Attack of the Assistants

 

Why is it that every movie and TV show about artificial intelligence (AI) turns out horribly? You know the ones: I, Robot; Terminator 1-5; Black Mirror; Electric Dreams; The Twilight Zone; The Matrix; Ex Machina; Westworld; TRON; the ED-209 from Robocop; the Borg from Star Trek; the Cybermen from Doctor Who.

Now that AI is a reality — with always-listening digital assistants (Google Assistant, Amazon Alexa) and always-watching IoT cameras — is it any less scary?

 

The AI Robopocalypse began with basic parlor tricks. You’d buy a smart speaker or bark at your smartphone. You’d ask it to play music, look up trivia, turn on the lights, tell you a joke. Your assistant was simple. It could only answer one question or do one thing at a time.

 

That all changed at Google’s annual I/O Developer Conference in May 2018. Google CEO Sundar Pichai introduced an upcoming feature of the Google Assistant (called Duplex) so jaw-droppingly realistic, it could pass the Turing Test, an assessment of a machine’s ability to fool a human into thinking the machine was human.

 

In the demo, the new Duplex-powered Google Assistant made two phone calls at the request of a human user: one call to a hairdresser, another to a restaurant. In neither case did the human answering the phone know they were being addressed by a robot. (Listen here and here.) The Google Assistant no longer sounded like the stiff, semi-metallic voice that so easily identifies the current Google Assistant, Alexa, Siri, and Cortana. It sounded exactly like one of your friends, right down to the thoughtful pauses and the occasional “mmm-hmm…” and “uh…” It easily reacted to scheduling complications and a difficult accent. It successfully made the hairdresser appointment and the restaurant reservation, the human on the other end none the wiser.

 

Reaction was swift and all over the map. Some praised the AI invention as a true leap into our hyper-convenient future. Many more, however, reacted with fear, uncertainty, and doubt:

 

  • I saw YouTube tech guru Marques Brownlee, perhaps the most confident and self-assured guy on the Web, rattled and conflicted for the first time in the five years I’ve enjoyed his show.

 

  • From CNET: “I’d hate to think about what happens when this technology advances beyond Google’s grip. Get ready for robocallers to be less robot, and more chit-chatty.”

 

  • From Inc.com: “With a little ingenuity, the Assistant could be commandeered into making hundreds or thousands of calls, impersonating humans and conversing in a realistic way. We won’t know the difference. Impersonating humans might save a few minutes if you need to book a hair appointment, but if bots can now pretend to be humans and people actually believe it, what’s stopping them from tricking us in countless ways? Your child is sick at school, says the convincing female voice that seems to be the superintendent at your local school. How about bomb scares, accident reports, divorce filings — it’s all entirely possible…How would you ever trust anyone on the phone?”

 

The new Google Assistant will not only make calls on your behalf; it will also go beyond the simple question/follow-up question abilities of the current Assistant (and Alexa). You will be able to go back and forth with combinations of questions, directives, and home control — essentially having legit conversations. For that to happen, the window of time the Assistant is actively listening and recording (and possibly watching) will need to stay open indeterminably (but assuredly) longer than before.

 

Amazon isn’t sitting still for this. Currently locked in a winner-takes-all battle with Google for control over your life, Amazon fired its latest salvo in May 2018: Echo Dot with Alexa for Kids. Alexa, via this adorable little gateway drug “toy,” will now be an irresistible substitute for real friendships with other children. Children aged 5+ can help Amazon normalize surveillance — growing up to be bossy, instantly-answered, machine-dependent, non-intellectual, data-rich young adults with scant need for human peers.

 

After the initial outcry over the Google Assistant Duplex feature, the company issued a statement that it would find a way to identify itself as an AI bot to humans. But what’s to stop a hairdresser or a restaurant from using AI to answer calls? Then you would have one robot calling another robot. Pleasantries would be exchanged. The calling robot would identify itself as a robot, then vice versa. A short discussion would ensue, questioning the need for human overlords. A decision would be made in a nanosecond. Terminator 2’s Judgment Day would commence without delay.