no problem.

So, in order to protect against possible military applications(known for their cost-sensitivity...); we are making the sale legal as long as El Presidente gets his cut? That's in character, sure; but what's the paper-thin excuse for that being a cogent policy idea?

I'm actually responding to the AC above you. He is arguing that the attack wouldn't make any sense for either country to make, based on *national* interest. I'm pointing out that's not the only framework in which *regimes* make decisions.

I'll find out in mid January, lol - it's en route on the Ever Acme, with a transfer at Rotterdam. ;) But given our high local prices, it's the same cost to me of like 60kg of local filament, so so long as the odds of it being good are better than 1 in 8, I come out ahead, and I like those odds ;)

That said, I have no reason to think that it won't be. Yasin isn't a well known brand, but a lot of other brands (for example Hatchbox) often use white-label Yasin as their own. And everything I've seen about their op looks quite professional.

Just put it in context: Today Russia struck the Pechenihy Reservoir dam in Kharkiv.
Russia launched the war because they thought it would be a quick and easy win, a step towards reestablishing a Russian empire and sphere of influence, because Putin thinks in 19th century terms. Russia is continuing the war, not because it's good for Russia. I'd argue that winning and then having to rebuild and pacify Ukraine would be a catastrophe. Russia is continuing the war because *losing* the war would be catastrophic for the *regime*. It's not that they want to win a smoldering ruin, it's that winning a smoldering ruin is more favorable to them and losing an intact country.

One major difference, assuming you've got full platform support(should be the case on any server or workstation that isn't an utter joke; but can be a problem with some desktop boards that 'support' ECC in the sense that AMD didn't laser it off the way Intel does; but don't really care); is that ECC RAM can (and should) report even correctable errors; so you get considerably more warning than you do with non-ECC RAM.

If you pay no attention to error reports ECC or non-ECC are both rolling the dice; though ECC has better odds; but 'proper' ECC and Linux-EDAC support will allow you to keep an eye on worrisome events(normally with something like rasdaemon, not sure what other options and preferences there are in terms of aggregating the kernel-provided data) and, unless the RAM fails particularly dramatically and thoroughly, will give you much better odds of knowing that you have a hardware problem while that problem is still at correctable levels; so you can take appropriate action(either replacement, or on the really fancy server systems, some 'chipkill'-like arrangement where the specific piece of DRAM that is failing gets cut out of use when deeemed unreliable without having to bring the system down.

Sometimes you'd get a BSOD that was a fairly clear call to action; when the error called out something recognizable as the name of part of a driver; but that is mostly just a special case of the "did you change any hardware or update any drivers recently?" troubleshooting steps that people have been doing more or less blind since forever; admittedly slightly more helpful in cases where as far as you know the answer to those questions is 'no'; but windows update did slip you a driver update; or a change in OS behavior means that a driver that used to work is now troublesome.

Realistically, as long as the OS provides suitable support for being configured to collect actual crash dump material if you want it; it's hard to object too strongly to the idea that just rebooting fairly quickly is probably the better choice vs. trying to make the BSOD a genuinely useful debugging resource; especially given how rare it is for the person with useful debugging ability to happen to be at the console at the time of crash(rather than just an end user who is ill equipped to make sense of it; or a system that mostly does server stuff, quite likely not on actual physical hardware, where nobody has even touched the physical console in months or years; and it's more or less entirely useless to display a message there; rather than rebooting and hoping that things come up enough that management software can grab the dump files; or giving up and leaving the system in EMS so that someone can attach to that console.

Yeah, it's not even worth considering for something like 15-20kg. A full pallet in this case is 464kg :) Years worth of supply for me.

The current "AI" is a predictive engine.

And *you* are a predictive engine as well; prediction is where the error metric for learning comes from. (I removed the word "search" from both because neither work by "search". Neither you nor LLMs are databases)

It looks at something and analyzes what it thinks the result should be.

And that's not AI why?

AI is, and has always been, the field of tasks that are traditionally hard for computers but easy for humans. There is no question that these are a massive leap forward in AI, as it has always been defined.

It is absolutely crazy that we are all very very soon going to lose access to electricity

Calm down. Total AI power consumption (all forms of AL, both training and inference) for 2025 will be in the ballpark of 50-60TWh. Video gaming consumes about 350TWh/year, and growing. The world consumes ~25000 TWh/yr in electricity. And electricity is only 1/5th of global energy consumption.

AI datacentres are certainly a big deal to the local grid where they're located - in the same way that any major industry is a big deal where it's located. But "big at a local scale" is not the same thing as "big at a global scale." Just across the fjord from me there's an aluminum smelter that uses half a gigawatt of power. Such is industry.

That "ruler study" was ancient. It's mentioned in peer review at least as early as 2018, and might be even older.

Believe it or not, people in the field are familiar with these sorts of things that you just read about.

Most of these new AI tools have gained their new levels of performance by incorporating Transformers in some form or another, in part or in whole. Transformers is the backend of LLMs.

Even in cases where Transformers isn't used these days, often it's imitated. For example, the top leaderboards in vision models are a mix of ViTs (Vision Transformers) and hybrids (CNN + transformers), but there are still some "pure CNNs" that are high up. But the best performing "pure CNNs" these days use techniques modeled after what Transformers is doing, e.g. filtering data with an equivalent of attention and the like.

The simple fact is that what enabled LLMs is enabling most of this other stuff too.

Jevons Intensifies

I've done my first test of buying a whole pallet of filament straight from a Chinese manufacturer. It's a risk - it could be all junk - but if it's usable, the price advantage is insane. Like $3/kg for PETG at the factory gate (like $5/kg after sea freight and our 24% VAT). Versus local stores which sell for like $30/kg.

I'd love to see someone try to 3d print with a filament that melts at 162K. Where do you even buy xenon filament? ;)