I think finding AI wouldnt be a problem. Not finding it seems to be a bigger problem…
It’s nice that there’s progress on this, but it’s not yet at the level where I’m thinking in my head how to better catch cheating. Give you the idea that AI catches itself during a game still exists in live and says it will serve the managers without the culture of a human. This is an idea that you don’t know how to do. So I abandoned this idea.
I thought that too. I even thought that a prominent streamer might say “Hey, OGS is taking action, that’s cool”. But no 
You would think so, but don’t under-estimate the price paid in complaints 
Nonetheless…
… we agree, and we’re working hard on that.
2 Likes
This makes me wonder what is the “false positive” rate or “abuse rate” of the report button (like someone reports any game they lost, etc.). And the workload for the report responses
If it is low and acceptable, maybe the announcement of the report function would be helpful (we see some players actually didn’t know this function existed), instead of announcing the account banning.
Just for clarity, we have never “announced” an account suspension (which is a personal thing).
We notified the other player, in their game chat, of the annulment of that game for the reason.
I think the people want to know that there is action, rather than wanting to know that they can complain 
Except that when there is action, they don’t want to know about it 
The false-positive rate is uncomfortably high, but also inevitably I think. Since reporters can deduce that it might be (or in their mind must be) AI for all sorts of reasons, we will inevitably get these.
We do receive enough AI reports, and also proactively hunt for it, that we don’t need to solicit more 
(At the very least, every game played here with enough moves to make it valid has a scan for “likely AI use”, which feeds into the detection process).
3 Likes
Well, the question is how accurate are you at finding out that it’s not an AI that you thought up yourself? I know you’re not allowed to reveal this information, but I’m guessing it’s a pretty high success rate.
I remember seeing some posts about being reported as AI here on the forum from time to time, and also posts like this complain about bots everywhere. It seemed to be a big gap between these (at least some of the players) and the team working behind the scenes.
I’m not sure that we have many posts complaining about being reported as AI.
Usually those posts end up coming to me, and I can only remember one this year.
That is one complaint from over a thousand reports handled.
I turned out to be an honest assessment mistake, we reverted, and moved on.
I wonder why we have this perception?
I’m in the team working behind the scenes. My impression is that I respond to complaints and questions whenever I see them.
Of course, as yahel mentioned, it’s a tricky topic because of the inevitable frustration that we won’t talk about what exactly we’re doing to detect AI … for obvious reasons.
2 Likes
Maybe the aim isn’t to have a fool-proof solution, but just a better perception. Maybe we can have a vote of how good or worse OGS is compared to other platforms. I for one will say that I feel OGS has better effort compared to IGS/pandanet, maybe even better or at least on par with foxwq (different rank range might have very different perceptions as well)
2 Likes
Level in OGS is better than Fox. Level in Fox for a man Dan 5 where he is Dan 1 or 2 in Europe and America. Times for playing Fox is better than OGS.
I’m grateful to the mod team for what they’re doing. Personally, I’m too low level to encounter AI cheaters, but I can see it would leave a bitter impression on the opponent.
1 Like
We are talking about the perception of bot detections on different platforms, not the comparison of players’ strength.
I said what they must do is make a robot that checks all games in real time. If it detects suspicion, it stops the game. It informs their manager who the player is that the robot suspects is using AI.
Why a robot, when we can hire a mystical garden gnome to do the same?
2 Likes
That’s not what you replied me, these two are clearly very different responses. And what I said had nothing to do with either of what your replies - surveying the perceptions from players themselves about different platforms compared to OGS.
This is a glitch in the connection. I suggest they develop an AI that can watch in real time and see who is using which player and stop the game and notify the administrators.
It may become quite hard to detect someone using an AI, unless their skills are suddenly drastically better, because there are so many AI programs now trained at various levels. You could (I theorize) even train an AI to play like you, but let it train itself to be just a little bit better, or a lot. The 10 Dan AI will not see a lot of similarity with what it thinks are the best moves, but you might still play whatever amount of stones better, depending on how you configured it.
Perhaps a full AI understanding (trained AI with that persons games as input) of how someone usually plays, ideally including verified games (such as real life I guess), then you can compare the next game against the expectation, and try to judge if the improvement (if any) is impossible or not. This might still be hard to proof in cases where the AI used is only a little stronger (I guess), but it might give a reasonable clue if the way someone plays is suddenly very different or impossibly much better ? If faul play is suspected but the player claims he just improved, he could then later prove that in later games, where he might play worse due to the stress (as Lee Sedol suggested in a cheating case might happen) ? (I hope I understand things correctly.)
1 Like
Can you really blame him for saying that though?
As a go streamer, the amount of AI, and stream sniping the dude has to take on the chin far exceedes that of an ordinary person… and on top of that, its quite litterally part of his dayjob to play go online… Getting botted has a very real affect on him and his work.
In addition to that, ive never seen Dwyrin actively say that OGS has an AI problem in any context that makes it on par or worse than any other Go server…and the dude plays on Tygem, a server where you can litterally pay to bot midgame. He also frequently tells new players that OGS is the place to be because it has “lots of human players!” (An actual quote).
you, and the mods of OGS can have beef with Dwyrin, he can be marmite for a lot of people… but i would also consider crediting him with the benefit of the doubt at the very least… and also recognise that much of what he said online and in stream is very much an overthetop exaguration of his character and how he thinks/ feels… but then its quite presumtuous of me to speak on behalf of him, as it is for you to essentially suggest that he doesnt care that OGS is at least trying to get better at dealing with Abuse of AI in games.
In my opinion thats a broad generalisation to make against anyone.
Its also unfairly, in my opinion, putting too much weight on the opinion of one person. At the end of the day, Dwyrin is just one person… and while Go is their job, weighing the opinion and position of one person against many people who potentially do want to see a more active and open response to dealing with problematic use of AI… is kinda a bad idea.
((I also want it to be known, that im not just posting this out of some misplaced loyaly to Dwyrin. Yes, i enjoy him and his content, but im also going to be the first person to call out any bullshit when i see it. I speak here, out of a concern for using the opinion of one well known person as a potential cudgel against doing something because of some precocieved idea of backlash))
Whoa, I didn’t mean my comments to be “blaming Dwyrin”!
I’m a fan of his!
I used his comment only as an observation of how people in general react.
He was a publicly visible example I could point to that illustrates the general problem of publicizing the policing of AI cheating.
(Actually, there are a few problems with it, this is only one of them.)
Personally what I took away from it is that we need to be as transparent as we can about what we’re doing when asked, but avoid proactive advertising about it. People just want to NOT experience AI cheating. If we can achieve that quietly, so much the better.
I will say that not everyone in the AI detection team agrees with this “low profile” approach, so it will be interesting to see how we go
2 Likes
He’s an influencer playing Go and it doesn’t matter to him if there was cheating or I didn’t understand correctly? (Dwyrin)