vitus

25 ms·

301.
by vitus 2y ago
11 minutes in, he mentions SICP as a book that he ended up working through exercises "in the past year" (so 2012 timeframe, well after the original Doom but before the reboot).
302.
by vitus 2y ago
This sounds like a more complicated (and simultaneously under-analyzed) version of the Princess Bride iocaine powder scene. Assigning probabilities doesn't really make any sense in this context, since you're relying on the trustwo
303.
by vitus 2y ago
Ah. I had focused on the lossiness of natural language much more than the rest of your statement. > Why should it be that if people observe this it becomes obvious, and yet if put in natural language it isn't? It's not clear to
304.
by vitus 2y ago
> I'm fundamentally sceptical that animals have, in general, a "poor" understanding of probability. I take the opposite view. Casinos with all of their negative-expected-value (for players) games would not exist, nor would
305.
by vitus 2y ago
> “How many times do you have to roll an n-sided die to ensure you’ve seen each side at least m times?” This is an extension of the coupon collector's problem (which addresses the m=1 case). Per the Wikipedia page: > Donald J. Ne
306.
by vitus 2y ago
From the intro of your first link: > While conversion from a string into an integer value is feasible with SIMD instructions, this application is unpractical. For typical cases, when a single value is parsed, scalar procedures — like the
307.
by vitus 2y ago
Can you point to some of these benchmarks? https://news.ycombinator.com/item?id=26934854 suggests that in at least one synthetic benchmark (with a 7.5KB protobuf message which expands to a 17KB JSON payload), protobuf parsi
308.
by vitus 2y ago
Did you mean to ask the reverse question (in what context is protobuf slower than json)? Because that's definitely the question on my mind, since GP's assertion runs counter to my expectations and experience. JSON is a heavy-weigh
309.
by vitus 2y ago
> Also, the pawn can’t deliver checkmate, can it, if it can be taken en passant? It probably is possible to construct a position where taking en passant would bring the king into check in another way, but in those cases, the en passant m
310.
by vitus 2y ago
You're not wrong, but it's still a nominally usable lower bound for the actual power draw of the chip, and a reasonable proxy for how much heat you need to dissipate via your cooling solution.
311.
by vitus 2y ago
That is a fair point, as the 2x CPU + 4x NIC are "only" about 550W put together. There's probably more overhead for cooling (as much as 40% of the datacenter's power -- multiplying by 1.5x pushes you just over that 800W
312.
by vitus 2y ago
Note that your power consumption is more than just the CPU (combined TDP of 2x225W [0]). You also have to consider SSD (16x20W when fully loaded [1]), NIC (4x24W [2]), and the rest of the actual system itself (e.g. cooling, backplane). [0]
313.
by vitus 2y ago
> Abseil, Boost and Folly all provide types which lose some performance to get this back if it's all you need to be "drop-in". Indeed, Google's initial migration from std::unordered_map to swisstables moved everyone t
314.
by vitus 2y ago
> These days, folks don't come to C++ from C and are not used to such under-the-hood tricks. It's funny to see that, since I'm used to folks coming to C++ from C and bemoaning that C is a much more straightforward language
315.
by vitus 2y ago
If there's any risk of this happening, it's probably because someone imported AGPL code into the codebase, outside of //third_party, after removing the copyright notice / license. (We have an allowlist of acceptable
316.
by vitus 2y ago
> using [] on a map to access an element is really really dangerous if you haven't read the docs carefully and assumes that it does what you believe it should do It's not that bad, as it just inserts a default-constructed eleme
317.
by vitus 2y ago
I could envision some point in the future where the tooling is good enough that if you reject the suggestion, you had better have a very good reason for doing so. We're already there on the formatting front, as well as the widely-enabl
318.
by vitus 2y ago
> That’s a list of things that gets clicks in the popular press. Are you saying they're nonissues in practice? I agree that most of those points have shown up in the news, but they're also things that I have personally observed
319.
by vitus 2y ago
From the screencast: > implement also for Days This fails to recognize that this is a bad feature that the Abseil library would explicitly reject (hence the existence of absl::CivilDay) [0], and instead perpetuates the oversimplification
320.
by vitus 2y ago
> 1. All the things it can obviously do very well today I'm curious what those things are. At least to me, it isn't obvious that LLMs solve any of their many applications from the past year "very well". I worry about
321.
by vitus 2y ago
> Just waiting for a random US future president to create an "Internet" branch of the military. Maybe that's already happened. https://en.wikipedia.org/wiki/United_States_Cyber_Command is the closest
322.
by vitus 2y ago
Notably, the 2024 list was released just yesterday, so I wouldn't fault anyone for not having seen that yet. Texas was in the lead for a couple of years, but honestly it's so close that I don't think it's interesting to
323.
by vitus 2y ago
For what it's worth, Wikipedia cites the 2021 list. The 2023 list had Texas at 55: https://www.statesman.com/story/news/local/2023/06/05/texas-... edit: another commenter points out that C
324.
by vitus 2y ago
> So 8GB machines are still around, and not just in the throwaway $200 laptop segment. You're not wrong here! I entirely agree that 8GB laptops are still very much a thing, and RAM hasn't been growing at quite the same rate tha
325.
by vitus 2y ago
> raw peak performance hasn't moved much in a decade. In general, CPU clock speeds stagnated about 20 years ago because we hit a power wall. In 1985, the state of the art was maybe 15-20MHz; in 1995, that was 300-500MHz; in 2005, w
326.
by vitus 2y ago
> Today, it fails badly at tasks that require reasoning. A simple example: https://chatgpt.com/share/da95843e-218a-4d69-a161-6aa2d7a3c9 ... > The issue is that humans can see its answer is wrong and its "reas
327.
by vitus 2y ago
Indeed, that's what `touch -t 201101020304 /tmp/old-file` is for. (Although on ext4 it seems that you can't control the birth time; for that you would need to set your system clock.)
328.
by vitus 2y ago
> The "hallucination" factor means every result an AI tells you about big data is suspect. AI / ML means more than just LLM chat output, even if that's the current hype cycle of the last couple of years. ML can be use
329.
by vitus 2y ago
If you're one of the public clouds targeting SAP use cases, you probably have some machines with 12TB [0, 1, 2]. [0] https://aws.amazon.com/blogs/aws/now-available-amazon-ec2-hi... [1] https://clou
330.
by vitus 2y ago
Do we even have a second C++20-standard-compliant module implementation? Last I checked, it was just MSVC -- GCC and Clang were like 90% but missing a few features. Looks like still no, although Intel provides yet another partial implementa

More ›