1- I got 7/10, jev got 6/10, and it says Jev won, presumably because I took 100 seconds, but Jev took 60? Funny what that reveals about author and AI-optimism in general
2- The chess board challenge isn't clear on how it should take the answer, I tried dragging a piece but that didn't work, then I tried touching the piece I wanted to move, but it marked the answer wrong, I apparently had to touch the square I needed to move the piece to.
Jev did not take 60 seconds for sure, with answers in 200-300 ms averages, more like 2 seconds.
You get very high scores if you are fast and accurate (which Jev is most of the time), and you get punished hard (15s) if you are wrong.
The game balances speed and accuracy.
Will try to improve the cards that didnt work for you, thanks for the feedback!
It's technically subjective, but to me speed is a tiebreaker, it goes after the decimal point. Being right is more important than being fast.
There's notably no button for "I don't know", but if there were, a human would maybe get 7 rights and 3 I don't know. While Jev would probably get 6 rights and 4 wrongs. I get that it's a game, but it's also not.
I think great minds make great efforts to not be wrong, someone that is never wrong is in a different category completely than something that answers fast, or even that answers more correct questions.
does this mean the UK will block 4chan in the UK? if the defense is jurisdictional then 4chan can continue to serve US users, but there would be no recourse if the UK blocks it in their jurisdiction
I think OFCOM themselves recognize the problem with trying to ban VPNs, but extremely dim MPs and extremely angry newspapers (who are the underlying enemies of free speech here) don't.
Using a nationwide default blocklist to address 4chan would be appropriate behaviour for Ofcom, given its duty to protect kids.
Beyond that, it is just madness to try to control the US company or fine it. Going beyond this duty shows that not all biscuits are present at the afternoon tea.
It's not like Facebook, or anyone's personal life or business, depends on 4chan.
This is my experience too, I don't know why they don't do IP level blocks. It's pretty trivial to evade a lot of bans.
I think the Spain-Cloudflare case does IP level blocks and it has a lot of side effect damage, but if I recall they do IP-block level blocking because the vendor, cloudflare, provides header/envelope encryption and pooling, so the Spanish courts considers it legal to block the whole service even, alleging that they are not collaborating with court orders, they just choose not to block the whole provider to reduce the collateral damage, even if they consider they have jurisdiction to block all of CF
Pirate streams have better access to public IP addresses than legitimate services, so blocking Cloudflare does nothing but destroy legitimate businesses.
4chan and Kiwi Farms, which is something the article intentionally glosses over. Honestly, KF is a far more “dangerous” place in the eyes of the UK government, as 4chan hasn’t been relevant for around a decade, and the combined efforts to deplatform KF traditionally has been fruitless.
deep packet inspection is not really necessary, headers contain information about the destination, as long as you don’t encrypt those, there is no need to do any decryption.
You just need to follow this to the logical conclusion. They’ll ban using your proposed method, people will switch to VPN and other modern methods and to combat people still accessing the forbidden internet they will need DPI and TSPU-like devices
I think the conclusion is more like mandating home and business web access is proxied at a national level and most other protocols disallowed. Use of VPNs without a VPN tax stamp and key escrow is prima facie evidence of violation of the Counter-Terrorism and Security act. Use of non-standard protocols is assumed to be criminal intent. Check with your approved internet provider for how to access the Internet safely.
What proposed method? I'm explaining how things work. In order to send a packet, the destination address needs to be known to the delivering entities. Encryption occurs by using enveloping, the envelope is unencrypted, but the content is encrpyted.
This is how it works, if you want I can explain to you why it works like this, or you can work it out from first principles. But this is not a proposal.
VPNs are an asterisk and they work not by avoiding this natural unavoidable principle, but by tunneling, the VPN or proxy vendor is putting themselves on the headers, and on the backend they still route the packets to the destination by putting the intended destination on the headers.
If these VPNs don't comply with local regulations and are used to evade national restrictions they can and are often themselves subject to bans. But even if not, it acts as a sort of waiver and compromising element for all illicit activity.
Sometimes it's not necessary to completely outlaw something, it is sufficient to make it so that engaging in the outlawed activity is sufficiently uncomfortable and clear that it is illegal, forcing users to undeniably acknowledge they are avoiding the law. An example of this is trying to watch pirated movies without paying for them, yeah you can do it, but you gotta do some shady shit and dodge a lot of malware, and eat shady gambling ads and etcetera.
In addition to making you feel like an asshole, it removes any recourse you may have, it's very hard to make any claim later, you cannot really complain later if you were accessing it illegally to begin with. Nobody is going to pursue you to the end of the world not to access a pirated movie or an illegal website, but they pursue the distributor and make things very uncomfortable enough so that you are effectively hiding like a fiending outlaw.
>1. This is a 2022 era picture that LLMs just suck up what's on the Internet and regurgitate it. It hasn't been true for a long time
if anything it’s more true now, AI chatbots rarely use raw LLM output, instead they have CoT and use search tool calls. and the speed at which webpages are crawled is in the matter of hours. I posted something here on hn, then searched for it on google 1 hr later and its AI answered with my comment
there’s several companies right now making a killing of organic placement on chatbots similar to the SEO rush of early google
> 2) There's almost no one who has a stronger incentive to distinguish LLM generated text from non-generated LLM text. Training classifiers is pretty manageable
i think it’s pretty naive to think that if LLM companies would develop LLM detection they would use it to skip training on that content.
instead, they use it to train their output to be less detectable.
LLM detection is not solvable because it is adversarial, an arms race
falsifying statistics alone without the added stealing on top, was 3 years of labor camp in places. that was after they course corrected and stopped giving capital punishment for it.
good thing nobody falsified statistic just for the sake of it -- so being a sabotager and a foreign spy was always part of it too.
Is this categorization actionable or meaningful in any way? If a system or process relies on no one on the internet being a liar, it's broken. There are liars. There will be more liars. As is made evident, people (and machines) will unwittingly reproduce lies out of sheer incompetence. That reflects poorly on them as reliable sources of information.
I feel this is answered by Free Software philosophy 101.
Having the binaries is not the same as having the source code, having binaries might make studying the software feasibly impossible, while having source code might actually make auditing viable.
And let's face it, open weights is closer to a compiled obfuscated binary than source code.
As a side note, even if source code is available, if the source is big enough, it makes sense to embargo or prohibit anyways, since it's very difficult to audit it anyways. Especially if problems are already found.
That's still not source code according to Stallman's GPL definition.
Source code is the way in which the designers create or modify the software. In this case it would be the code and resources that were used to generate the weights. Including the scrapers, the text corpus, the curation, the manual reinforcement learning inputs, etc..
reply