Facebook put out some great figures for their first quarter 2016 earnings call yesterday, but a few words from Mark Zuckerberg springing from an after-earnings Q&A session may be a bit more interesting. When an investor asked about future instances of leveraging the tech behind Facebook’s A.I. based chat bots, Zuckerberg didn’t have anything in particular to say, such as a specific product or service that the technology would be benefiting. What he did say, however, was that he believes that, within the next 5 to 10 years, we may see a day when A.I. is better than humans at basic cognitive tasks, opening the door for things like actual, unsupervised learning and the infinite possibilities that come with it.
As with any chat about advanced A.I. that includes topics like machine learning and neural networks opening up A.I. to massive possibilities, the elephant in the room was, of course, what may happen to the human job market. In theory, if an A.I. can learn, can perform basic tasks at a superhuman level and can be backed by robotics, just about any non-creative job that a human can perform could be done by an automaton. Zuck addressed this head-on, saying that he sees A.I. bots edging out humans in low-level fields like customer service, or becoming training liaisons for new employees entering higher-level positions, once the A.I. has gone through its own training under the supervisor who would normally train the individual that the bot will be training.
He went on to talk about the various solutions currently on the table to address the possibility of a much, much smaller available labor pool for humans in the future. Although almost everybody having free time all the time may sound nice, it does raise some rather hairy economic questions. One solution he said was being talked about extensively was the concept of a guaranteed basic income for everybody, regardless of their education, social standing or job. Another possibility was the transition of the full human workforce to higher-level jobs, as well as the creation of different and new categories of work that A.I.s may have a hard time with. There is still the issue, of course, of A.I. systems not being terribly good with unassisted and unsupervised learning, a hurdle that would need to be overcome before bots could affect the human labor pool in any truly sweeping and meaningful fashion.