coherentpony

8 ms·

31.
by coherentpony 1y ago
> Converting complex HTML pages with navigation, ads, and JavaScript into LLM-friendly plain text is both difficult and imprecise. Oh my god. It just occurred to me that LLMs may have a better experience “browsing the internet” than hum
32.
by coherentpony 1y ago
I did an in-person class. I highly recommend it.
33.
by coherentpony 1y ago
> Real-time accent feedback is something language learners have never had throughout all of human history, until now. Do you have a source for this? It doesn't seem plausible to me, but I'm not an expert.
34.
by coherentpony 1y ago
Ah yes. Ok. Yes, that makes perfect sense. Thanks.
35.
by coherentpony 1y ago
> Correct. Which means that every 20ms pixel slices two or three frames. Which is a really really bad way to profile! If 20 ms is a reasonable frame time for a modern game, why is it an unreasonable thing to profile? I understand other,
36.
by coherentpony 1y ago
> Ain’t no game using 20ms frames! A frame every 20ms equates to 50 frames per second. Doesn't seem too unreasonable for a modern game. 60 frames per second would be one frame every ~16 ms.
37.
by coherentpony 1y ago
It shouldn't be "an idea" at all. Profile the application, find the hotspots, understand what the performance limiter is, and fix it. Granted, this is all Hard Work. I understand that. But it's the right thing to do.
38.
by coherentpony 2y ago
Yes, I meant the RDMA verbs API. Thanks for the insight.
39.
by coherentpony 2y ago
> everyone knew the ways in which that was a bad idea The state machine pattern is used currently in verbs. That was the state of the art for many decades before higher level interfaces appeared. Why was it a bad idea? Can you elaborat
40.
by coherentpony 2y ago
> we’ve built our most advanced image generator yet into GPT‑4o. The result—image generation that is not only beautiful, but useful. Sorry, but how are these useful? None of the examples demonstrate any use beyond being cool to look at.
41.
by coherentpony 2y ago
Here's a timelapse of the aurora: https://science.nasa.gov/asset/hubble/hubble-time-lapse-of-a...
42.
by coherentpony 2y ago
It bothers me that the word 'hallucinate' is used to describe when the output of a machine learning model is wrong. In other fields, when models are wrong, the discussion is around 'errors'. How large the errors are, th
43.
by coherentpony 2y ago
It likely will never be an option. You could consider spoiling your ballot instead? I have no idea if spoilt ballots are counted in America, though.
44.
by coherentpony 2y ago
> I need a how-to right now and all I get is an explanation. This isn't a reasonable expectation. Your current state of documentation may be very different from some-other-software's current state of documentation. There may
45.
by coherentpony 2y ago
> Responses referenced the contributions the development of the infrastructure supporting the basic research itself have made, which is fine, but not the benefits of high energy physics discoveries. I was one of those responders. There w
46.
by coherentpony 2y ago
https://en.wikipedia.org/wiki/CERN#Scientific_achievements Here's a couple, in case you don't want to read the page: - CERN pioneered the introduction of TCP/IP for its intranet, beginning in 1984 - CERN
47.
by coherentpony 2y ago
https://link.springer.com/article/10.1007/s10676-024-09775-5
48.
by coherentpony 3y ago
The first paragraph of the preface on page 7 in the book gives an explicit example of “higher order”: https://terrytao.files.wordpress.com/2011/03/higher-book.pdf
49.
by coherentpony 3y ago
> Nice, just like in the good old days with serfs and feudal lords. I can't tell if you're being sarcastic or not. Serfs were largely agricultural workers bound under the feudal system to work on their lord's estate. Fore
50.
by coherentpony 3y ago
> I think sometimes we forget how fuzzy language is. I'm willing to bet a reasonable sum of money that the folks that designed the dark pattern didn't forget how fuzzy language is. In my opinion, the entire spirit of the desig
51.
by coherentpony 3y ago
No, I'm a software engineer.
52.
by coherentpony 3y ago
> They're not lying to you That's correct. But it's correct in the sense of "technically correct". It's a language-lawyer type of correct that nobody actually expects when communicating in plain English. Y
53.
by coherentpony 3y ago
> I am sure some AMD folks are on HN The person you're responding to is an AMD employee.
54.
by coherentpony 3y ago
Do you realise that doing nothing is also making a choice about the state of the future?
55.
by coherentpony 3y ago
Well, I can't control the past. My options are the present, and the future, and the latter depends on the former. So what are my options?
56.
by coherentpony 3y ago
https://mlir.llvm.org/
57.
by coherentpony 3y ago
Disclaimer: I have no idea how machine learning algorithms work. I work with problems so huge they do not fit on a single device. Multiple devices each own a small piece of the global problem. They solve a local problem and they must com
58.
by coherentpony 3y ago
What do you mean? Are you talking about capacity? Or bandwidth from RAM? I'm in the HPC space, and pretty much everything I do on the GPU is bound by how quickly I can get data in and out of DRAM. The point at which data motion to&#x
59.
by coherentpony 3y ago
One of the many idiosyncrasies of the English language. The sentence > Don't make fire rings where they already exist means both: 1. If there is an existing fire ring, do not re-use it. Make a new fire ring somewhere else instead;
60.
by coherentpony 3y ago
> 1 exaflop To be clear, this is floating point quarter-precision operations when using the FP8 tensor core arithmetic unit [1]. [1] https://resources.nvidia.com/en-us-grace-cpu/grace-hopper-su...

More ›