> The project was met with significant legal challenges from authors and publishers which was eventually overcome
I don't think they were overcome. As far as I remember Google couldn't make the books available so they abandoned the project. They possess the scans (if they didn't delete them) but they won't be made public.
IIRC the courts ruled that because it would be implausibe for a person to use google books previews to read an entire work (you'd have to make a whole bunch of separate accounts to do so), it could not plausibly impact the market for that work.
The consensus among copyright lawyers is Google lost Authors Guild v Google, but for some reason the media and Wikipedia do not clearly report it that way.
In some countries they do, in some they don't, in others every PhD student even gets a scholarship. Every person I know who did the PhD did it because they loved the topics they were studying.
The author of the blogpost is from Russia, I bet he didn't have to pay for his PhD.
My uneducated guess says, not much. For running massive models you still need a ton of high-bandwidth interconnects between many individual chips/GPUs/etc since you need to do math across a few TB worth of weights. That's simply going to require more power (and more die area in I/O, and therefore more cost). Being able to run small models in tiny power envelopes is incredibly useful to people, but I believe it will be covering a different niche than what datacenters can provide. Likewise, you'll still need crazy amounts of high-end memory to populate whatever goes in these datacenters.
The only thing that will crash prices is reduced demand (duh) or, more interestingly, increased production. In particular, if CXMT is able to get their DDR5 fabs up to a reasonably high yield, that could add some downward price pressure (as could government subsidies). As well, if Micron/Kingston/Hynix think that CXMT is going to start cutting into their market share, they might be willing to either increases supply or drop prices. Unfortunately CXMT looks to be taking quite a while to get their new fab up to max capacity so that may take a year+ before anything manifests.
If you're interested in following the (publicly available) info on these sorts of things, check out what companies like Axelera, DeepX, and MemoryX are doing today and have on their roadmaps, as well as the sorts of chips/SoCs Qualcomm, Kinara (now NXP), and Ambarella currently have announced (or have on the market). And remember, that pretty much all of these chips on the market today were in initial development more or less when ChatGPT first launched. If you knew what you knew today (or a year ago) about what requirements current- and next-generation models would have (from a silicon perspective), what might you do differently? Think for instance, host system interconnects, amount and speed of on-package or on-die memory, image/video decode capabilities, int8 vs fp8 vs fp16 vs bf16 compute units, etc. And, consider that most "AI" stuff in development a few years ago was all 15nm or 12nm - because who was gonna pay big money to get fab capacity at 3nm to run some object detection models? So most of the stuff on the market today is on very old nodes and therefore not super power efficient.
Pretty well, to the point where the US administration complains that there's a trade imbalance. (which is true for manufactured goods, false when we count in services)
Chrome sends home the urls you visit together with the page performance data (and probably more). That's how they build Chrome User Experience Report (CrUX) for the most popular sites: https://developer.chrome.com/docs/crux
I remember one Polish diplomat recalling that his boss told him "I am sending you to the safest outpost in the world" when offering him embassy in North Korea.
I think I've read somewhere that JIS was phased out after Philips standard included some of JIS features and that theoretically modern Philips screwdrivers should be compatible with JIS.
The JIS, ISO and ASME standards for cross head screw drivers are fully compatible.
The myth that JIS needs special screwdrivers is caused by people using either using the wrong size (easy at small sizes), using a worn-out or poorly-made screwdriver, or using Pozidriv (which is genuinely very different).
Anecdotally, about 5 years ago I bought a set of off-brand "premium" screwdrivers, where it turned out the Phillips head ones (yes there are two L's in there) were so far outside the standard I had to just throw them in the trash.
You got the wrong standard. JIS B 1012 type H is Philips equivalent, but what people refer to as the JIS screw is JIS B 1012 type S. JIS B 1012 Type S ("JIS") and ISO 8764 (Philips) are entirely different screw and scredriver profiles. See https://www.peterverdone.com/jis-cross-head-screws-and-drive....
The Philips screwdriver has lobes with a large multi-stage taper that ejects the screwdriver from the screw upon torque. The JIS screwdriver lobes are in contrast almost perfect 90 degree angle with a tiny filet in the center - not dissimilar to the Frearson screw.
A Philips screwdriver will be unable to insert all the way into a JIS screw due to this taper, exactly similar to how the extra lobes on a Pozidriv screwdriver collides a Philips screw. Attempting to apply torque with this partial engagement puts you at very high risk of stripping the screw.
A JIS screwdriver however does work perfectly fine in Philips screws, often even better than the Philips screwdriver as you did away with one half of the "screwdriver ejection feature".
There's a reason that iFixit goes out of their way to make both JIS and Philips bits, and warn you in their guides. They aren't really in the business of trying to cram more bits down your throat for fun.
For me the biggest problem was never memorizing the suffixes but determining whether the verb belongs to group 1 or 2. While you can tell immediately that some verbs belong to group 1, the others (e.g. kaeru mentioned in the post) are not so easy (as far as I know there's no algorithm for that and you just need to memorize with every verb to which group it belongs).
I don't think they were overcome. As far as I remember Google couldn't make the books available so they abandoned the project. They possess the scans (if they didn't delete them) but they won't be made public.
reply