@diPero @Kusokosla @Onyx_Chess @petri999 @Sarg0n @pawnedge @clousems I think it's limited to focus on AI cheat detection using only board moves for data.
Technology is rapidly improving, getting cheaper and getting smaller. AI can also be applied in other ways if we provide more data inputs.
Surveillance cameras are now extremely cheap (especially as a bulk purchase). Imagine if in order to play rated games on lichess, you must install a 1080hd camera behind you that has clear view of both your screen, your hands and your body. This camera would be activated and record (saving the last 16h to the cloud) when you log on to lichess. It would be accessible only to AI data analysis and lichess moderators. Moderators can read your typed private messages at any time anyway, so there is minimal difference in introducing cameras.
AI could then iterate pattern recognition to learn typical cheater body language. For instance, perhaps 20 stockfish moves in a row are typically correlated with a hunched, restless and angry posture or body language. Then a higher % of cheat analysis resources are automatically allocated to players with this posture. Efficiency goes up significantly.
Early on in the implementation of cameras, when a player is reported, the video footage could also be reviewed, with screen content and head/eyes direction analyzed by a moderator to see what the player was doing. Eventually, more and more of the data will be handled by AI. But at the start, moderator actions would also be useful data points for AI to learn. Just like crashes are (unfortunately) the most insightful data points for Tesla autopilot to learn from.
The only counterpoint is the question of cost. Users could buy them or lichess could buy them. Even the most difficult case of Lichess purchase is easy to justify from an investment standpoint. Lichess generates revenue through donations. Let's say bulk purchased cheater monitoring cameras cost $10 each. Let's say they are a treated as a 3 year investment. Therefore, if the introduction of a $10 camera per user leads to the generation of at least an extra $10 in donations per user across 3 years then this investment is worthwhile for lichess.
I predict that this extra revenue is likely, since lichess would almost completely eradicate both existing and potential cheaters. You cannot deny it if you are caught on camera doing it. There is also the case where lichess automatically publishes the video footage in order to shame cheaters and deincentivize cheating even more, but that is for another discussion. Most importantly, I'm sure most people here would donate $10 to play in a cheat free environment for 3 years. We could also put constraints on the initial investment, such as only sending cameras to users who have previously been marked for cheating or artificially changing their ratings. Since offending predicts reoffending, this is the highest donation yield investment for lichess.
There is also the option that lichess requests that you order a camera yourself (from their store) and set it up in order to access rated games. This would also work, given that most players here are addicted and only play on this site. $10 is a small barrier to entry.
Since camera footage would collect more data points and iterate AI pattern recognition much more quickly, moderator and dev workload would also decrease quickly. This would mean that they can spend more time improving lichess rather than undertaking arduous policing work that should be automated. It benefits everyone and is another incentive for donations. Join me in petitioning for surveillance cameras with AI analysis to be introduced.
🙏 Warm regards, Burrower 🙏