Blog Details

AI with machine learning banned 20,000 “toxic” cs: go in a month and a half

Faceit is a platform for competitions, including Counter-Strike: Global Offensive. In order to strike at caustic behavior among the players, the owners of the site enlisted the support of the cloud service of Google Cloud and the technological incubator Google Jigsaw. As a result, Minerva was born.

Minerva version 0.1 is already working on Faceit. So far, it is focused on text chat: if II detects toxicity, the culprit receives punishment, and abusive messages are marked as spam.

The punishment depends on the number https://slotsnogamstop.co.uk/ of violations: the more violations, the more serious the punishment. First, you can get rid of the warning and temporary detachment from the matches, but malicious toxics grab the ban. Minerva is able to make a decision within a few seconds after the end of the match.

As Faceit writes, AI-Admin trained for several months to minimize false works. The full -fledged launch of Minerva took place in August: the developers stopped manually interfering in its decisions.

The results were as follows:

• At the stage of multi -month training, over 200 million chat messages were analyzed. Of which 7 million were recognized as toxic.
• For the first one and a half months of work, Minerva wrote out 90,000 warnings and 20,000 banks for insults and spam.
• A month after the launch of miner, the number of toxic messages decreased by 20.13 %.

CS: GO Conditionally paid, they also do not require money to register on Faceit, therefore, SMS data is confirmed to combat Smurfs on the site. In order to take part in official matches at Faceit, to users whose account is suspected of smoothing, boosting, potential cheating or toxicity, first you will have to bind your phone number.

Faceit owners are satisfied with the results of Minerva and are going to develop it further – including adding AI to “Other Systems” (apparently, this is meant other games).

コメントを残す

メールアドレスが公開されることはありません。 が付いている欄は必須項目です