Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPU World

gpuworld.org

Post reply on HN
Print view View original

241–250 of 303 posts

Re: GPU World

#241

Post by Gareth321 »

munksbeer wrote: post #235
Gareth321 wrote: post #195

Earlier quoted context omitted.

My codex tasks regularly cross 8 hours, and I'm only using Sol High. It's not unusual for tasks to span much longer. It just requires instructions to continue working until the spec is complete. Anthropic and OpenAI are currently obsessed with getting humans out of the training and improvement loops. It's going to happen very soon and when it does I think we see staggering improvements in a very short space of time.…

Are you breaking up your tasks and spawning new sessions for each, or are you just yoloing and letting it auto compact when it blows the context many times on such a long running task?

Yolo. I keep a large project markdown document + incident/logs/feature documents which its instructed to review and update as necessary. It still occasionally misses stuff but it's surprisingly effective. Disclaimer: this is for hobby software. For work I'm more cautious - usually.

Re: GPU World

#242

Post by gwern »

sudohalt wrote: post #79

Will any of the data submitted be used in training? We are all given the prompt and you will get a few thousand (maybe tens of thousands of submissions). Seems like a great source of training material.

We do not intend to train on them, no. This never even occurred to me to address in the rules. (What would be the point? There will be few good submissions and they would be a vanishingly small % of existing training data and this doesn't seem like a super-valuable task in general that would make a LLM that much more valuable...)

Personally, I would encourage participants to publish their pieces independently. Note that we only require a non-exclusive CC-BY-NC license for the ones we select as winners, which is intended to allow people to republish the good pieces themselves, especially commercially.

Re: GPU World

#243

Post by aand16 »

chrismorgan wrote: post #222
zamadatix wrote: post #221

Earlier quoted context omitted.

100k prize announcement for a competition right up a HN writer's alley and the top discussion on it is how the announcement page doesn't load correctly on a browser with half the features disabled. We are truly a special group :D.

It’s currently at +39. I did not expect that! I expected it to settle at the bottom like something that settles at the bottom—I dunno, detritus, lees, Titanic, something like that. Still, I’ll take the opportunity to advocate for turning off fonts. Firefox: Settings → Fonts → Advanced settings → Allow pages to choose their own fonts, instead of your selections above . I’ve been using the web so for 6½ years and it ma…

+40 I hate websites that do stupid stuff. My latest pet peeve is Qwen's blog, when I pinch to zoom, to be able to read their tiny graphs, there is a black bar that progressively covers the entire page.

Re: GPU World

#244

Post by hansvm »

zamadatix wrote: post #221
chrismorgan wrote: post #149

This website has the distinction of being the first basically-just-content website I have ever encountered to break when web fonts are blocked (or otherwise fail to load). Other than the weird 5×5 colourful tile image thing near the top, all the content is in the document , but they go out of their way to break it in various baffling ways: ① if JS is disabled, it hides all the content behind a “Please enable JavaScri…

100k prize announcement for a competition right up a HN writer's alley and the top discussion on it is how the announcement page doesn't load correctly on a browser with half the features disabled. We are truly a special group :D.

I have a fully functional, updated Chrome derivative with all the features enabled, and they went out of their way to dynamically load a

  Update your browser to access this website
message as the only visible content rather than test features (which they're clearly doing to GP, so how much work would it be to properly use that feature testing as the main tool for compatability checking) or just letting the browser try to load it -- displaying one of those big-tech "update your browser now" nag messages if they really feel the need to be annoying.

Re: GPU World

#246

Post by hansvm »

chrismorgan wrote: post #222
zamadatix wrote: post #221

Earlier quoted context omitted.

100k prize announcement for a competition right up a HN writer's alley and the top discussion on it is how the announcement page doesn't load correctly on a browser with half the features disabled. We are truly a special group :D.

It’s currently at +39. I did not expect that! I expected it to settle at the bottom like something that settles at the bottom—I dunno, detritus, lees, Titanic, something like that. Still, I’ll take the opportunity to advocate for turning off fonts. Firefox: Settings → Fonts → Advanced settings → Allow pages to choose their own fonts, instead of your selections above . I’ve been using the web so for 6½ years and it ma…

It probably just literally doesn't work for most people. If you click through, see a broken page, and see that the top comment thinks the way in which it's broken is interesting, that'll probably be more interesting to engage with than the contents of the page.

Re: GPU World

#247

Post by Gareth321 »

bwfan123 wrote: post #217
Gareth321 wrote: post #195

Earlier quoted context omitted.

My codex tasks regularly cross 8 hours, and I'm only using Sol High. It's not unusual for tasks to span much longer. It just requires instructions to continue working until the spec is complete. Anthropic and OpenAI are currently obsessed with getting humans out of the training and improvement loops. It's going to happen very soon and when it does I think we see staggering improvements in a very short space of time.…

> It just requires instructions to continue working until the spec is complete Try putting an LLM agent in a deterministic workflow without humans in the loop. My experience with this is not encouraging. Getting it to work requires sprinkling some context magic and hoping and praying the LLM does the right thing. More astrology or religion and less science. Great for use cases with humans-in-the-loop, but less than i…

> Try putting an LLM agent in a deterministic workflow without humans in the loop.

I would not use a sewing machine to repair my deck :) LLMs are non-deterministic, by design.

Re: GPU World

#248

Post by hansvm »

re6tor wrote: post #245

Do submissions need to be in English? If so, that'll prevent a huge chunk of us from participating.

I'd be surprised if the 3 judges share a common language other than English at the requisite level for judging writing.

Re: GPU World

#249

Post by hansvm »

I'm not sure what else might happen, but at a minimum the Babel Fish will finally be real, even without internet.

Re: GPU World

#250

Post by david-gpu »

HarlequinHair wrote: post #188
ipsod wrote: post #177

Earlier quoted context omitted.

> I don't know if anybody consuming it really wants to. You don't know of people who ask LLMs questions? Or just chat with it for fun and to learn? Maybe your electrician isn't interested. A pharmacist and a nurse I talk to are. Virtually everybody I've talked to uses it to some degree or another. I think it'll soon replace Alexa/Siri for most people, as well as google search. That's pretty far from nobody.

I think the first parent post all this conversation originated from, pointed that LLMs are not groundbreaking inventions. If people use it to replace voice assistants or search engines, you can see how this new invention is not really adding anything you already hadn't. AIs in general are very useful, but apart from the campaign they did to change everything's name into AI (also well known algorithms and "old" machin…

If people use [the Internet] to replace [Sears catalogs] or [paper mail], you can see how this new invention is not really adding anything you already hadn't.
Post reply on HN