Teaching machine systems 'how the world works' is, of course, a hugely complex concept. The approach is yet to be worked out, but could include systems drawing from YouTube for tips, Norvig hinted.
“If we want it to get all those things right we’re going to also have to say, you know, here’s all the YouTube videos in the world, and here’s how glasses fall off tables and now in addition to understanding the structure of sentences you also have to understand the structure of the world. And nobody has tried to do that yet. But to get to perfect translation we’ll need that.”
The implications for accurate and instant translation are huge, not just for the companies pursuing it but potentially all of humankind.
As Zuckerberg put it: “Understanding someone's language brings you closer to them, and I'm looking forward to making universal translation a reality.”
Norvig agrees: “We did so much better than I thought we would ever do with this end to end machine translation. And having easier communications between people who otherwise would not be able to talk to each other – I’m excited about that opening up the world.”
Sign up for CIO Asia eNewsletters.