Microsoft aims to filter voice conversations in real time with AI
by pxhere
Microsoft has announced that Artificial Intelligence (AI) will apply filtering that automatically limits content based on user settings so that inappropriate content is not displayed in text messages received by Xbox Live users. In the future, it is reported that a function to filter voice chat performed on Xbox Live in real time will be installed.
Putting You in Control of Your Gaming Experience-Xbox Wire
Microsoft unveils Xbox content filters to stop the swears and toxicity-The Verge
https://www.theverge.com/2019/10/14/20913323/microsoft-xbox-live-content-filters-messages-party-chat-moderation
It has become commonplace to play games and collaborate with people from all over the world via the Internet, making it easy to communicate via email and voice chat. However, playing games with strangers can sometimes include inappropriate content, such as bad or discriminatory messages.
In Xbox Live, it is now possible to filter the content contained in text messages by selecting from four stages: Friendly, Medium, Mature, and Unfiltered. Friendly is the most child-friendly level and is designed to exclude all potentially negative messages. On the other hand, Unfiltered is a setting that does not perform any filtering.
Also, if you own an account with each family, parents can use family settings to set and manage child account filtering from Xbox apps on Xbox One, Windows 10, iOS and Android .
by
A reporter of the technical media The Verge said that he was able to pass through the filtering many times by sending dirty words to his Xbox Live account in order to check the accuracy of Microsoft's AI filtering. However, even if we try existing techniques to avoid filtering, Microsoft's AI picks up certain words and messages and excludes them, so it seems that it is possible to achieve fairly high-precision filtering.
However, The Verge says, “Phrases and memes for attacking specific communities are immediately generated on the Internet, so even Microsoft filtering is absolutely uncertain.”
On the other hand, even if it is a negative word, it is not always used in a bad way, so high-precision filtering requires a very complex algorithm. For example, when you propose a strategy with a message saying 'Let's kill the enemy in XX first!' In online multiplayer, the message contains the negative word 'kill', but there is a problem based on the context It is not a conversation.
Microsoft's Xbox operations manager Dave McCarthy told The Verge: “In a game space, the context of a conversation is very difficult to handle. “It ’s a never-ending battle to find a way to understand.”
Microsoft officially stated in the official blog: “In the end, it complements existing efforts and leverages our efforts in AI and machine learning technology to filter all kinds of content on Xbox Live and to individual players. We are aiming to provide control to. ' This means that we are considering filtering Xbox Live voice chats in real time, not just text messages.
by USAir Force
According to The Verge, Microsoft has already developed a speech-to-speech function to transcribe speech in real time, and is considering applying this technology to Xbox Live. However, if you analyze the voice in real time, there will always be a time lag, so it can be said that it is quite difficult to apply it to in-game voice chat, which requires immediateness.
Regarding the filtering to voice chat, Mr. McCarthy said, “Development started from the question that if voice can be raised in real time and filtered, voice conversation can be filtered. “Development of filtering is just getting started, so I ’ll be working on it,” commented.
Rob Smith, Program Manager of the Xbox Live Engineering Team, said, “The goal is to detect bad words in real-time voice conversations and beep them to users who do not want to see or hear them. “This is a big goal, but we need to take one step at a time,” he said.
Related Posts:
