spider-mario

4 ms·

31.
by spider-mario 1mo ago
It’s the same principle. The “trivial interval” from the first link (the one you called pedantry), which is applicable to a Gaussian and tweakable to have different coverage than 50%, clearly shows that being an n% confidence interval, on i
32.
by spider-mario 1mo ago
If you just meant it in the sense that “if I pick one interval at random and don’t look at it, it will (future) have an n% chance of containing the true parameter” then sure. It’s a pre-data statement (“I have an n% chance of sampling dat
33.
by spider-mario 1mo ago
It’s not reaching. See my response here: https://news.ycombinator.com/item?id=49622516 > Like saying the probability your uncle died of a shark attack is either 0 or 1. For a frequentist, it is. If you want the Bayesian
34.
by spider-mario 1mo ago
> Correct, which is why it's crazy to consider a Jeffrey's credible interval as valid, but not a confidence interval. The confidence interval is “valid” for what its definition is. But: - saying “there is a 95% chance that this
35.
by spider-mario 1mo ago
https://link.springer.com/article/10.3758/s13423-015-0947-8#... https://sami.boo/jaynes/confidence-intervals-vs-bayesian-int... (+ Cauchy example below) It’s like with test accuracy. Test acc
36.
by spider-mario 1mo ago
About the same amount of energy per image as a second of hot shower.
37.
by spider-mario 1mo ago
Yes. It’s maybe easier to reason about by imagining that we are trying to estimate the parameter of a Gaussian distribution. Let’s say that the “true” distribution of the data has mean μ=100 and standard deviation σ=15, but we don’t know th
38.
by spider-mario 1mo ago
MTPLX is this software: https://github.com/youssofal/MTPLX I tried it with the author’s 4-bit quant of Qwen 3.8 27B: https://huggingface.co/Youssofal/Qwen3.8-27B-MTPLX-Optimized... (but no need to
39.
by spider-mario 1mo ago
> Qwen3.8 27B seems like it was clearly supposed to be a high-end consumer open-weights model, but the t/s is so low for me on my old M1 Max 64GB that I hope others are getting use out of it. Have you tried it with MTPLX? I get arou
40.
by spider-mario 1mo ago
> Second, besides noise (bars are Wilson 95% confidence intervals, very conservative for run-to-run noise), there is little difference down to 4-bit; only the 2-bit scores a bit lower. Confidence intervals have nothing to do with run-to-
41.
Implementing FMA and finding bugs in C and Rust standard libraries(shnatsel.github.io)
3 points by spider-mario 1mo ago|0 comments
42.
by spider-mario 1mo ago
“your own website and then asking for donations […] business model” make it sound as if it’s for profit, which it isn’t.
43.
by spider-mario 1mo ago
No, I don’t think they’re intelligence either.
44.
by spider-mario 1mo ago
The post ends with: “Time to lawyer up...”
45.
by spider-mario 1mo ago
For what it’s worth, there’s an emoji dial in the settings. I’ve set mine to: Warm: Less Enthusiastic: Less Headers & Lists: Default Emoji: Less
46.
by spider-mario 1mo ago
I don’t follow your logic. “The gifted person is highly intelligent, just not good at deadlifting 500kg” does not diminish the 500kg deadlift and I wouldn’t expect an IQ test to measure it.
47.
by spider-mario 1mo ago
That sounds a bit like saying that the words-in-noise test is not a good hearing test because neither dolphins nor octopuses can speak.
48.
by spider-mario 1mo ago
Sometimes, it’s quite presomptuous and doesn’t search when it should. I’ve had to argue too many times with it that, yes, the Nintendo Switch 2 is real.
49.
by spider-mario 1mo ago
Starting at 0 is really only useful if you’re plotting a ratio measure and not just an interval one ( https://en.wikipedia.org/wiki/Level_of_measurement#Interval_... ), which I’m not sure this “intelligence index” is.
50.
by spider-mario 1mo ago
Nah, it reads much more like ChatGPT than like Claude to me.
51.
by spider-mario 1mo ago
> Claude is genuinely good. Better writing than most of the alternatives, Is that really how people feel? That’s not my subjective experience at all. I find Claude’s writing style “annoying”, for lack of a better description.
52.
by spider-mario 2mo ago
Right, but not a model you have already downloaded onto your machine.
53.
by spider-mario 2mo ago
Try 3.8 27B in MTPLX; I get about 30 tok/s with the same hardware as you. (Although it does use around 90-95W of power, compared to the ~60W that 3.6 35B-A3B uses to generate 55 tok/s. That’s about 3 J/tok instead of 1.)
54.
by spider-mario 2mo ago
Nokia buying Qt in 2008 brought an LGPL option.
55.
by spider-mario 2mo ago
It also means you can’t easily know the set of formats that the server could respond with, which in turns makes e.g. archival more difficult.
56.
by spider-mario 2mo ago
Most JPEGs are slightly smaller as progressive. The drawback is rather that it’s slightly more resource-intensive to decode, and if you overdo it and split the chroma, you can get weird effects: https://cloudinary.com/blog
57.
by spider-mario 2mo ago
macOS and Windows’ respective file explorers and built-in image viewers both support it, as do Photoshop, GIMP, Krita and a few more. https://en.wikipedia.org/wiki/JPEG_XL#Official_software_supp...
58.
by spider-mario 2mo ago
Two builds of the same source code don’t always produce exactly the same binary either, but we don’t use that as an excuse to distribute object files and call it open-source.
59.
by spider-mario 2mo ago
It’s not the “natural conclusion” at all; it’s a fallacy of relative privation.
60.
by spider-mario 2mo ago
The point is to parody the pathologisation of cognitive functioning that differs from the societal norm (see: “neuronormativity”). You’re just reasserting that pathologisation.

More ›