Some updates:
- AI solved A, B, C, D, E, F, and H in this round (reference: bugfeature). A cheater that gave the problem statements to GPT quickly would have placed at rank <= 4 in this round. I'm not sure about G, as most cheaters had already been banned by the time I checked, but looking at JohnNash01, it seems plausible that AI might have also solved it (which would imply the first(?) div-1 AK by AI, and also rank 1 for a fast enough cheater).
- Tourist has now been defeated by AI cheaters in the two most recent div-1s. GPT has turned out to be the real conqueror_of_tourist we made along the way.
- In an entirely unrelated turn of events, Panda_V is the new Indian GOAT, swatting aside Dominater069. We're being ushered into an unprecedented golden age of Indian competitive programming.
I think copes like "just get to 2100, there aren't any cheaters in div-1" are clearly invalid now. Ratings don't have much value any more, and will probably lose the little they still do in the coming months.
i think we should just wait
system testing and skipping isn't over yet
https://mirror.codeforces.com/blog/entry/146629#comment-1311335
I'm obviously not blaming Mike or any of the Codeforces staff here, as it's essentially impossible to prevent AI cheating at scale in the long term (and anyone who deludes themselves by dreaming up comically contrived countermeasures is coping). This post is merely meant to acknowledge the observable progress in AI capabilities and the breaching of important milestones that were once thought to be unbreachable.
hmm... and i think cheating is a problem which can't be prevented at this point
there are a lot of cheaters on cf, and they will continue to do so, and one day they will stop cheating after realizing companies dont give a f..k to their rating, and then they will leave cf forever
even i was a cheater approx 8 months ago, but after sometime i just lost interest in cheating, created new account (this one) and tried solving problem again, and trust me the happiness of solving a really hard problem is unmatched
so they will never be able to feel the dopamine rush we go through, and that is only benefit we have
dont forget that there are people who cheat a little (implement the idea given by ai)
Well, as ratings mean less, people would be less incentivized to cheat, eventually there should be some kind of balance. We are already seeing fewer participants (this round only has 17k). Hopefully by the end we get a small group of people who truly love competitive programming to stick around.
yeah and they've been invalid for at least $$$2$$$ months. the top $$$100$$$ of div $$$1$$$ s are probably like over a fifth cheaters after rollbacks. it's pretty unfortunate, but I think that the situation could be a lot better if more were being done about it. like if you click on one of their accounts, it's pretty much obvious if they are cheaters. their accounts usually all follow a similar pattern: $$$500-5000^{th}$$$ place in div $$$2-3$$$ for the first few contests, then $$$ \lt 50^{th}$$$ place in div $$$1$$$ s. and there are hundreds of accounts out there like that (that are $$$\ge$$$ $$$CM$$$). if someone can easily tell that they are cheaters, then they should be banned, but they aren't. so clearly the situation could be a lot better than it is now.
Looks like the only way to defeat ChatGPT now is to sneak in "is there a seahorse emoji?" in the problem statement.
there was actually a seahorse emoji, but we just switched timelines which really threw off the $$$AI$$$
copium level set to max
Some serious actions have to be taken to protect the sanctity of CP
Till what rating is AI able to solve consistently (say, more than 75% of the time)? If it is ≤1900, then “just get to 2100 and you won’t be affected by cheaters” isn’t a cope.
how about https://mirror.codeforces.com/profile/phanvien
Waiting for cheaters to destroy div1 just like they have already done in div2 and div3