facebook寻求在恐怖袭击视频之后改进人工智能

In further reaction to the New Zealand terrorist attack, circulated via video on its network, Facebook (FB -0.7%) says it will pursue faster reactions from artificial intelligence when it comes to reacting to live video on its platform.

The company uses AI to detect and prioritize videos that are likely to contain suicidal or harmful acts, and enables the company to proactively detect most of the content it removes. "But it's not perfect," Facebook says.

For one thing, AI systems need large volumes of training data but terror events like this one are "thankfully rare."

The company says it will improve matching technology "so that we can stop the spread of viral videos of this nature, regardless of how they were originally produced." It also will explore how AI can be used for faster reaction on live-streamed videos, and get to user reports faster.

Facebook says it removed the attacker's video within minutes of outreach from New Zealand police, and that it was viewed fewer than 200 times during the live broadcast (during which time no user reported the video, it says). Overall the video was viewed about 4,000 times before removal from the site.

Then in the first 24 hours, Facebook removed more than 1.2M videos of the attack at upload -- before they could be seen -- and another 300,000 copies were removed after they were posted.