The public scare about long half-lives is particularly weird when considering other aspects of nuclear vs. fossil power. Nuclear is known for rare freak events such as Chernobyl, which kill a bunch of people at once, while it's fossil fuels that are killing a lot more people in the long term. This is so even if we don't consider global warming, due to effects such as fine particle pollution. Here nuclear is the scary one, because there are no sudden deaths due to fine particle pollution, and because people are bad at statistics and long-term thinking. Besides, we're just more familiar with fossil fire. A fireplace symbolizes cozy, old-fashioned life, even if it's actually a worse polluter than a car due to the incomplete combustion.
But when it comes to nuclear waste, suddenly the hoi polloi worries about long-term effects. I'm not saying we should ignore the radiation of long-term nuclear waste, but it seems easier to contain than the CO2 and fine particles from fossil fuels.
Have you looked into the Gentoo binary package host project? Lets you mix & match binary and source packages. You should be able to install a binary version of LLVM that way.
I'm aware of it and I've set it up, but suitable packages are rarely available, especially for 32-bit x86.
A piece of nuclear waste contains a certain amount of energy. If it's released steadily* over 1e5 years, the radiation can't be very intense. Conversely, materials with short half-lives are "hotter". So I'd be more afraid of having the latter in my back yard, but for some reason the public scare is always about the long time scales, rather than intensity.
*(Exponential decay is not strictly "steady" as in linear, but for these purposes the details don't really matter.)
AI or not, chip makers and IT service giants aren't going anywhere. While the Big 7 companies are overvalued due to AI, they don't just do AI, so picking stocks based on binary AI criteria is problematic. Things are clearer with the smaller AI-only companies, but they don't count as much in the big picture of index funds.
Even 2D desktop stuff has used OpenGL and other "3D acceleration" features for years. 2D is a subset of 3D, and there's no point in maintaining a separate API and hardware features for 2D.
More technically speaking, the "3D" part is a slight misnomer, as the end result of OpenGL rendering is always a 2D picture. Besides, OpenGL internal coordinates are actually 4D, and of course you can use any dimensionality internally in shaders. I can't speak for other graphics systems, but I'd guess they work rather similarly.
Grants from art foundations are very different from basic income schemes. Slashdotters might be more familiar with grants for scientific research, such as PhD or Post-doc programs, and it's basically similar for arts. There are many more applicants than available grants, and you have to spend a considerable part of your time on the application process, rather than doing the art/science itself. You also need to show your worth in some way, basically working for years and years without grants to build a decent body of work.
As another commenter pointed out, it's the beginners that need the money more than the established professionals, so basic income makes much more sense.
I bought a used Thinkpad last year. A few years old. Amazing computer. It looked brand new so probably was from some clueless corporation as "surplus".
I also have a used Thinkpad that seemed intact, except for some wear on the outer case corners. I'm guessing it was always used with a dock, so nobody has touched the keyboard or the display.
OTOH, I also have a used Chromebook with a school's logo painted on the top cover, but otherwise looked intact. It doesn't have a docking connector (traditional or USB-C) so it could be a "surplus" item, although that too seems unlikely for a school.
I'm a Finn and I do mind!
Finnish homes don't traditionally have cooling AC, because it hasn't been necessary. In the past, we might get a few days over 30 C (with high humidity) in the summer. In recent years, we've experienced much longer heat waves, such as a few weeks in a row. Consequently, a lot of people have fitted air heat pumps, and those can also be used for heating in the winter using reversed flow direction. But even now, cooling AC is rarely installed in new buildings, because apparently the design/construction companies are full of climate change deniers. Office buildings usually have proper AC, so those with traditional indoor jobs don't suffer that much from daytime heat.
Paradoxically, global heating has also caused colder winters in some sense. Right now we're in a middle of an extra long cold spell. This is apparently because global warming affects polar regions much more, and the polar vortex/jetstream/something has been disrupted. As a result, we get weeks and weeks of the same extreme weather, be it summer or winter. I won't even go into the possible disruption of the Gulf stream, which would cause further cooling up here.
I've always thought active cooling in large scale is kind of idiotic, because you're pumping the heat out into the atmosphere, thus resulting in need for more AC. While it might not contribute that much to actual global warming, the local effect is quite noticeable in cities such as NYC. Somehow, people in the Middle East and elsewhere have figured out better cooling solutions centuries ago, such as https://en.wikipedia.org/wiki/... . There are also basic building methods that reduce the need for AC, such as painting the outside walls white, sensible positions and sizes of windows, and using awnings on them. Growing trees to shade the buildings also helps a lot in a number of ways.
If there's a "theorum" about that, they'll find it eventually.
FFS, the summary has a link to https://en.wikipedia.org/wiki/... named "Star of David theorum". Ever heard of copy-paste, or does that need an LLM these days?
But do NPUs lock you into doing AI? After the AI bubble bursts, your CUDA/Nvidia hardware can be used for other things. While I prefer more open solutions such as OpenCL on AMD, the same OpenCL code will also run on Nvidia hardware.
I don't know about the details of NPUs, but I've got the impression that efficient AI uses very low precision arithmetic, such as 8-bit and even lower. So while it could be possible to use NPUs for general vector/matrix math, the low precision would be a deal breaker for a lot of applications.
Nothing happens.
