As I’m sure you all know, use of AI to help in our games is categorically ruled out in the OGS Terms and Conditions.
It is not allowed.
However, the question has come up a few times “could I be allowed to use AI to help me experiment with play unranked vs Bots”?
Personally, other than the “thin end of the wedge”, I can’t see a real reason why not.
What do you all think?
I feel ogs should hand over grey names way more than in the current situation. Make everyone know that it is perfectly OK to create a new account for botting and request grey name for it.
EDIT: I mean, this server already has very good system for running a bot-account, sadly most ppl dont just know about it, nor that grey names are given on by request-basis.
OGS should advertise this feature more, let everyone know how easy and simple it is to run a bot here (and how easy it is to play (or avoid playing) with bots since they have grey names), and we could prolly get lot more users who are interested on bot-related stuff
I think it is a great idea too , but as a primary student almost middle/secondary don’t know how you like to call it I won’t really be able to make a AI.But can’t you just log in to that AI account you created and challenge another bot then spectate in a unranked game?
Just to clarify - this isn’t about “setting up a bot”.
This is about using an AI that you have - probably Lizzie - to help you play the game against the AI on OGS under any circumstances (other than a tourney agreed for this purpose with unranked games).
You totally are not allowed to do that at the moment.
The question is whether you should be allowed to experiment with this in an unranked game with an AI here.
Why would it not be allowed in unranked play? Is there any good argument? People who want serious games can play ranked.
I think I have also violated this rule, if it’s not allowed under any circumstances. During teaching games it can be quite useful to do a small check with AI halfway to see if whatever you’re explaining is correct, I’ve done this a few times, just like I would do during a review.
I think it would be useful, actually. I don’t see anything wrong in doing so.
TBH, I think that the Terms of Service weren’t really thinking about unranked games.
That being said, they are pretty unambiguous
I too have done an AI check while teaching, and so clearly I consider that to be an obvious exception already
I do think that players who play unranked vs humans outside of teaching don’t expect to be facing someone using AI assist.
As an aside, players who play unranked are often a bit fuzzy about their expectations: some of them expect that their opponent’s rank will be representative … to the extend of reporting sandbagging … even though clearly their opponent plays unranked at least some of the time, so their rank is unlikely to be representative …
… But it does go to show that players use “unranked” for games that they expect to be somewhat “realistic”.
But as I said, I too can’t see why it would not be allowed unranked vs an AI - who doesn’t care about that sort of thing
It’s a bit similar to something like Dwyrin playing unranked sandbagging games with nobody complaining, because it results in educative videos.
I think it should be allowed, but they should first request that their names will be made grey before they do this. Then they are free to use their lizzie or smth on ranked and unranked games against other bots or humans.
I suppose this is actually a solution, come to think of it.
It’s a very “heavy weight” solution though!
I think using an AI tool in any normal (non-teaching) game against a human should remain strictly forbidden, unless explicitly agreed beforehand under special circumstances (and in an unranked game, of course), since players may be specifically seeking out to play against an actual human, which have a different playing style than bots.
In terms of using an AI tool for the sake of exploration in an unranked game against a bot, I think this may likely be less of a concern. However, bot accounts are often run by volunteers and do consume their personal computing resources to provide a public service. Perhaps some of those bot operators would want their resources prioritized toward offering games against actual humans, rather than having their bots be used for analysis in conjunction with another bot. A bot operator might want to limit the number of games that their bot accepts from other bots in order to set aside enough capacity for accepting human challenges.
If someone wants to use an AI tool against a bot account, I think it may be beneficial for both accounts to be marked as bots, since then the other bot could use that information to decide whether or not to decline or accept the challenge.
Maybe even an additional type of account should be classified? Perhaps, a “centaur” or “cyborg” designation for a human that is also using an AI tool for assistance. Of course, I think such accounts should not play ranked games, and be careful that their opponents are aware of their nature before offering and accepting challenges.
It’s perfectly fine, and I hope nobody has actually enforced the rules as written in a vs-bot situation.
If the bot operator wants to benefit from volunteers spending their time to test and provide data for their software, they will have to accept that the public will interact with it in all kinds of unpredictable ways.
Do your students know you use a bot to assist in teaching?
If not, would you go back and tell them you used one to see how they felt?
I’m strictly against my opponent using AI, ranked game or not*. Go players seem so old-fashioned to the minutest details, but I guess “gentlemen’s agreement” flies out the window at the drop of a hat. When we offer or accept a challenge with a human, we expect to play a human. If I want to play AI, I’ll choose a bot. It’s baseline respect.
As far as the guidelines are concerned: In general, I’m wary of people searching for the loophole on principle. Some people seem to drive straight to “how can I game the system”, and I don’t like those people.
*Yes, I know I’m not able to find out or even tell if my opponent uses AI.
The first thing I think about the bots is “yay, someone spent time and resources to provide the community with a bot”, not “eh, they get data, I can exploit it”.
Eh, I can understand if someone doesn’t like it. Dwyrin is nice and all, but other players don’t exist for his videos.
EDIT: And about the original question: for me, the right thing to do would be for the player to write in the name of the game “AI used against bot”, if the game is part of their record anywhere. If someone has 50 wins against kata-bot and they don’t disclose they used AI, I’m pretty sure their intention wasn’t to explore the endless game possibilities.
Would you rather have a kyu player* teach things that are wrong, because they didn’t check their ideas? If I teach something else and I’m not sure about it, I also look up the answer before opening my mouth and spreading misinformation.
*In fact even for dan players and professionals it should apply
@Gia I get your point about botting in human games, but that’s really besides the point here - we’re talking about bot games.
And why shouldn’t I look at machine analysis on my own side? To call this “exploitation” of the operator’s resources is a real stretch.
If “no running your own bot” is your position as operator, what are you really offering? If you’re just running a bot for the enjoyment of others, why put such inconsequential restrictions on that enjoyment?
You didn’t actually answer the question… But uh…
Can a Kyu player interpret an above-pro bot move correctly, and then have the ability to communicate to a ddk why that move is good?
Often you have to actually read the 40 move deep sequence (which includes multiple global positioning judgements and counting and influence estimates) to find out why.
A Kyu player should be teaching a ddk how to play like a Kyu, not like a bot that can beat all the pros 100-0.
I guess you look up the answer to the best of your abilities.
I don’t mind if a teacher would use AI, but I would wonder which part of the teaching is them going forward.
I’m sorry, I thought they were rhetorical. But sure, I can tell them. I even do occasionally say that the AI considers the position favourable to this and that player by so many points (if it’s insightful).
I don’t believe you answered mine either.
Of course not, even Michael Redmond struggles with explaining above-pro bot moves. That doesn’t mean that even weaker players cannot make use of bots to improve their teaching. If I’m telling a player that they should throw-in in some life/death situation, but because of my kyu-ness I can’t see that the throw-in doesn’t work because there is a very complicated 40 move fight that ends badly, then I can change my “in this situation it’s good to throw-in” to “in situations like this I would throw-in, but I’ve run a test with AI, and it seems that the situation is actually pretty complicated, so let’s not throw-in and try to find a more peaceful way to play”.