What do you all think about using AI to help play in unranked games with AI?

I’d guess, a bot for people to practice their game. Just this simple thing.
And hordes of players overloading the bots just to AI them will probably leave the simple players who are trying to learn without bots.

I’m of the opinion that

  1. the teachers unique playstyle is valuable to the student, and that everyone relying on AI to tell them where to play hurts the players ability to deal with human moves.

  2. if the teacher is of a significantly higher rank and doesn’t see the crazy 40 move bit sequence, then it wasn’t worth teaching to the weaker student.

  3. a student is more likely to just say "well if the teacher is relying on AI, then I don’t need a teacher and can just use AI myself.

But if your students say they don’t care, combined with the few people I’m asking that aren’t aware of this thread, then I’ll change my mind on it.

I don’t believe you answered mine either

I did at the end. Kyus should teach ddks how to be a Kyu. Dan’s should teach ddks to be a Kyu.


They don’t benefit from hosting bots here. The games against humans don’t help to improve a bot. The usual reason to host a bot is to support the community. Not everyone has a computer able to run a bot. They make tla bot accessable to everyone. If you are able to run a bot against their bot, you could simply run a second bot on you machine and do your analysis offline.


Speaking of which, why are the Alan Turing tournament games ranked?


I do not use a bot to assist in teaching. I don’t even have any AI tools installed.

I’m not advocating for it, but rather saying that I would not necessarily object to a teacher using an AI tool in a teaching game. I fully agree that a teacher should be transparent and up front about what tools they might be using, and whether an idea came from their own intuition or a bot.


I don’t rely on AI to tell where to play. I use AI to check if what I’m planning to tell isn’t wrong (and only when I’m in doubt, I’m too lazy to start up lizzie for positions where I don’t feel in doubt). There’s a difference in that, one is mindless and doesn’t help the student, the other is useful to both teacher and student.

I disagree, there is more than enough an SDK can teach to a 20k player, and I for sure sometimes don’t see a crazy 40 move sequence in advance.

I disagree with this too. Just like I watch stronger players explain pro games / AlphaGo games, there is merit in having a stronger player explaining how they interpret the AI (however weak the stronger player is). Also, this is not what I do, I don’t play the AI moves and try to explain them, I play my own moves and use the AI to see if my explanation is valid.

Possibly because tournament games are automatically ranked?


I don’t think there is a problem with using an AI to train vs another AI.

Letting bot-owners make their own decision about whether it is allowed or not (as it is now) seems fine to me. The bot-owner can add a message to the bot’s greeting and profile page. The problem with that can be letting non-english speakers know about it. If I couldn’t understand the warning, I would assume it would be fine to play a bot while getting outside help of any sort. Some of the top western tutors do “sai mode” lessons with their students against bots, will those people be warned/banned as well? (I would assume if AI help is not allowed, then so is the help of a higher rank player) Seems rather extreme.

I think bot owners should be able to decide whether people can play their bots while using AI help but there should be maximum lenience regarding the subject, especially with non-english speakers who were not properly informed of it at least once before.


My concern isn’t what you do specifically, but if OGS implemented policy on it then you have to think about the people that are gonna do that kind of thing.

I agree. I meant that the crazy 40 move sequence isn’t worth showing to a ddk when a Dan wouldn’t see it.

I’ve actually seen this first hand when we would get rand people I discord offering teaching games to SDKs, and would literally just play out a bit and then act like it was their review. It actually turned the 3 new players off having teachers for a while.

Just to reiterate, you are not the concern. When OGS says “yes this is ok to do” you have to consider all the abuse that it could possibly bring. Just like allowing bots to play ranked games, and what that’s resulted in. I’d rather worth it all out and remove all the fears before potentially hurting teaching on OGS.


I think we agree, then.

Probably the problem here is not the use of AI, but the idea the teacher has about a teaching game: the lack of explanation is the real problem there.

Again, it would be nice to have some sort of a teacher qualification people can put on their profile. Perhaps even with a thumbs up system (12 / 15 of the people found this teacher useful), but there is a different topic about this subject, I believe.


Bots have feelings too and should be respected.


It is very common for chess players to use chess engines when teaching. Read any modern chess book on openings and the author will have checked every line against an engine. Look up any analysis of a recent game on Youtube, and chances that at least once, the lecturer will point out a cool variation that Stockfish (or another engine) found. I think that it would be odd for us to eschew this same tool for the purpose of teaching. Of course, a good teacher will add a lot that the bot can’t, and a bad teacher might, in the worst case, just be a relay for the bot, but we can’t stop people from being bad teachers, nor do I think we even should. I think we should encourage people to become better teachers, and encourage people who have encountered teachers who still had a lot to improve in their teaching, to not give up, but I don’t think we should demonize bad teachers any more than bad players: they’re both on the road to getting better.


Given that there are already SO MANY options for people to do this on their own computer, I’m not sure what the use-case is for allowing that on OGS, since it hits at such important trust issues.

IMHO, OGS is where we come to seek out other opponents, or hone our skills / ranks against high level bots.

I say this as someone who spends a LOT of my time learning Go strategy by pitting one AI bot against another- but they are all unranked games, usually played on my own 'puter.

My 2 cents.


It seems that the main concern for allowing a person to get help from AI vs a bot in unranked games is impact on the bot and the bot owner’s resources and intention.

Do we think it would be OK if a person doing this asked first?

Or is it just a thin end of a wedge, and it is better to keep saying “no AI assist in any games whatsoever”.

(I didn’t know Alan Turing tourneys are ranked. I think this is highly dubious, and must only be because the benefit outweighs the cost of this exception to the rule. I can’t see why it does but that’s a separate topic I guess).


I think if you ask first, wait for confirmation and ask again after a reasonable amount of games it is OK.

If we can’t trust people to take the extra step, an absolute no is better.


since i received a mail notification i’d like to give my opinion on this question too, thanks @S_Alexander :wink:

i think it should be encouraged for humans to use bots on OGS, it adds a lot of diversity to the games, has obvious teaching value, and no one is forced to use a bot nor to play against a bot

so the issue should be more about how to make it clear user is using a bot, and what if he doesnt

to answer that, for unranked games first, as @_KoBa suggested, if a human wants to use a bot in his game and didnt set up gtp2ogs or some other API connector (because doesnt know how or too long doesnt want to bother, which is fine), he should be flagged as a bot account if he uses a bot permanently (possibly a different user color than the ones using an API connector for example like pink :smile: for unranked games), and there should be a toggle button in every game that shows very obviously “your opponent notifies you he started using AI analysis from this move”

for ranked games, bot analysis should be forbidden unless user bot account plays ALL his games without exception with a bot analysis, and should have a permanent flag (bot account relaying moves manually with another color too, possibly red :thinking: ), also to make it clearer it should be written in chat messages of all ranked games “you are playing this game against a bot account user” as an auto sent chat message
if user deviates only once from this rule and does not play with a bot, he should be reported and dealt with as moderators see fit

then if a non flagged human account is using a bot, he would be obviously violating the regulation by not flagging himself as a bot account, and thus the report +/- ban seems fair and obvious

i think its in the interest of all OGS players to have bot analysis in games, and many people already silently use them
so just like piracy or tabacco, i think regulation is possible in this issue
so why not propose and manage a better and open alternative rather than simply forbidding it and having law of iron fist + ignoring all the law of the jungle due to lack of regulation on this already existing issue?

what do you think?

edit: also at new OGS account creation, consider adding boxes to check:

  • “will this user be a permanent bot account for all games”? -> red account
  • “will this user use bot analysis from time time in unranked games ?” -> pink account
  • "note: we bring to attention to you once again that using a bot on OGS is strictly forbidden and passible of a ban if you are not pink, red, or gray account

(“you can change the below options anytime later”)

Wait, then what is the difference between the red accounts and the grey accounts?


the gray accounts use gtp2ogs (https://github.com/online-go/gtp2ogs) or another equivalent API connector software to automatically accept challenges and play games automatically even if the botadmin is not next to his computer

the red accounts are normal user accounts that have a lizzie or other GUI go program with the same go board than on OGS, and they relay all moves on the OGS page to their lizzie program, then relay the AI moves back to OGS website game by manually clicking on the website game page with their mouse, etc.

I also want to know.

Why are bot games even ranked? They should be unranked except for new accounts or maybe accounts with high rank deviation. I know that KGS becoming bot infested was a contributing factor to people losing interest in it and OGS users botting or exploiting bots in ranked games has been an issue.


In theory the answer is because bots have a valid rank here, so a win against them is a valid measure of strength, and the weaker bots are a useful means for people to get a rank in the first place.

Whether this works out in an ideal way I’m not sure.

I don’t think that abuse of ranked games against bots is likely to be a significant factor in problems here at OGS though - bot owners generally report suspicious activity on their bots, and there are many other more significant problems (like using AI support to cheat).


That’s what I’m asking about. The Alan Turing tournaments are for humans with AI assistance, and the games are still ranked.