FloatingUpstream

754 posts

FloatingUpstream banner
FloatingUpstream

FloatingUpstream

@FloatngUpstream

가입일 Ocak 2021
1.1K 팔로잉60 팔로워
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@badlogicgames Look at the proposal and timeline. Nothing of this is "finished" or will be concrete before 2028.
English
0
0
0
73
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@EU_Commission This discussion started at the beginning of 2025. And you are looking at 2028. This is too slow. Much too slow.
English
0
0
1
6
European Commission
European Commission@EU_Commission·
We are introducing EU Inc. To make building and growing a business across the EU faster, simpler, and smarter. 🔸 Start a company in less than 48 hours 🔸 No minimum capital requirement 🔸 Fully online and borderless
European Commission tweet media
English
609
1.2K
7.6K
2.3M
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@CostasAnest @peterrhague @Liv_Boeree Basic nociception is not suffering or pain. Speaking of which: crabs go beyond this and are amongst animals that can actually feel pain or pain-like experiences. I.e. they do not just act on stimulus reflexes.
English
0
0
1
17
Costas
Costas@CostasAnest·
@peterrhague @Liv_Boeree "Meaningfully suffer" does a lot of legwork here. We suffer from outside stimuli, hormones as those are interpreted in our brains. The stimuli and hormones exist in many animals. We don't know how they experience it mentally, but we can't be sure they don't suffer "meaningfully".
English
1
0
3
169
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@NerdShinobi @IceSolst Basically, a Miata is the bubbly, cute girl from school. An RX7, and to a lesser degree RX8, are her brothers. Both have massive issues with multi substance abuse, anger issues, and breakdowns. But the whole family is fun at parties.
English
1
0
2
40
solst/ICE of Astarte
Idk how to convey the emotional significance of a Miata to future generations
solst/ICE of Astarte tweet media
English
16
1
93
2.5K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@hkarthik @groby There were many early 2000s IRC channels that averaged around 10k concurrent users, with peaks beyond that. So what are you even talking about? This is absolutely not the big brained problem you imply.
English
0
0
0
42
Karthik Hariharan
Karthik Hariharan@hkarthik·
C'mon IRC was notoriously unreliable. Lag, consistency guarantees were garbage. You'd hop around 20 EFNET servers trying to find a good one. All us 90s kids remember it. Definitely was not a solved problem back then. We just accepted shittier solutions becuase we weren't paying for them.
English
3
0
16
2.8K
Karthik Hariharan
Karthik Hariharan@hkarthik·
10 years ago, it was considered technically impossible for a company like Slack to build a chat room that could serve 30K users simultaneously. Today this is a standard problem given in a system design interview in Big Tech for a Staff+ engineer.
Benjamin Booth@benjaminbooth

@GergelyOrosz Slack could not scale to 30K people in a single ‘townhall’ which had become a cultural hallmark at the time. HipChat was our only tool for eng on call at the time of our rocket growth but its mobile app would go down for multiple hours at a time, frequently.

English
19
15
660
129.4K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@NOTimothyLottes Tho, if you want to make more than the efficiency argument, I don't think morphological AA is really the best example. Static CNN graphs approximate that well and "reasonably" fast. Pick sdf raymarch if you want a harder case.
English
1
0
0
27
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@NOTimothyLottes I mean ok. You have essentially a logic/pattern matching algorithm. Tensor cores do dense math. Therefore tensor cores are the wrong hardware.
English
1
0
0
30
NOTimothyLottes
NOTimothyLottes@NOTimothyLottes·
Lets talk about the ML koolaid. Somehow it isn't obvious to sheeple that having {a fixed hierarchical network of fixed low precision matrix multiplies feed into a fixed non-linear operator} isn't necessarily the best most efficient solution to all of the worlds problems.
English
3
4
88
6.2K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@Gr8b8P @Old_Tobold @sleepy_devo That is a very narrow constitutional / US legalese view. E.g. you absolutely have an entitlement wrt emergency medical treatment under EMTALA. If you choose to call it a "right" is a purely semantic distinction.
English
0
0
1
72
wd
wd@Gr8b8P·
@Old_Tobold @sleepy_devo That’s correct. A right is something the government can’t take away from you.
English
2
0
0
239
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@sleepy_devo @Old_Tobold But even the US implements many positive rights / statutory entitlements that require policy and are implemented as law. The US just opts to not put them right in the constitution.
English
0
0
1
28
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@sleepy_devo @Old_Tobold It can be a positive right that requires / is state policy basically. Practicality of that position is another topic.
English
1
0
1
377
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@Old_Tobold @sleepy_devo Article 25 covers access to food. The US just doesn't do many legally binding things that are basically positive rights / require policy. If you look at the number of countries that ratified e.g. ICESCR and do fuck nothing concretely, that might be a sensible position.
English
0
0
0
166
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@NOTimothyLottes Is this ragebait? CNNs implement maxpooling explicitly, and e.g. RelU can be used to implement min max reduction indirectly in a trivial way?
English
1
0
0
30
NOTimothyLottes
NOTimothyLottes@NOTimothyLottes·
Ask an ML network to do non-linear logic (like MIN and MAX reduction trees that are common in things like morphological processing for AA), yeah those fixed matrix multiplies are just dead weight, also ML sucks at that, better to just use the min3 and max3 shader ops.
English
2
0
22
1.1K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@jstrocchi @VladVexler Yes, but if you predict the correct conclusion based on erroneous causal reasoning, your prediction is still just a lucky guess.
English
0
0
2
23
Jack Strocchi
Jack Strocchi@jstrocchi·
@VladVexler Mearsheimer argued Ukraine war was driven by UA preference to be a US client state through nato expansion, rather than neutrality as mandated: 1. 1990 USSR-US "not one inch eastwards" 2. 1991 UA constitution Istanbul Accords proved Mearsheimer was right, it was all about NATO.
English
5
0
1
154
Vlad Vexler
Vlad Vexler@VladVexler·
No, Mearsheimer did not predict the Ukraine war. Prediction has to be causally substantive. Mearsheimer said the war would happen for national security reasons. It happened for regime security reasons, supplemented by a dollop of mystical collective self-realisation.
Collingwood 🇬🇧@admcollingwood

@Talleyrand1738 He predicted in 2014 what would happen if the west did not change its policy. It didn't. What he said would happen happened. Which part of that is getting things wrong?

English
14
25
341
10.7K
J, 4%
J, 4%@J_B_N_P·
@Adriksh struct __attribute__((packed)) There. I just saved you a petabyte of RAM. You can thank me later 😎
English
1
0
4
596
Adriksh
Adriksh@Adriksh·
worrying about climate change when C++ struct padding is silently wasting 40% of the world's ram is wild.
English
53
78
2.9K
136.9K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@thdxr There is also an interesting amount of "technologists" (or some variation thereof), with an apparently deep resentment towards developers.
English
1
0
1
132
dax
dax@thdxr·
with every wave of tech the scammy part has the same message "see those winners? they're gonna be losers now and you're the one that's gonna be on top" saw it with crypto and now again with ai
English
37
36
704
26.9K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@dcolascione @reichswanderer Don't have a hard preference for either. I would much prefer a uniform solution, that is also closed. The downside of the dual approach is people claiming correctness, when they are really ignoring a complete error class.
English
1
0
1
13
Daniel Colascione
Daniel Colascione@dcolascione·
@FloatngUpstream @reichswanderer Indeed, but as you indicate, the closedness of a function's error channel is orthogonal to whether we spell that error channel as a sum type or as nonlocal control flow. I'm still of the opinion the latter leads to clearer programs.
English
1
0
0
37
Daniel Colascione
Daniel Colascione@dcolascione·
Few realize how badly Rust bungled its error handling. You pay for both error objects and exceptions without gaining the predictability of the former or the clarity benefits of the latter. The availability of panic=abort means lowest common denominator code. Then there's unwrap.
English
6
1
35
5.6K
FloatingUpstream
FloatingUpstream@FloatngUpstream·
@reichswanderer @dcolascione Effect systems in general. Several languages implement *something* like this, Java's checked exceptions are a very limited version. Monads come close, but are also not sufficient. Practically, Ada/Spark contacts are the most complete solution that I am ware of.
English
1
0
0
16
Nnanna
Nnanna@reichswanderer·
@FloatngUpstream @dcolascione these are actually very valid concerns. i don't know that rust will ever be able to fully implement this, because it involves reasoning about code outside rust. that being said, i don't really know any other language that implements this, you see?
English
2
0
0
25