cbolton

That's using he most enthusiastic fanboy predictions as the baseline by which to judge success, not a recipe for reaching the right conclusions.

I'd rather look at all the data available. If you want to compare to Loongson, you have to include the 20 years head start that this company has in developing microarchitectures, a big part of which can be transferred to the Loongson designs. And factor in the strategic investments in China who doesn't want to depend on ARM and x86, while in the Western world people were happy with x86 and ARM and big money is mostly flowing to AI chips, where RISC-V is successful.

Then there is Xiangshan Kunminghu... For some reason the most powerful open source CPU is RISC-V. Chinese and not Loongson.

So it's late to deliver hugely complex cores according to some optimistic timeline? That's a weird measure of success. The ISA is successful in the low end and is seeing huge investment on large cores, though most of the promising companies are now focusing on AI or getting bought by large players like Qualcomm. So yeah it might take a while still before we see competitive high end consumer cores. I would still put my money on RISC-V rather than Loongson for the long term but we'll see.

Not sure this quasi-religious framing does RISC-V any good.

By some coincidence RISC-V became successful just before/while AI took off. I'm glad it did. The design is pretty good for low end cores and probably good enough for high end cores, so there's a chance it will become one of the dominant ISAs in the mid-term.

We're probably getting to a point where AIs will just generate optimized binaries directly, so I'd rather have them generate them in a machine language that is human friendly. RISC-V being designed by humans with a focus on simplicity is close to the best we could wish for.

This is just a webpage some individual put up, right? I mean as far as I can see there's no institute, no company, no government behind this. Just one RISC-V enthusiast on the contact page.

> Taler has audit logs containing what token was used from which customer(id) with which merchant

This is completely wrong. The Taler exchange doesn't know which tokens you have. When you spend them as a customer, the merchant passes your tokens to the exchange. It's like physical coins: you give them to the merchant, the merchant deposits them in their account. The exchange know which merchant deposits the tokens but has no information on the customer because the tokens (like coins) are anonymous.

Some form of AI RSI is likely, but why would you expect the improvement at each level to stay proportional the level itself (which is required for exponential improvement), or even why would the rate of improvement stay above a particular constant (required for linear growth)?

To me the most likely scenario is RSI with diminishing returns, as the models get closer to the limits of the general approach to AI that they embody. The decreasing rate of improvement might look like a log curve, or even an horizontal asymptote. Not that scary.

Please digitize and post some of these recordings!

You make it sound like the center party is playing both sides but as I understand it is not part of that block, and they would be OK with social democrats but not far left in government. What's wrong with that?

That's... not what happened? See the timeline I posted here: https://news.ycombinator.com/item?id=49445897

JPEG XL is largely a Google effort including in creating the specification, the reference implementation and later a safer Rust implementation. Why would Google executives want to stop it? From what I gather it's Chrome engineers that didn't want it because it was huge, had little adoption (at the time), and was rather redundant with AVIF and WebP (for Web purposes). Not saying they didn't have related mandates from executives but they also have a lot of say in what goes in and what doesn't.

The handwavy section on RISC-V is underwhelming. Yes the ecosystem is still nascent but something like the SpacemiT K3 is far from "abysmal" for vector operations and would make a nice test case with its two core types both supporting RVV 1.0 (one with vlen 256, the other with vlen 1024). On the more industrial side, for example the SiFive Intelligence X280 has been used in TPU by Google for years already and I doubt they are the only ones so calling the ISA completely irrelevant is a bit of a stretch.

Just for the sake of curiosity it would be nice to have a peek at what SIMD in Rust looks like on RISC-V today. Yes, even if it requires some "obscure compiler flags" for now (while we wait for the Oilsm extension).

This video has 30 examples: https://youtu.be/nw2QfPREu-Q

Not saying you're wrong on the main message (and those examples are with filters rather than true CRT) but it's still impressive how it can make the pixel art better looking in those cases.

Not so clear in the article is that also with consoles the game pixel grid didn't match the CRT grid (even in "pixel count") so just by geometry you have blending of one pixel data into several CRT "pixels" (plus varying intensity even in a single phosphor patch as the analog RGB signals rise and fall in intensity).

As for CRT filters one can argue how realistic they are but they show that scanline gaps and blending at least can produce amazing perceptual improvements as shown in this video: https://youtu.be/nw2QfPREu-Q

> Over 10% of the country will be in the top 1% of earners at some point in their life

Though most of them only for one year due to temporary revenue, so it's not that rational.

The republican program was in large part vetoed by Clinton though and the spending cuts that passed were largely offset by the tax cuts. From what I read the reduction in deficit is mostly due to the dot-com boom, helped with the "deficit reduction act" of 1993 that was pushed by Clinton and happened to hike taxes at the right time to tax the boom.

I completely agree.

Why the aggressive tone and the strawman rhetoric? I never said there was a consensus. Yes I'm talking more about the "optimistic" commenters and pointing out that this chess thing is a good datum to temper their enthusiasm. What's wrong with that?

Also these claims are not completely without merit, it's just that LLMs seem to excel at specific "cognitive" tasks and it's interesting to see where they fail.

Neither. As I said I want to measure cognitive abilities.

Your "ability of the box" is like "economic potential" in my previous comment. If that's what you want to measure, fine. But I want a deeper understanding: what is the thing doing, how is it solving problems? I want to get a sense of its abilities that is richer than a one-dimensional scale.

Well it's not a perfect test so you need a bit of care in how you use it. If you have no idea what the subject is doing, then you don't know if you're measuring cognitive ability or something else (like cheating ability, or algorithmic sophistication or whatever). But failing the test is a pretty clear sign of certain cognitive abilities being poor.

It's a great test of cognitive abilities. There are many claims that current LLMs surpass humans in cognitive abilities so it's noteworthy that they underperform on that test.

Letting the model execute a chess program (that it wrote) would make sense if you're measuring its economic potential, but for cognition that would be cheating just like if you let a human run a chess program. The fact that the human would have a much harder time writing a useful program is irrelevant.

Sorry if that was confusing, I have edited my post to clarify a bit: That sentence was about a feature specific to JPEG XL, which is that if you have JPEG files you can re-encode then as JPEG XL to save space without any loss of quality. It's nice for peace of mind, you can bulk convert a large archive and you know you still have the same pixels everywhere (later you can even turn them back to JPEG without loss if you want). AVIF cannot do that. Now if you are OK with some loss in the conversion, you can also do much better than 20% with JPEG XL.

The article makes a good point: JPEG XL is amazing but not specifically for the typical Web use cases, compared to AVIF. But the conclusion doesn't follow. Having an excellent and versatile format supported by browsers is very useful!

Maybe a lossless re-encoding of my website JPEGs to gain 20% size is not worth the 33% longer decode, and maybe the website doesn't need very high resolution images or more than 12 bits per channel. But my personal archives can definitely use that and it's great that such files can be viewed everywhere without specialized tooling, including served on the intranet and indeed the Web.

In summary JPEG XL easily beats AVIF on the not-so-long tail of use cases and having compatible viewers everywhere is quite useful. And with JPEG XL becoming part of PDF, it will soon be common use cases too.

As for the rest of the article: there are interesting points in the benchmark section but I think it's too early in the JPEG XL adoption cycle to draw conclusions. One thing bothers me though: There is a single picture comparison (with tag line "results speak for themselves") and I find it quite misleading: it's just picking one point on the "bytes received" line that looks best for AVIF compared to JPEG XL. Try it yourself and you'll see JPEG XL shows something already at 2KB while AVIF has nothing until 8KB, and JPEG XL looks better than AVIF after 100KB.

Looking it up I was actually remembering from Feynman's QED book (path integral approach). I don't have it at hand but I think it's a description based on photon particles. Anyway I agree my description with photon travel "between atoms", emission and absorption was quite bad (especially if you think absorption and emission as slow incoherent processes instead of a general way of describing interactions which is what I meant).

But still I think saying "the primary wave gets absolutely and completely delayed" is not helpful. Using a wave description as in Feynman's lecture[1] is more enlightening: the incoming wave travels at "full speed" through the medium, but doing so it interacts with atoms such that they emit an additional wave, and the sum is a slower wave.

You can say it's the same since there's only one electric field in space and so the only "real wave" is the sum of all effects. But I find it quite helpful to think that one of the components in this sum is the original wave traveling at the speed of light in vacuum, also in the space occupied by the medium.

[1] https://www.feynmanlectures.caltech.edu/I_31.html

Yeah I didn't find that helpful. What I remember from Feynman's lectures is that photons still travel at "full speed" c between atoms, but if you look at the global progression of light as photons get absorbed then emitted it progresses slower than c.

Maybe it's projecting but what I see is kids ganging up to break a target that is often human/animal-shaped, and that feels different to me.

Yeah I'm not saying it's bad for the children, but it's unpleasant for me. It seems violence is a concept that gets deeper and uglier as you grow. For young children it's often abstract and maybe harmless. I just read a book to my child that turned out to have violent scenes that horrified me a bit but he barely thought about it and only in technical terms (like where is that artery ripped by the wolf). Same with people that had fun watching horror movies as teens, then find them unpleasant when they're older.

It never occurred to me to consider efficiency in piñata, but I dislike them because it's making a game out of violence.

What did you do to get the money?

Isn't it a bit arrogant to pass judgement with such confidence based on the flimsy evidence of a three word sentence? Anyway maybe this will convince you:

https://news.ycombinator.com/item?id=25948632

https://news.ycombinator.com/item?id=24862016

https://news.ycombinator.com/item?id=11943786