Companies have rushed to embrace deep learning’s potential in their efforts to automate their enterprises, often with an eye towards replacing as much of their human workforce as possible or to scale their operations without expanding their hiring. An endless stream of success stories tout AI’s success in replacing an ever-growing array of traditionally automation-resistant jobs, while developers are hard at work finding ways to replace the rest of them.
Facebook represents this contradiction. While publicly touting the company’s AI-driven future and investing heavily in building a world-class AI research staff, the company is also rapidly hiring human content moderators. Even as the company increasingly deploys AI algorithms to moderate speech on its platform, it has more than 15,000 community operations staff and growing.
Much like an infant’s simplest movements spark wonderment in new parents, so too do the most basic of AI accomplishments give rise to an image of intelligent machines just around the corner. The simple fact is that while impressive compared to past algorithmic approaches, today’s most advanced AI systems remain primitive compared to even the youngest human child.
Content understanding showcases both deep learning’s greatest strengths and its greatest weaknesses. Today’s algorithms are light years ahead of where they were just a decade ago, yet at the same time, they struggle immensely to cope with the complexities of human discourse. Machines still see imagery largely through the lens of metadata subject tags applied through classifiers, while textual posts are understood through similar classification or simple embeddings.