A recent explicit Meta report harassing people on Facebook and Instagram

its researchersFacebook shared it new statistics on the bulk of bullying, hate speech and harassment on its platform. The new figures, released by the company’s most recent reports, come as Meta faces a major overhaul of its ability to protect consumers and implement its global standards.

His most recent report is the first time a company has shared metrics “spread” around tortured and tortured on his platform. “Quantity” is the number that Facebook uses to track the violations that go through its cognitive machines. “This represents the amount of error that people see on another person’s computer,” Integrity Vice President Guy Rosen told reporters at a briefing.

According to the company, the prevalence of violence was between 0.14% -0.15% on Facebook and between 0.05% -0.06% on Instagram. “This means that harassment and bullying were seen between 14 and 15 per 10,000 Facebook shows and between 5 and 6 per 10,000 views on Instagram,” the company explains in a statement. Instagram in particular has faced questions about its ability to deal with harassment and bullying. The company introduced a number of innovations hatred of violence measures earlier this year after a number of UK players described in detail their experiences racist violence on the app.

Most importantly, the company claims that this “spread” simply takes away what Facebook and Instagram remove without a user report. This means that the statistics only take a small amount of all the bullying, because bullying is not always difficult for the machine to detect.

This distinction is confirmed by the revelations of the Facebook Papers, a series of posts that were announced by former co-founder Frances Haugen. According to the posts he shared, his Facebook researchers compare that the company is capable of dealing with about three to five percent of hate speech on its platform, meaning that many go unnoticed and are allowed to defame News Feeds readers.

Facebook has done it over and over again pushed back on these claims, and has also spoken of the “spread” which he shares in his visual reports. But as researchers have done he pointed, a company accounting that “spreads” can hide the actual amount of breach on the platform. This is because Facebook machines are not always reliable, especially when recognizing content in languages except English. The revelations echoed the claim that Facebook puts profits ahead of user protection.

“We have no incentive, whether commercial or not, to do anything other than ensure that people have a positive experience,” Rosen said Tuesday. “I think it’s not true that our algorithms are simply designed to end a relationship. We always look at how we choose to deal with these problems.”

In a recent report, Facebook also reported that hate speech dropped sharply in the fourth quarter, and the rate dropped from 0.05% in the last quarter to 0.03% this quarter. The company also reported the spread of hate speech on Instagram for the first time, saying hate speech was at 0.02% or about 2 out of every 10,000 episodes seen on its platform.

However, it is important to note that even those with high hopes for these statistics – 0.03% and 0.02% on Facebook and Instagram, respectively – could still mean that millions of people are experiencing hate speech every day, due to the sheer number of users and pieces of content. it is posted on the platform every day.

Individually, Facebook added its researchers is working on a “new phase of AI research called ‘shot-shot’ or ‘zero-shot’ learning,” which will help them to teach AI models more quickly. Instead of relying on large datases to teach models, say hate speech, it can help models who “can learn to recognize something from just a few academic examples, or just one example,” the company wrote. Facebook did not say how long it would take to conduct the research, but it does indicate that the company is still developing advanced AI to address major challenges.