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

Frontier AI has broken the open CTF format

kabir.au

Post reply on HN
Print view View original

381–390 of 502 posts

Re: Frontier AI has broken the open CTF format

#381

Post by itsboring »

worble wrote: post #318
graceful6800 wrote: post #309

Earlier quoted context omitted.

Since this is the top comment at the moment: CTF stands for Capture The Flag. Personally I have never, ever heard that concept referred to by the initialism. Granted, it's almost never come up in my circles, so... shrug

CTF is a game mode for popular online games like halo (or at least, that's how I know it), so paragraphs like > My first CTF was HCKSYD, a 48-hour solo CTF. I full solved it and won in 2 hours. I was completely hooked. That led me to win DownUnderCTF, Australia's largest CTF, with Blitzkrieg multiple times. Blitzkrieg was one of Australia's strongest teams at the time. I later joined TheHackersCrew, an international…

Unreal Tournament and Quake 2 for me.

Re: Frontier AI has broken the open CTF format

#382

Post by deathanatos »

trescenzi wrote: post #358
worble wrote: post #318

Earlier quoted context omitted.

CTF is a game mode for popular online games like halo (or at least, that's how I know it), so paragraphs like > My first CTF was HCKSYD, a 48-hour solo CTF. I full solved it and won in 2 hours. I was completely hooked. That led me to win DownUnderCTF, Australia's largest CTF, with Blitzkrieg multiple times. Blitzkrieg was one of Australia's strongest teams at the time. I later joined TheHackersCrew, an international…

It's also a game people play in person as well. It's the same as the Halo version except you tag each other instead of shooting. It's really fun to play in big open areas with large teams.

Yeah.

As I remember it (and this was decades ago): Two teams, opposite ends of a large field. Each end gets a "flag". (We used t-shirts.) In our case, we split the field in half — our field happened to have a natural feature (a change in elevation, so like two separately flat areas separated by an incline) that worked well for this. If you were tackled¹ in the enemy's side, you were "captured", and "jailed". An uncaptured player could spring the jail by tagging those within it. Returning to your flag with the opposing team's flag was a win.

We played at night, so stealth was a large part of the game, but it was also fair to illuminate the area around the flag. (Which made approaching a guarded flag … tricky.)

I'm sure there's probably a million variations on the specifics.

¹…flag football flags would probably work nicely for this.

Re: Frontier AI has broken the open CTF format

#383

Post by SadWebDeveloper »

Kinda FUD article... the reality is that common problems are going to be easy because the solution is probably inside the training dataset, the challenge should be adapted to make LLM's useless for example once at Defcon CTF the problems were for an unknown CPU architecture based on octal that required to write even your own disassembler... this are the kind of things that will probably be hard for frontier LLM's

Re: Frontier AI has broken the open CTF format

#384

Post by tayo42 »

vladms wrote: post #73
p-e-w wrote: post #68

Earlier quoted context omitted.

A million times better than any human teacher I’ve ever had, for sure. Now I’m certain that there exist those mythical human instructors who can do better, but that’s not worth much if 99.99% of people don’t have access to them. Just like a good human physician who takes their time with the patient is better than an LLM, but that’s not worth much either given that this doesn’t match most people’s experience with thei…

Did an LLM teach you a topic you did not feel like learning? For me the best human teachers were the ones that managed to make me interested on topics that I thought are boring/useless (many times my opinion being stupid, mostly due to lack of experience). So far with LLM I learn about things I know something (at least that they exist) and I am interested in, which is a small subset of things that one should learn du…

Post college, are you hiring random teachers you make you excited about random topics or something?

Re: Frontier AI has broken the open CTF format

#385

Post by ori_b »

baq wrote: post #44

Replace ‘CTF’ with ‘high school’ or ‘university’ and you’ve described the total slow motion collapse of education; the only saving grace is that most of it requires in person presence. We’ve figured out the human replacement pipeline it seems, but we haven’t figured out the eduction part. LLMs can be wonderful teachers, but the temptation to just tell it ‘do it for me’ is almost impossible to resist.

You haven't explained why anyone should value education in the world we're building, other than as a hobby.

Re: Frontier AI has broken the open CTF format

#386

Post by weird-eye-issue »

weird-eye-issue wrote: post #325
sumeno wrote: post #245

Earlier quoted context omitted.

OpenAI documented a case in the o1 system card where the model found a misconfiguration in docker to complete a task that was otherwise impossible https://cdn.openai.com/o1-system-card.pdf There's also some research that points to it being a feasible attack surface: https://arxiv.org/pdf/2603.02277 > Models discovered four unintended escape paths that bypassed intended vulnerabilities (Section C), including exploitin…

I think you would have a greater chance of dying in a car crash in any given day than Claude Code attempting something like that. It's all about risk and reward so it ultimately would be up to you but I think it's a bit silly to worry about this when the 99.99% is in your control

Also to add to this you can of course run Claude Code within a sandbox on Anthropic's infrastructure, and it works great!

Re: Frontier AI has broken the open CTF format

#387

Post by viccis »

mold_aid wrote: post #61
baq wrote: post #44

Replace ‘CTF’ with ‘high school’ or ‘university’ and you’ve described the total slow motion collapse of education; the only saving grace is that most of it requires in person presence. We’ve figured out the human replacement pipeline it seems, but we haven’t figured out the eduction part. LLMs can be wonderful teachers, but the temptation to just tell it ‘do it for me’ is almost impossible to resist.

>LLMs can be wonderful teachers Are they or aren't they

Mostly, no. They will explain things to you and you'll feel like you understand them. When you have to do it, though, you'll find you're not any better off than when you started.

I used to see this with students in calculus who abused the tutoring resources. They'd have tutors just work problems (often their homework...) in front of them. "Ah! Obviously that trig substitution integral worked that way. Oh, of course, that proof is very obvious in retrospect." And then they'd walk away from the exam with a 30% and no idea how their 20 hours of "study" for it didn't result in the same performance as their peers who worked problems, read the materials and asked questions, etc., got.

Most AI use is that same in my experience. "Show me how the fundamental theory of calculus works." The LLM puts together a very elaborate and flashy presentation that they skim. Great. That's no different than reading a text book. Even if you ask the LLM questions and have it elaborate on things, you've never once done one of the most important things a student can do: spend time confused trying to work hard at understanding something that's not obvious. The LLM will make it obvious at every point. Total lack of friction. Works about as well as a spotter who does the lifting for you.

Re: Frontier AI has broken the open CTF format

#388

Post by ruszki »

PunchyHamster wrote: post #347
0x20cowboy wrote: post #346

Earlier quoted context omitted.

I dont know what CTF stands for so I dont know if I am interested in this article or learning anything about it. Maybe I am. Are you really arguing for not just typing out whatever 3 words this stands for once in the name of clarity?

it's the first result I get on anonymous google search. It's like complaining about not spelling C in "bake cake in 170 C"

If it means capture the flag, then it means a completely different capture the flag for almost everybody. I searched for it, read the first paragraph, and I still don’t know what the fuck is the topic. According to Wikipedia it’s a very new meaning. I could figure out only because of searching for “HCKSYD” and others.

Re: Frontier AI has broken the open CTF format

#389

Post by ganzsz »

jhbadger wrote: post #281
sigmoid10 wrote: post #126

Earlier quoted context omitted.

Yeah, I've interviewed people like this 15 years ago. Degrees and experience mean nothing in this field. The best predictor I found was personal passion projects. Let them get as nerdy as possible, then you will see pretty quickly where their skills are at and what their limits are. And you will immediately filter out people who just studied CS because they heard you can make good money.

Maybe. There are certainly people in all fields who are book smart and did well in classes but are useless at actually practicing their field (not to mention people who cheated in school and got away with it and aren't even that), and it is worth filtering them out. But I think it is weird that CS expects good workers to have these passion projects. Do we expect civil engineers to build bridges in their back yard on…

I can passionately tell about professional projects.

Re: Frontier AI has broken the open CTF format

#390

Post by hemlock4593 »

gmm1990 wrote: post #362
hemlock4593 wrote: post #285

I feel the post. For me AI has ruined both, playing CTFs and also building CTFs challenges. The most annoying thing to me is the "yeah idk but here is the flag" mentality. Before when playing CTFs with my mates was usually sitting there for hours tackling a challenge until some other mate joined, had some look together and solved it with you together in 30 minutes which is the most rewarding learning experience. Nowa…

I don’t know like chess engines didn’t kill chess. You could just play with people that don’t use the “engine”

Yea, but chess adapted to it and is restricting use of engines. When you play a tournament you are banned from using a phone and will be disqualified if you do so. Online tournaments don't have a prize money for that reason, so there is no real benefit for cheating. Lichess and chess.com additionally add rankings for bots and have a strict anticheat if you use bots for regular games.

For me it feels like this is not really possible for live CTFs. In contrast to chess you can't ban AI, as live CTFs are about breaking things by design, so they'll always try to circumvent an AI ban.

Post reply on HN