Difficult for AI to identify hate speech: Facebook’s Mark Zuckerberg

Difficult for AI to identify hate speech: Facebook’s Mark Zuckerberg
Mark Zuckerberg  |  Photo Credit: Reuters
26 Apr, 2018

In an earnings call with analysts, Mark Zuckerberg, the founder and chief executive of social networking giant Facebook, said that artificial intelligence cannot easily distinguish hate speech from other offending content, a report in financial news website Business Insider said.

“It's easier to build an AI system to detect a nipple than what is hate speech,” the report quoted Zuckerberg as saying.

Facebook has been using artificial intelligence to weed out objectionable content, including pornographic and terrorism-related material, the report said.

The CEO also said that Facebook's AI engine was able to identify a majority of terrorism-related content and remove it immediately but failed to identify hate speech reliably. According to Zuckerberg, it will take AI 10 years of development to flag offensive content, Business Insider reported.

A few years ago, the company was involved in a controversy over sharing of breastfeeding images. It had barred users from posting any photos exposing nipples for a few days but later allowed them as long as no exposure was identified in photos, the report added.

Zuckerberg’s comments come a day after the social networking site released a rule book for the types of posts it allows on its platform. While the company earlier provided a brief and general version of what is permitted on its network, it released a longer document on its website to clear up confusion and be more open about its operations.

Facebook has faced fierce criticism from governments and rights groups in many countries for failing to do enough to stem hate speech and prevent the service from being used to promote terrorism, stir sectarian violence and broadcast acts including murder and suicide.

In a separate development, the Indian government has sent a second note to British data firm Cambridge Analytica in order to gain more information on the company’s operations in India, The Economic Times reported. In response to the government’s first note, the firm said that it did not have Facebook data of Indian citizens, though it has worked on projects in the country. The government deemed the response by the firm as ‘cryptic’ and ‘evasive,’ the ET report added.

The government also sent out a notice to Facebook seeking information on its data security architecture, the report said.