@meistermeier Smudo hält sein Gesicht einfach für jede Werbung hin
RE: https://az.social/@lerxst/115789867119287844
We are a weak people, secured by protective oceans, endowed with fabulous natural resources, but with a grossly overrated character, fed by myths and misconceptions that served us well for a time. That time has passed.
For those who haven't got the backstory here, a GenAI company basically pre-ordered 40% of the world's RAM to lock out competitors... then Micron, who make their profits by making RAM, decided to also stop making RAM to chase GenAI fad.
So now PCs and servers are basically unaffordable. Nobody is quite sure what will happen.
If you would like to know why the price of DRAM and certain storage such as SSDs has exploded:
"On October 1st OpenAI signed two simultaneous deals with Samsung and SK Hynix for 40% of the worlds DRAM supply. "
40%...of global supply
Neither of the suppliers knew about the other deal and they'd be announced on the same day.
"OpenAI’s competitors, OEMs, and cloud providers scrambled to secure whatever inventory remained out of self-defense, "
This caused other companies to lock down supply as well.
"Cloud service providers are behaving similarly. High-density NAND products are effectively sold out months in advance. Samsung’s next-generation V9 NAND is already nearly booked before it's even launched. Micron has presold almost all of its High Bandwidth Memory (HBM) output through 2026. Contracts that once covered a quarter now span years, with hyperscalers buying directly at the source."
The first reference below goes on to explain other factors and potential impacts. The second reference provides more context for preexisting pressures on the DRAM supply.
Reference:
https://www.mooreslawisdead.com/post/sam-altman-s-dirty-dram-deal
Video version of the article above: https://www.youtube.com/watch?v=BORRBce5TGw
@GossiTheDog Apologies if the following was covered elsewhere in the thread...
In addition to OpenAI buying 40% of the global DRAM supply OpenAI’s competitors, OEMs, and cloud providers scrambled to secure whatever inventory remained out of self-defense. This locked down a significant about of the global supply. This is already lead to death of certain video card models and will likely lead to other products being killed or having restricted production runs. Unless something surprising happens the impacts will get worse and it is unlikely to improve much in the next 12 months.
Hello and a good meowing everyone!
Have a most wonderful #Caturday and stay safe! ♥️😸
@stux I kind of think of cats being able to do that naturally!
@cstross I am increasingly convinced the AI bubble exists solely to sell the idea of being able to do everything as a means of concentrating money.
The point is to participate in a mammonite magic to have all the of the love of god (that is, all the money) and the thinking involved is as completely detached from any material concern as a stylite.
(The "and everything will be easy!" part is something of a tell, here; that's the language of attaining paradise, not material ambition.)
@BartoszMilewski If you invert the question, it feels less puzzling: why do we treat “66 million years” as a naturally meaningful chunk of time at all?
Most of our deepest claims about reality - causality, locality, stable constants - are extrapolations from an absurdly tiny empirical foothold: a thin slice of time and a minuscule patch of accessible space.
So maybe the real surprise isn’t that our theories feel incomplete—it’s that our pea-brain models generalize as far as they do.
@mathemagical
The biggest example of hubris is to imagine that our monkey brains should be able to figure out the workings of the Universe
Uh … AI bubble costing the US as much money in 12 months as *two major wars over 13 years* is quite the buried lede!
"If it can't go on forever it will stop." — economist Herb Stein, 1985. (The "it" is a placeholder for any economic process.)
https://researchbuzz.masto.host/@researchbuzz/115782207999466966
@cstross I am increasingly convinced the AI bubble exists solely to sell the idea of being able to do everything as a means of concentrating money.
The point is to participate in a mammonite magic to have all the of the love of god (that is, all the money) and the thinking involved is as completely detached from any material concern as a stylite.
(The "and everything will be easy!" part is something of a tell, here; that's the language of attaining paradise, not material ambition.)
I write technical articles on my blog.
AIs show up in large numbers to read them, crawl them, learn from them.
Time passes. I publish a new post.
And right on schedule, someone comments:
"This was clearly written by an AI".
Which is fascinating, really.
I write.
Machines read.
I keep writing.
Then humans accuse me of being the machine.
At this point I am not sure if the problem is that AI sounds too human,
or that humans have forgotten what a human who actually studies sounds like.
Either way, I will keep writing.
Worst case scenario, the AIs will enjoy it.
Best case scenario, one day a human will too.
@ChrisMayLA6 @econads Add to that the extreme difficulty in performing good experiment when your subject of study is society. It is close to impossibl to create to equal societies and compare one under socialist rule and one under libertarian rule.
That is why I think the best we can do are those type of histo-empirical studies, and then it can be argued, naturally, that past performance is not a guarantee of future performance, that you are comparing apples to oranges, that science and technology today makes X possible, while it was not possible 200 years ago, etc. etc.
To me, this is both the charm and utter frustration with the social sciences.
Indeed; that's why I also find social sciences that is based on modelling from first principles frustrating; really one needs to start from the historical record.... but as the debates about public debt as a share of GDP show, even there some significant issues of data reliability & accuracy can quick undermine seemingly relatively straight forward arguments.
Small hint here: this analysis only applies to people who, in living memory, did feel listened to by the government. Marginalized groups tend not to go fascist for many reasons, including that there is no recent Glorious Past to return to.
(To a first approximation; thanks to intersectionality, oppression is fractal all the way down, and there are always people who choose to believe lies about the past, e.g. tradwives.)
✏️ Dear Mozilla, I don't want an “Al kill switch”, I want a more responsible approach for all
a new blog post! (*I do want the kill switch, but responsible rollout matters even more in the grand scheme of things)
So überraschend, da fällt mir selbst nichts mehr zu ein.
Pic. 1. Human Interface Guidelines, Apple, 1992
Pic. 2. macOS Tahoe, Apple, 2025
@nikitonsky
The key word being ”arbitrary.” You can dislike icons in menus, if you wanted an option to disable them if I’d support it, but an icon matching the menu item is not arbitrary.
"In [the Unaccountability Machine], I diagnosed the problem of populism as being that people had got tired of not being listened to by the systems that governed them, and responded by using the only remaining communication channel they had. To take the title of my post on the US elections, the gap between governance and governed had got so big that the only signal that channel could carry was a scream. This was (and is) my theory of Trump, Farage, Meloni, etc etc."
https://backofmind.substack.com/p/what-is-once-sprung-cannot-necessarily
@ellyxir Yeah, "good old fashioned AI" at most, but most is just some maze generation or cellular automata with a lot of hacky special case rules to get the map style you want.
I used to have some game design articles which included a few ways to do it, but I never reposted them.
I do like this book:
https://pragprog.com/titles/jbmaze/mazes-for-programmers/
@quinn I'm hopeful (but beware—I'm a pathological optimist) that the AI bubble is inflating *so* fast that it'll burst before too many of the data centres/coal-burning power stations being built for it come on stream. So it'll be hugely wasteful but in the final analysis just a brief blip on the multi-decade climate change graph.
And hopefully the next investment bubble will be in PV panels and grid scale batteries.
@cstross I also share that hope, though it's going to result in absolute misery for so many people, and I suspect an era of madness, especially in my home country.
All of this will end in tears.
Can you guess why the microwave emissions from OH26.5+0.6 show two peaks? Every OH maser star does.
For more on this star, check out this paper:
• O. Chesneau, T. Verhoelst, B. Lopez, L. B. F. M. Waters, Ch. Leinert, W. Jaffe, R. Köhler, A. de Koter and C. Dijkstra, The mid-IR spatially resolved environment of OH 26.5+0.6 at maximum luminosity,
https://www.aanda.org/articles/aa/pdf/2005/20/aa2235-04.pdf
For more on astrophysical masers:
https://en.wikipedia.org/wiki/Astrophysical_maser
This particular star is a 'Mira variable', and they are quite interesting too:
https://en.wikipedia.org/wiki/Mira_variable
(2/2)
Casimir effect b/c there's a magnetic field?
Merry Christmas to everybody, except that dude who works for Elastic, who decided to drop an unauthenticated exploit for MongoDB on Christmas Day, that leaks memory and automates harvesting secrets (e.g. database passwords)
CVE-2025-14847 aka MongoBleed
Exp: https://github.com/joe-desimone/mongobleed/blob/main/mongobleed.py
This one is incredibly widely internet facing and will very likely see mass exploitation and impactful incidents
Impacts every MongoDB version going back a decade.
Shodan dork: product:"MongoDB"
@GossiTheDog It's ok. They'll fix it with LLMs. No need to bother humans. (sarcasm)
@davidallengreen Blasphemer! The series actually got me reading the book (and others by Austen) at a time my English was considerably worse than now. (Still bad, I know 😋)
@frauxirah As a gateway drug is ok.