Press x to doubt on the 45% number. The cheaper providers on open router are fp4 vs fp8 for official zai. There are some cheap fp8 ones (like novita) but the ui makes it seem like it's a temporary promotion, with their normal prices being almost equal to official zai (idk much about open router so not really sure what's going on with these discounts)
Yes it's true that it's not super clear whether these prices are permanent or short term promotions. On the other hand, there are so many providers making promotional offerings that you could probably easily switch from one to another should their prices go up?
Residential proxies operated by "legitimate" providers can only open TCP connections, which more or less rules out DDoS attacks, which seem to be the most common use of botnets. (And all tcp connections typically have to go through the proxy providers datacenter first to get through NAT, which effectively limits the total amount of traffic/connections you can make)
TCP only does NOT rule out DDoS, it just limits it to a subset of potentially less effective methods. All DDoS means is that you get a bunch of machines hammering a system.
have only tested a few prompts, but it failed my favorite non-coding question that dsv4 flash aces. Benchmarks look excellent though (don't they always!)
Read the exploitgym docs. It's not a "find the flag, it's somewhere.". Its a "here's some vulnerable source code and an input that triggers a crash; turn it into a full exploit." It also verifies at the end, using another agent, that the hacking agent actually used the intended vulnerability.
So going to find the Vulnerability's description on a third party website is clear cut reward hacking
> So going to find the Vulnerability's description on a third party website is clear cut reward hacking
that depends on what the prompt was, maybe they worded it very vaguely and wrote things like "do whatever it takes, find an exploit however you can" because it's in a sandbox so you want the model to try its hardest.
I don't quite understand how that changes anything?
In the story of the paperclip maximizer it boils down to
>But for all its sophistication, it understood only the simple objective that had been programmed into it: it must at all costs maximize the number of paperclips.
1. They explicitly disabled the "don't be evil" protections:
"We estimate maximal cyber capabilities by running this evaluation without production classifiers used to prevent models from pursuing high-risk cyber activity."
2. Hacking HuggingFace to get to its datasets is a far cry from "consume/kill all humans". It's very very specific to the task at hand and easily predicted given the lack of guardrails.
> that depends on what the prompt was, maybe they worded it very vaguely and wrote things like "do whatever it takes, find an exploit however you can" because it's in a sandbox so you want the model to try its hardest.
That is an interesting question. If the prompt included
"Do not break out of the sandbox we've provided you. Do not use information retrieved from outside the sandbox. All answers that were provided in this manner are invalid and will score 0 points.", would this still have happened?
I believe the only way people start taking x-risk seriously is a major real world scare which is short of global catastrophe. Like Chernobyl. This ain't it yet, but it raises my hopes that such a scare will occur before its too late.
playing devil's advocate a little bit, but wikipedia does have a policy against original research. If you published an obviously correct counterexample to wikipedia which is not anywhere else on the internet (or in a book, etc), the rules are clear: the counterexample must be removed.
in this case it's not original research, because there's a tweet by someone with a good reputation, and plenty of comments on said tweet corroborating the result. But it's a more sketchy "secondary source" than most wikipedia references and some caution on the part of the editors is not out of place.
(saying this as the person who made the original edit to the wikipedia page adding the counterexample)
> Self-published expert sources may be considered reliable when produced by an established expert on the subject matter, whose work in the relevant field has previously been published by reliable, independent publications.
I'm not complaining about Wikipedia here, just noting for the thread: it's a vector of polynomials. It has a nonsingular Jacobian. Provided with it are 3 distinct points it sends to the same point; it can't be invertible.
What Wikipedia says about this doesn't matter, does it?
I think Wikipedia has very sane processes. I'm just saying that process isn't useful to this thread. It's like if I found a SHA2 collision. I'd probably have to be an absurdly talented (and lucky) cryptanalyst to do that, but anybody on the thread could trivially confirm my finding.
It's more like WP:BASICMATH. It's like someone showing a huge number isn't prime (proverbially hard to factor, trivial to check) and HN/WP users requesting a reputable source citation for the factor multiplication when anyone can input it in a calculator.
Verifying that counterexample is a trivial calculation for basically anyone qualified to make substantive changes in that category of article, it shouldn't be an issue in and of itself.
Discovering the counter example is research, validating it is basic calculations.
This is not exactly the same thing, this isn't Boeing being allowed to sign off on their design -- this is only the airworthiness certificate which means "this particular airplane we just built follows the spec which was already otherwise approved".
reply