Shutterstock / Pannochka
Conforming to Betteridge’s law of headlines, the answer is no — or at least, not yet.
But since you already clicked it is worth listening to Googler Peter Norvig’s full response when posed each of these offbeat questions on Wednesday morning, during the Q&A session following his talk at the NASA Innovative Advanced Concepts symposium at Stanford University.
Norvig is certainly the guy to ask. He’s a director of research at Google, where he worked on the company’s award-winning machine language translation efforts. The team achieved major advances in the field by setting machine learning algorithms to work on lots and lots of data, including cross-translated transcripts of U.N. proceedings.
His polite response to both questions, which arrived one after the next, is that machine translation depends on at least two sets of data, the raw language and what it translates into. Until we have the latter set from animals or aliens, machines won’t be facilitating any cross-species or intergalactic conversations.
“The more free (your models are), the more data you need to squeeze that model down — and I don’t know how much data we’re going to have when we start communicating with the aliens,” he said, smiling as he got around to the alien bit.
He did raise the possibility that we could closely track animals and possibly piece together a rudimentary understanding of what some of their pips, squeaks or barks mean — but stressed it would “tough.” One person in the audience pointed out the average pet’s vocalizations may boil down to little more than variations on “I love you” and “give me food.”
Norvig noted there’s an added complexity when it comes to communicating with animals, a point he made by referencing a line from philosopher Ludwig Wittgenstein.
“If a lion can talk, we wouldn’t be able to understand him,” Norvig quoted him saying, then added, “because we don’t know what his world is like and what they think about.”
Listen to Norvig really earn his paycheck at around the 59:40 mark: