Veedrac

24 ms·

301.
by Veedrac 4y ago
Debris removal is less an R&D problem than a market problem. The highest impact thing a government can do is put a price on debris removal, and then guarantee that price for existing debris for an extended period of time. Flight volume
302.
by Veedrac 4y ago
This is mostly just an angry rant, yes, but equally it is just true. Marcus is intellectually dishonest.
303.
by Veedrac 4y ago
After Nanite, it seems to me that mesh shaders are just a transitory point before everything inevitably moves primarily to raster in compute. Sticking to fixed-function hardware doesn't make sense unless that hardware is faster. (As I
304.
by Veedrac 4y ago
https://docs.python.org/3/library/stdtypes.html#configuring-...
305.
by Veedrac 4y ago
> But memory has felt like an exception to Moore’s Law for a while, at least in practice. That's because it is. Memory is failing to scale. That's why there's so much investment in alternative memory technologies, includin
306.
by Veedrac 4y ago
> Traffic was about 25% of PM2.5 in 1995, combustion was like 90% of that. 1995 numbers seem like a poor choice, given engines ran vastly dirtier back then. > "Every bit helps" though right? Except PM pollution is global in
307.
by Veedrac 4y ago
It's quite easy to find websites that claim it is, do you have something more authoritative that claims otherwise? Eg. https://www.tiretechnologyinternational.com/news/regulations... cites a UK study.
308.
by Veedrac 4y ago
Stupid title aside, air pollution is one of the most important issues humanity has faced ever, even ignoring climate change, and it's a terrible oversight that we've focused so little effort on it. One would think after lead that
309.
by Veedrac 4y ago
Sort of off topic but also sort of relevant, pixel-accurate mouse motion blur should really be a thing by now. The lack of it makes mouse movements significantly less smooth than need be.
310.
by Veedrac 4y ago
Making intentionally and obviously bad plays that directly work against the agency's goals because that way they get more free government money is corruption. The ‘a 50 year old rocket that was the first of its kind was expensive too
311.
by Veedrac 4y ago
> By such logic, we should have wished the other shuttles had blown up in the 90s too. Yes. If Shuttle failed faster NASA would be less motivated to make intentionally terrible decisions. The greatest irony of these defenses is perhaps t
312.
by Veedrac 4y ago
That is a very unfair comparison. You are giving numbers for Starship going to NRHO, then landing the whole upper stage on the moon, then returning it to NRHO. That is a completely different ballpark to SLS's ‘dump a capsule in a l
313.
by Veedrac 4y ago
Absolutely. The Fukushima disaster: https://en.m.wikipedia.org/wiki/Fukushima_Daiichi_nuclear_di... The earthquake: https://en.m.wikipedia.org/wiki/2011_T%C5%8Dhoku_earthquake_... Zero known fatal
314.
by Veedrac 4y ago
Nuclear is uneconomical because safe, economical nuclear is illegal. That's it. If everyone were pro-nuclear, those laws wouldn't exist (or would exist in a sane form), and nuclear would be cheap.
315.
by Veedrac 4y ago
> Besides, if your corpus is asymptomatically everything, why wouldn’t fitting it perfectly fitting it be the goal? That's called Google, and it serves a different purpose. Perfectly learning the distributional properties of text in
316.
by Veedrac 4y ago
1. They aren't overfit. 2. It was far from obvious a priori that this could work at all. The existence of the sentiment ‘lol you trained a big model on X so obviously it produces a good probability distribution on X’ only exists beca
317.
by Veedrac 4y ago
There is plenty of clean energy, here on Earth, using existing technologies. There is even more than that if we go proactively grab a fraction of the the billions of galaxies of resources just meaninglessly burning themselves away.
318.
by Veedrac 4y ago
It is legitimately coming sooner than that.
319.
by Veedrac 4y ago
Perceiver AR does solve it that way, yes, albeit only for MIDI. Direct audio generation is still too long for it to keep a full piece in context.
320.
by Veedrac 4y ago
For context, here's some recent computer music, both MIDI and direct audio generation: https://magenta.tensorflow.org/perceiver-ar Here is an older project that generated audio from lyrics: https://openai.co
321.
by Veedrac 4y ago
My design just used grooves cut into the cylinder, which the two pin types would ride at different heights, so it didn't need to be any bigger. An insert, aka. a rectangular metal bar with cuts in it, is inserted into a slot cut down t
322.
by Veedrac 4y ago
I designed something like this before. Multiple shear lines with a two stage isolation-then-check is intuitive. The details matter, though, and this looks clean. I could see it manufactured. My design tried to avoid it being possible to int
323.
by Veedrac 4y ago
Lots of people are saying that. I am saying that. OpenAI has it as a foundational mid-term goal.
324.
by Veedrac 4y ago
You mean James Webb? That's a prime example of how we've failed. It's a decent $500m telescope with a moderate but not particularly ambitious scope, that overran its budget by a factor of 20 and launched fifteen years late. T
325.
by Veedrac 4y ago
That's only a counterargument if you're ignoring the value of coordination with the greater EA movement. I don't know why one would take that stance.
326.
by Veedrac 4y ago
Your question is the source of its own answer: if we want EA to be the most effective force for good it can be, we should make it something people will be glad to be a part of. It sounds like slight of hand, to say that things aren't o
327.
by Veedrac 4y ago
Wow, I am totally going to need to wait for more experimentation before believing any given thing here, but this seems like a big deal. It's one thing if DALL-E 2 was trying to map words in the prompt to their letter sequences and fail
328.
by Veedrac 4y ago
Those are all very different things. It is probably possible to argue that using a JIT would have solved some of Itanium's compilation issues, like it would make it easier to make compilation decisions about where to do software data m
329.
by Veedrac 4y ago
Disabling branch prediction would have such a catastrophic effect on performance, there is no way it would pay for itself. Actually this is true for most parts of a CPU; Apple's little cores are extremely power efficient and yet they a
330.
by Veedrac 4y ago
Itanium was a really badly designed architecture, which a lot of people skip over when they try to draw analogies to it. It was a worst of three worlds, in that it was big and hot like an out-of-order, it had the serial dependency issues of

More ›