Yup, Siri/Alexa can't detect my Manc accent properly so I wish them the best of luck.
Activision will add real time in-game voice chat moderation in Call of Duty, AI powered detection - Live Today in MWII/WZ in NA, worldwide with MW3
- Thread starter Bishop89
- Start date
Yup, Siri/Alexa can't detect my Manc accent properly so I wish them the best of luck.
OP
OP
it says actions wont be taken instantaneously and are subject to review
some of you could at least pretend to have read more than just the word "AI" in the thread title lol
The AI shouldn't ban people. Just detects and reports the players.
Just for comedy sake, if someone uses this defence they must on the spot demonstrate the correct pronunciation of Niger in French and/or any of the local languages. Obviously still ban them, but I want to hear the attempts anyway.
As for the system, I think it's actually not a bad use for AI, as in this is a job that can't feasibly done purely by human power, and shouldn't be done by pure human power as that would be a mentally shit job.
A little dystopian? Maybe, but if anything encourages people to clean up their act it's the panopticon listening in on their hate speech.
"Listen here you faintcolt ****, you are now banned so stick your ******** **** in a blender, now **** ***. Have a lovely day, heil *****"
View: https://www.youtube.com/watch?v=gmhEXjM7iTw
Glad we're now forcefeeding robots the racist diatribes of CoD players. This will absolutely give them a pleasant, optimistic view of humanity when the time comes for them to overthrow us. /s
Seriously though, this seems like an interesting system. Curious to see what the results are.
Seriously though, this seems like an interesting system. Curious to see what the results are.
This will likely be a net positive for the game in general. Haven't used game chat in about a decade since party chat is just so much more effective.
Not really sure if we can expose AI topics around Era but Vertex AI from Google can distinguish a lot of languages with high accuracy. I already tested on my work with Portuguese and works really well.
Good idea, assuming that there's a human actually listening to the clips and verifying the bans before they go through.
I'd also like to see some comparative quality benchmarks between languages, if they exist. There's no way the other 12 languages have as much training data as English and Spanish.
I'd also like to see some comparative quality benchmarks between languages, if they exist. There's no way the other 12 languages have as much training data as English and Spanish.
It says they "may" be subject to review. The technology only flags voice clips as probably toxic, coming up with a moderation policy based on that data is up to Activision and they could easily just set it up to auto-ban someone who gets flagged X times.
Yeah I'm good, Absolutely no way I can trust where they store that Data. I'll stick with talking outside of the game
Imagine if it wasn't AI...
I wouldn't be surprised if some people would argue that this is stealing jobs that could go to people instead. Good to know a human will have to review the content that is automatically flagged. Sounds like it would be a nightmare having to just listen to all audio to see if something bad pops up with no automation.
That aside, I'm gonna have to make sure the people in my COD group know about this because they have some proper potty mouths. We mostly stick to Discord anyway, but sometimes they like to shout some British-favourite words at people.
Honestly? It's an interesting idea, especially since it reflexes to a human for confirmation. Not really possible to have humans moderating in real time constantly, so trying this out to see how it does is pretty neat. If it does a great job, then great! If it false flags, then humans will see that and they can try making adjustments. If it's just totally awful then they'll likely get rid of it after a year or two.
It feels like CoD has done more novel/interesting things to combat toxicity and/or cheating the last couple years than most online games.
It feels like CoD has done more novel/interesting things to combat toxicity and/or cheating the last couple years than most online games.
I think this is very fair when it comes to stuff like party chats, but I wouldn't expect any sort of privacy in team chat of a public matchmaking lobby or a server hosted by the developers.
A big selling point of the service is that they filter the audio data before processing to only find the relevant parts and cut down on costs.
Though they don't really say exactly how the triaging or voice analysis works, which makes me kinda sus. This is the real [citation needed] in all this, especially because different games have different codes of conduct and they're obviously not training a new model for each game type.
I don't believe for a second this can be reliably automated in a real video game environment. It's like saying they've found a way to detect hard R's.
The reason why I'm not worried about the overall system is because it's going to be tuned to cut down on the amount of audio processing being done for cost reasons, which in turn will lower the false positive rate. It could still go to shit if they auto-ban people based on these results.