Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 7:04 am
by dust
Basing a decision on fans' and spectators' reactions would unfairly turn it into a popularity contest.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 7:41 am
by Javaness2
I wonder they cheated so blantantly. Is it unreasonable of me to suspect that they simply wanted to be caught and punished?
Personally, I would find a catch the cheater event to be much more interesting. Have 6 players play a round robin tournament. In each round, one is assigned the cheater-role. The aim is to find out which player cheated. After each round the audience votes for who they think the cheater is. Those who vote correctly are allowed to live. Those who vote incorrectly are sent to the Vortex of the Rangdo.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 8:07 am
by Mike Novack
Just to bring another game into it, consider the Reese-Sshapiro matter in bridge (1965). At the time I was playing a littlew tournament bridge, had a few "red points", and met a few of the world's top players across the table (will always remember the thrill of getting a good result against Kay-Kaplan).
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 9:11 am
by Bill Spight
Javaness2 wrote:I wonder they cheated so blantantly. Is it unreasonable of me to suspect that they simply wanted to be caught and punished?
They were kids.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 9:18 am
by Bill Spight
Mike Novack wrote:Just to bring another game into it, consider the Reese-Sshapiro matter in bridge (1965). At the time I was playing a littlew tournament bridge, had a few "red points", and met a few of the world's top players across the table (will always remember the thrill of getting a good result against Kay-Kaplan).
At my first US Nationals I played against Kay-Kaplan.
I was playing with a pick-up partner, a high school kid. Kay and Kaplan were our first opponents. We had agreed to play the weak NT. On the first hand I opened 1NT and he raised to 3NT. When the dummy showed, he had only 10 HCP.
I went into the tank, trying to figure out how to go down only 1. Kaplan picked up my convention card and noted that we were playing the weak NT. He put the card back down and said, "Norman, there must be rumors going around about our defense."
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 9:47 am
by Ferran
Just to clear my head:
Is it just me or have we gone from "dismissal from the organization is way too harsh for a teen" to "well, if you really think 1 year is too much..." somewhere? And I don't mean in these last examples, but much earlier.
Thanks. Take care.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 9:52 am
by tapir
I am kind of confused that no one makes the case for her innocence. Afaik korean professionals questioned her on the back of "AI move" percentages below what many on this forum happily defended in mid-dan amateur players.
Of course with the culprit being a teenager and less of "good faith advocates" around Koreans actually got a confession.
Punish often, punish lightly.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 11:38 am
by RobertJasiek
tapir wrote:I am kind of confused that no one makes the case for her innocence.
Initially, I did question her guilt. For going further and advocating her innocence, I lack information. Meanwhile, we have firm statements of her admitting AI usage. Since she is still a child / teenager, it is a principle possibility that she might have been pressed to admit that while not exactly understanding exactly what she admitted or being convinced to adopt a guilt role voluntarily as the allegedly best still available option to her. Others might have applied so much pressure that she might have confused AI usage for study with AI usage during the particular game etc. However, with the currently available information to us English readers, such are merely theoretical possibilities. It bounces back to the reported firm statements of her admitting AI usage.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 11:53 am
by Ferran
tapir wrote:I am kind of confused that no one makes the case for her innocence. Afaik korean professionals questioned her on the back of "AI move" percentages below what many on this forum happily defended in mid-dan amateur players.
I can't speak for others, but, myself, I'm not commenting on guilt / innocence because I simply don't know. And then I'd have to focus on "she started denying then, under pressure...". If I start thinking on Eastern peer pressure and teenagers I'll go ballistic.
On this subject, there's an English commentary by Yeonwoo 2p on this. TLDR, she leans towards a 3-y suspension.
Take care.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 12:40 pm
by explo
I missed the start of the debate, but so far I've agreed with all of jlt's posts. Maybe it's because we're both teachers. If every student I've caught cheating got a definite ban from school it would cause a mess.
One thing that I haven't seen mentionned and that I would be worried about is that setting the first case to a definite ban would set the bar quite high for later, unclear cases. What do you do when you think someone cheated but you're not really sure? Do you ban them for 10 years, just in case? Pros studying hard with IA will not make every case easy.
Kirby wrote:And as a pro go player, if anything is worth being fired for, cheating in a for-cash tournament seems a good candidate. It amounts to stealing money from the sponsors.
I can see how it amounts to stealing money from your opponent but I fail to see how it steals money from the sponsors. The sponsors pay the same amount of money regardless of cheaters. I also think that at 13 year old people cared more about winning than about the prize money.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 2:42 pm
by SoDesuNe
John Fairbairn wrote:
But being fired by the only organisation who can actually employ you is quite a bit harsher than your average job loss.
But Kirby's example was "Insider trading may be an example of a "fire-able" offense in a company outside of the go world. " That's not an average job loss.
How about this for a compromise? Instead of disqualification for one year, you get disqualified totally, but have the right to reapply for membership after one year. The re-application is then dependent on what the rest of the membership feels, which will reflect one year's more information about the remorse shown by the cheater, one year's more worth of fans' and sponsors' reactions, and one year's more information about the scale of the cheating problem.
Actually I did google before and did so again right now: As far as I can tell insider trading does not mean you can never work in the financial sector ever again. Sure, there are consequences and I'm all for consequences. They just should be thought through and the first thing to think about should be prevention of further offenses (on an individual level and on a "collective" level).
That's also one of the reason I wouldn't trust masses to judge on a person's sentence (and certainly not when it comes to potential competition). For one thing putting the masses in a sitation to pass DOWN judgement and therefor elevating their position is not really to my cynical self's liking. For another reason, as per your choice of words "what the rest of the membership feels", "remorse shown by the cheater" and "fans' and sponsors' reaction": Drama does not really benefit a legal process as emotions are a really fickle thing. A "judge's" change of heart cannnot change a legal sentence.
As for the innocence of the player: I'm actually more interested in how cheating in go will be perceived and handled. Nevertheless I think the tempory ban was the right move from the pro association.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 3:17 pm
by Uberdude
A problem I see with removing her pro qualification and then allowing her to retake the pro exam later is that (assuming she wasn't always cheating and is a strong new pro) she is very likely to pass the test (even if banned from KBA she can still play and practice and improve online) so reduces the limited number of spaces for the rest of the cohort. So do you increase the number of new pros allowed that year? Or every 3 years there is one extra place for reapplying banned players and they can't play in the normal cycle?
As for her guilt/innocence and the quality of the evidence against her, there was some discussion of that on reddit for several days before I posted here after the new information that she apparently admitted cheating. I wasn't much impressed with the touted evidence of high matching against AI: if the KBA can't manage to organise a mouse which doesn't trigger a touchpad how can they be trusted with the much harder task of rigorous statistical analysis?
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 3:26 pm
by tapir
Ferran wrote:
I can't speak for others, but, myself, I'm not commenting on guilt / innocence because I simply don't know.
Well, there is a confession. And the confession exists only because Korean professionals had none of the concerns vented in this community when faced with a cheater in the European Team Go Championship. They acted on far less information as far as I understand.
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 4:15 pm
by Kirby
Ferran wrote:
On this subject, there's an English commentary by Yeonwoo 2p on this. TLDR, she leans towards a 3-y suspension.
Since you're summarizing this 17-minute video in a TLDR, I feel compelled to add context. Namely, I watched the Korean version of her announcement when it first came out:
At 15:55 in the video, she very clearly says that Kim Eunji should be unconditionally expelled (무조건 제명, which is *unconditionally* expelled/fired/dropped out). *Otherwise* (ie. if that's not working), a 3-5 year ban at minimum. In the English translation, she just mentioned that she advocates at least a 3 year ban. But given the context of the original Korean video, I believe that this really is a minimum - she mentioned kicking her out, first. Maybe she had a change of heart or played with the wording in the translation - but I think it's inaccurate to say that a 3 year suspension is exactly the punishment she'd opt for.
explo wrote:
Kirby wrote:And as a pro go player, if anything is worth being fired for, cheating in a for-cash tournament seems a good candidate. It amounts to stealing money from the sponsors.
I can see how it amounts to stealing money from your opponent but I fail to see how it steals money from the sponsors. The sponsors pay the same amount of money regardless of cheaters.
The sponsors pay prize money to the winner of the tournament for *playing go* - not for copying moves from an open source AI. Tournaments have rules, and you get prize money in accordance to those rules. If some non-go player barges into the next Samsung cup and steals the prize money without actually playing any games, the sponsor still dishes out the same amount of money... But it's still stealing, is it not?
explo wrote:
I also think that at 13 year old people cared more about winning than about the prize money.
Clearly not by her own ability...
Re: Young Korean pro Kim Eunji banned for 1 year for AI chea
Posted: Fri Nov 27, 2020 4:20 pm
by Kirby
Uberdude wrote:I wasn't much impressed with the touted evidence of high matching against AI: if the KBA can't manage to organise a mouse which doesn't trigger a touchpad how can they be trusted with the much harder task of rigorous statistical analysis?
I agree with this. Generally speaking, I don't think we are very good at detecting AI with enough accuracy right now. But that's another reason I personally believe that the punishment should be steep when we *are* certain of the cheating. From a game theoretical perspective, if someone has no morals, the value in cheating is: ChanceOfNotGettingCaught*ValueOfUsingAI + RiskOfGettingCaught*NegativeValueOfPunishmentForUsingAI.
Basically, if there's low punishment for using AI and you don't have other moral standards, there's greater incentive to cheat. Since we aren't good at controlling the risk of getting caught, yet, this is the only variable we can do something concretely about.