifts the liability onto them.
Besides, in the EU, the Product Liability Directive (Directive (EU) 2024/2853), is taking effect on December, 9th), i.e. you are responsible / liable for the software you sell, regardless whether it is (customized) open source, human-written closed source, or AI generated).
Good luck providing a fix for an AI-generated software you don't really understand in detail when the AI fails to resolve the issue.
For non-profit hobby communities, the onslaught of AI generated content has stirred up quite a bit of an uproar, to put it mildly.
Neither the attitude presented by the (radical) AI proponents ("hey, now we can finally get rid of those elitist developers, AI democratizes everything"), nor the one by the anti-AI fraction who slanders everything that remotely looks like AI was somehow involved, is very helpful here.
For some, the ongoing AI content flood has caused them to lose motivation because why invest months or years if the result gets drowned in 'mindless slop', anyway. Not sure this applies to the demoscene (there's no flood of AI prods as far as I can see) but this might change in the future.
The aforementioned AI rules posted on nesdev.org and discussed on EAB sound like a reasonable compromise / "code of conduct", e.g. for demo parties.
Last but not least, the general AI topic has been discussed to death (it's everywhere!) and this thread is about porting old DOS demos to browsers (my opinion on that is that I'd rather watch the original in an emulator *shrug*).
p.s.: ..and in case anyone wonders: I do not use gen AI myself (as long as nobody forces me to, e.g. maybe for work in the future) since I genuinely like coding (and music). As a "stackoverflow" replacement and "research" + "administration" assistent it's sometimes useful but for that online search engines are usually enough.
p.p.s.: besides, what's the point of a hobby if you let someone (or something) else do it for you
</wall_of_text>
AI will certainly not disappear but it's too early to try to roll it out at this scale, i.e. I think that (for now) it does not belong in the hands of everyone (local LLMs and search engines aside). For it to be useful (fast enough to run the trillion+ parameter top models) you need those data centers and heavy duty equipment (A100 etc), a gaming video card won't do. But the tech isn't there, yet (i.e. not efficient enough), and we also don't have unlimited-energy fusion reactors, yet (maybe in 20 years, har).
Another concern is that we are heading straight into an idiocrazy if everyone just "vibes" and never learns how things actually work.
It somehow works at the moment since the current seniors grew up without AI, gathered knowledge and know exactly what to prompt but the next generation already struggles to find employment because (some) companies don't see worth in them anymore ("they can easily be replaced by an AI" was already said two years ago). In every society there's this "social contract" which does not just mean that the young are simply paying our rent but that the old are also giving the young a fair chance to learn and evolve, and eventually become seniors themselves. Trying to cut them out for short term profits is a major no-go.
With tools like REA (reverse-engineer-all), and the general "everything's up for grabs" attitude that the legally questionable model training on copyrighted material made hugely popular, there are some tough times ahead for many software companies.
If what I heard is correct, some of the big players now offer AI subscriptions that protect their customers from lawsuits.
On the other hand, (some) companies require their employees to sign a document that shifts the liability onto them.
Besides, in the EU, the Product Liability Directive (Directive (EU) 2024/2853), is taking effect on December, 9th), i.e. you are responsible / liable for the software you sell, regardless whether it is (customized) open source, human-written closed source, or AI generated).
Good luck providing a fix for an AI-generated software you don't really understand in detail when the AI fails to resolve the issue.
For non-profit hobby communities, the onslaught of AI generated content has stirred up quite a bit of an uproar, to put it mildly.
Neither the attitude presented by the (radical) AI proponents ("hey, now we can finally get rid of those elitist developers, AI democratizes everything"), nor the one by the anti-AI fraction who slanders everything that remotely looks like AI was somehow involved, is very helpful here.
For some, the ongoing AI content flood has caused them to lose motivation because why invest months or years if the result gets drowned in 'mindless slop', anyway. Not sure this applies to the demoscene (there's no flood of AI prods as far as I can see) but this might change in the future.
The aforementioned AI rules posted on nesdev.org and discussed on EAB sound like a reasonable compromise / "code of conduct", e.g. for demo parties.
Last but not least, the general AI topic has been discussed to death (it's everywhere!) and this thread is about porting old DOS demos to browsers (my opinion on that is that I'd rather watch the original in an emulator *shrug*).
p.s.: ..and in case anyone wonders: I do not use gen AI myself (as long as nobody forces me to, e.g. maybe for work in the future) since I genuinely like coding (and music). As a "stackoverflow" replacement and "research" + "administration" assistent it's sometimes useful but for that online search engines are usually enough.
p.p.s.: besides, what's the point of a hobby if you let someone (or something) else do it for you
</wall_of_text>
