The world is waking up this morning and realizing that Google is much further toward a conversational AI than we thought. To be honest, I was very surprised by the advancement in their demonstration of Google Duplex:
There are immense issues around this whole demonstration. Many legal ones have been outlined here by a Privacy Attorney, but it is not merely the legal questions raised.
When the video shows the Google Assistant saying “hmm” or “mmm hmm” the audience laughs, but to be honest, it sounds like a very, very nervous laugh. And we should be very, very nervous.
This makes us all nervous because what is a common human “filler” in voice speech patterns is unmistakably being inserted artificially by a machine that does not need to pause. The machine doesn’t need to stop and ponder, it has the answer, and the next statement, and the next, all queued up.
This is a machine using “hmm’s” to slow down to our speed. To human speed.
This is AI treating humans like a QWERTY keyboard.
QWERTY keyboards have an odd layout with many of the most used keys and pairings far away from each other and on the statistically less-dominant hand. One theory for the design is that by separating the most common pairings, Christopher Sholes, who invented the typewriter in 1868, was attempting to slow down any human operator’s ability to enter input into the machine (typewriter). This was supposedly done to help minimize the machine jamming, which occurred when two keystrokes hammered the paper at roughly the same time.
From Urban Dictionary (underline mine):
QWERTY is the keyboard layout in the U.S for most keyboards. This layout was used in early typing days to slow down typists in order to have the typewriter in use not jam as often due to lack of typist’s speed. QWERTY is named as such because “Q.W.E.R.T.Y” are the first 6 letters on the keyboard.
Now it has been widely debunked that the QWERTY keyboard really was designed this way, however, the term “QWERTY” has stuck. So when an engineer says something is “QWERTY”, they are saying that the design is inherently an inhibitor to slow its operation.
The Google Assistant video shows the AI purposely slowing down it’s ability to respond and communicate with humans. The “hmm’s” are some of the first (of many to come) examples of machines purposely slowing themselves so as to not “jam” us humans up. And this is the reason for the uneasy laughter you hear.
Machine learning and AI have been around for a long, long time. But Google just demonstrated that, perhaps, we are much closer to an intelligence that recognizes the limitations of their counter-party in a conversation. Namely, us.