Ukv

66 ms·

841.
by Ukv 3y ago
> More importantly, most users seem to want the subs reopened Hard to prove one way or the other at the moment since the subs that vote to go private are now inaccessible, but almost all polls I've seen have been in favor of the bla
842.
by Ukv 3y ago
> 1. As mentioned, this article is about Full Self Driving Beta Pro Black Edition The article does mention both. Notably it uses autopilot+FSD crash count (but then only divides by the miles driven using FSD).
843.
by Ukv 3y ago
> Don’t they know these models fail horribly with math especially? > But I’m right though right HN? These models seem exceptionally bad at math questions. If you're asking "find the error in my proof" or "solve thi
844.
by Ukv 3y ago
> the conmen who are selling ChatGPT and the like are extremely irresponsible for the way they sell LLMs as magical AI that arrives at factually correct answers ChatGPT has a pop-up on first use, a warning at the top of each chat, a warn
845.
by Ukv 3y ago
> This is essentially a ban on removing AI content I think framing it as "a ban on removing AI content" is wrong/misleading and the reason for softwaredoug's question about whether you'd even be allowed to edit l
846.
by Ukv 3y ago
> The issue is if they show up as part of some meme chat and then you as the LLM creator get sued because people start harassing them. This seems a more obscure concern than extraction of data. > copying is a basic part of how these a
847.
by Ukv 3y ago
The point of the retort is precisely to point out the meaninglessness when "regurgitation" is extended to include systems with some degree of generalization/extrapolation/discovery capabilities. The expectation set on AI
848.
by Ukv 3y ago
There's no ban on removing AI content as far as I'm aware - admins just upped the evidence requirement for determining a post as AI-generated, seemingly in response to high false positive rates of hunches and current detectors: &g
849.
by Ukv 3y ago
Privacy, where there's a problem if some original data can be inferred/made out (even using a white box attack against the model), is a higher bar than whether an image generator avoids copyright-infringing output under non-advers
850.
by Ukv 3y ago
> AI models will make 1:1 copies of training data where artists [...] In general I don't think this is the case, assuming you mean generations output from popular text-to-image models. (edit: replied before their comment was edited
851.
by Ukv 4y ago
> Computers can't learn abstract concepts Goalposts can be moved on whether it has "truly learned" the abstract concept, but at the very least neural networks have the ability to work with concepts to the extent that you c
852.
by Ukv 4y ago
I don't hear people say "July 4" either, but you do hear both with the ordinals as in the examples I gave.
853.
by Ukv 4y ago
"4th [of] July, 2023" and "July [the] 4th, 2023" are both valid - I'd bet there's a correlation between spoken order and written order, but not sure which caused which (or maybe mutually reinforcing).
854.
by Ukv 4y ago
> So I asked ChatGPT, “Can you provide a link to the Guardian obituary?” expecting it to finally give up the lie but to my concern it went even further: > > Yes, here is the link to Alexander Hanff's obituary on The Guardian&#
855.
by Ukv 4y ago
> In August 2020, the two of us wrote an article for Technology Review, “GPT-3, Bloviator” where we illustrated GPT-3’s limitations with 6 examples that showed GPT-3 making laughable errors in commonsense reasoning. GPT by default is a g
856.
by Ukv 4y ago
In the event that Getty Images wins, it seems most likely that AI researchers would pay Shutterstock/Getty Images for their large existing catalogs of images. With the companies having a stronger position (getting to act as a gatekeepe
857.
by Ukv 4y ago
> They're supposed to be better at driving than humans. And that actually means all of it Why? Couldn't it be "better at driving than humans" by reducing the most common errors (lapses in attention), while performing
858.
by Ukv 4y ago
I'd rather not risk stunting progress in areas like language translation, malware scanning, DDOS prevention, spam filtering, product defect detection, scientific data analysis, autonomous vehicles, voice dictation, narration/text-
859.
by Ukv 4y ago
If training machine learning models is Fair Use (not totally settled, but seems likely) then they aren't bound by the license of the original content. Setting up robots.txt is probably your best choice. In the EU commercial text and da
860.
by Ukv 4y ago
> There’s a phenomenon called “chain of thought” where large AI language models perform much, much better if given prompts like “Think about it carefully and show your steps.” Mathematical word problems that trip up ChatGPT normally get
861.
by Ukv 4y ago
> You’re missing that the incident in question was days after the jet landed at the airport. DAYS. There was no following from earlier that day. I could be making a mistake but I don't believe this is true. Are we looking at the sam
862.
by Ukv 4y ago
> The incident happened in a car I believe the car was followed from the jet (possibly after the car dropped off Musk, or collected Musk's son from Musk), which was at Los Angeles International Airport earlier that day. The car itse
863.
by Ukv 4y ago
> It was days after and nowhere near the airport. > [Other comment:] days after. so by musk's own rules.. fine to post. wasn't real time. Location of the jet was shared in real-time to my understanding, checking with the lin
864.
by Ukv 4y ago
> So what would banning a jet tracking account have to do with a stalker From what I can gather and infer, a couple of days ago Musk's son got off the jet and into a car, then that car was attacked by a stalker looking for Musk hims
865.
by Ukv 4y ago
If the code is ostensibly available under permissive licenses, a well-intentioned human will also make the same mistake.
866.
by Ukv 4y ago
I think that's confusing between two issues in opposite directions: 1. Accusations that AUTOMATIC1111 (the web frontend developer) copied code from the NovelAI leak relating to the loading of hypernetworks 2. The leak revealing that An
867.
by Ukv 4y ago
> Do people really care that it kind of looks like they’re betting or it kind of looks like they’re really holding their cards, or do they just want to play poker and interact with people? I think people do care about factors that add to
868.
by Ukv 4y ago
> CoPilot changes the kind of work the engineer has to do, from authoring to editing, but the amount of work is not reduced. Disagree based on my own experience using CoPilot, but it would be interesting to think about ways to fairly tes
869.
by Ukv 4y ago
Record for PG-13 movies is apparently Gunner Palace with 42 uses of "fuck". But more to the point, Youtube is not the MPAA. The documentary seems to comply to Youtube's publicly given rules more so than a huge number of other
870.
by Ukv 4y ago
I agree that the main problem is the fact that this rule is uncommunicated. As the creator of the video in the article said, they'd be fine with adjusting if they could. Instead, they complied with the publicly given rules and in retur

More ›