Page 16 - Human Moderators News Today : Breaking News, Live Updates & Top Stories | Vimarsana

Conducting itself before. >> well, it really does. they have beefed up their artificial intelligence operation so they're using technology to target key words to locate locate content that is terrorism related or nudity related. they're using human moderators now. they've hired more staff to review content and find hate speech. that is a little bit more nuanced. you need to have someone with cultural -- to understand the cultural nuance to be able to identify hate speech. >> they say they disabled 200-plus apps like the one cambridge analytica had used to gather all this data. what does this m...

Score, you're using the word toxicity. would hate be a good substitute for that. >> it's interesting so we are experimenting with lots of different ai models. we began with toxicity because it was the easiest to measure based on the training data but were also exploring a sheen learning models that return a score of how insubstantial it is or how off-topic it is. we are experimenting with lots of things. i think what you said is important, when i tell people about this one of the first things they raises this question of isn't the censorship. my response is, look at the default position. there...

Score, you're using the word toxicity. would hate be a good substitute for that. >> it's interesting so we are experimenting with lots of different ai models. we began with toxicity because it was the easiest to measure based on the training data but were also exploring a sheen learning models that return a score of how insubstantial it is or how off-topic it is. we are experimenting with lots of things. i think what you said is important, when i tell people about this one of the first things they raises this question of isn't the censorship. my response is, look at the default position. there...

You're literally giving them a score, you're using the word toxicity. would hate be a good substitute for that. >> it's interesting so we are experimenting with lots of different ai models. we began with toxicity because it was the easiest to measure based on the training data but were also exploring a sheen learning models that return a score of how insubstantial it is or how off-topic it is. we are experimenting with lots of things. i think what you said is important, when i tell people about this one of the first things they raises this question of isn't the censorship. my response is, look...

"why have they been found to not violate guidelines or probably found "to not violate guidelines?" and he said that was a mistake, that the technology was marking these comments as "benign" which meant that they weren't being followed up for a human review and they weren't being taken down. he said that this technology is now being fixed and that instagram was taking all of these reports very seriously and that they would be following them up properly. well, i think that we know this has been an issue for awhile. racist abuse on social media is nothing new. and even earlier in the year, some l...