Google’s new AI is magic — but raises a lot of questions

Source: Pixabay
Source: Pixabay

“Any sufficiently advanced technology is indistinguishable from magic,” wrote sci-fi author Arthur C. Clarke. This week in Silicon Valley, Google (GOOG, GOOGL) proved it.

Several times during CEO Sundar Pichai’s keynote presentation at the company’s I/O developer conference, the audience of 7,000 gasped or burst into applause. That’s because some of Google’s massive investments in artificial intelligence are now paying off—in features that truly are magical.

Gmail can now autocomplete not just words you’re starting to type, but entire sentences. Google Photos can now colorize a black-and-white photo with a single click.

And in an absolutely mind-frying demonstration, Google Assistant booked a haircut appointment by placing a phone call to a human receptionist at the salon and having a conversation on your behalf, sounding indistinguishable from a real person. The receptionist probably never knew she’d been talking to AI.

But guess what? Here’s another famous quote, this time from Stephen Hawking: “Artificial intelligence could spell the end of the human race.

Google’s breakthroughs today don’t exactly smell like doomsday. They do, however, bring up a few questions that somebody ought to ask.

Smart Compose

Google is rolling out a new Gmail design to its millions of customers. Once you get it, you’ll be treated to an amazing new feature: autocomplete of entire sentences.

You type “I ho” …and Google proposes I hope you’re doing well.

You type “Does next f,” and Google proposes Does next Friday work for you?

You type “My a,” and Google proposes My address is 830 West Maple, Highland, NJ 12993.

You type “L” near the end, and Google proposes “Looking forward to seeing you!”

To accept one of these proposals, you tap your Tab key; you’ve just saved yourself a line of typing. Otherwise, just keep typing, and the suggestion vanishes.

Here’s the demo portion of the video from the keynote:

There’s no question this AI feature will speed up email processing and eliminate typos. There are a few other questions, though:

  • Won’t this mean that everybody’s email personality sounds exactly alike—like Google’s programmers? What will this do to “voice?”

  • Will people eventually recognize the prose of Smart Compose? Will they be offended that you’ve Smart Composed them instead of taking the time to write original prose?

  • Will you be able to customize the canned responses? To make up new ones? To tailor them to your voice?

  • How good can Smart Compose get? Will Google eventually come up with complete rejection letters, thank-you letters, and breakup letters? You type: “Dear J,” and Google proposes, Dear John: In my heart, I’ll always love you. But in the real world, I can’t keep going like this…

The phone-call assistant

Here’s the video of that phone call to the hair salon. Keep in mind that the male voice is Google Assistant—a completely automated entity. It does a perfect impersonation of a human being, complete with “um”s and “uh”s and seemingly improvised inflections.

You’ve just said, “OK Google. Make me a haircut appointment on Tuesday between 10 and 12.” Invisibly, silently, without your awareness, Google calls the salon for you and has the following conversation (this is an actual recording of an actual test call):

Assistant (calling the actual salon phone number): “Hi. I’m calling to book a woman’s haircut for a client. Um, I‘m looking for something on May 3?”

Human receptionist: “Sure. Give me onnne second….”

Assistant: “Mm-hm.”

Human: “Sure, what time are you looking for, around?”

Assistant: “At 12 pm.”

Human: “We do not have. 12 pm available. The closest we have to that is a 1:15.”

Assistant: “Do you have anything between 10 am and, uh, 12pm?”

Human: “Depending on what service she would like. What service is she looking for?”

Assistant: “Just a woman’s haircut for now.”

Human: “OK, we have a 10 o’clock.”

Assistant: “10 am is fine.”

Assistant: “OK, what’s the first name?”

Human: “The first name is Lisa.”

“OK perfect. So I will see Lisa on May 3 at 10 o’clock.”

Astounding. This Google developer post offers some more fairly amazing examples of this feature in action.

Now, the first concern you might have here is: But what about when the call doesn’t go flawlessly? Surely there are times when the receptionist cracks a joke, misunderstands, or otherwise goes off-script.

To CEO Pichai’s credit, he also played one of those examples—an Assistant attempt to book a restaurant reservation:

Fortunately, Assistant handled the flameout gracefully. It reports to your phone either that it was successful (and has put the appointment or reservation on your calendar) or that it wasn’t.

This robo-calling feature is still just a limited experiment, with no fixed launch date. By the time it’s ready for prime time, let’s hope that Google has found answers to questions like these:

  • Should the person who takes the call know that she’s talking to a robot? It seems somehow manipulative or sneaky to conduct a business transaction that, behind the scenes, is entirely automated. Or does it matter?

  • Google notes that 60% of small businesses do not offer an online booking system. Will they come to recognize when they’re interacting with Google Assistant (or Apple’s, or Amazon’s, or Microsoft’s version thereof)? Will they care? Will YouTube fill up with hilarious recordings of receptionists messing with Google’s AI caller just for spite?

  • Will Google eventually develop a corresponding AI feature for them, the businesses, so they don’t have to waste time on actual phone calls, either? Will the whole thing eventually boil down to customer AI apps interacting with business AI apps—no human conversation involved?

Google Maps Vision

I dearly love Google Maps. But there’s a sore spot: You emerge from a building or subway station, ready to start walking to your next destination—and you don’t know which way you’re oriented. Maps tells you, “Walk south toward Vermont Street”—but which way is south? So you start walking, staring at your dot on the map, hoping that it will start moving on the map so you can figure out which way you’re walking!

Product manager Aparna Chennapragada demonstrated a coming-soon feature: Maps uses your phone’s camera to recognize the buildings at your intersection, and thereby figure out which way you’re looking! At that point, a big arrow appears on the screen to show you which way to walk. (Google is even experimenting with an animated fox that bounds in the right direction, bidding you to follow.)

I really have only one question about this feature:

  • How soon can we get it?

David Pogue, tech columnist for Yahoo Finance, welcomes non-toxic comments in the Comments below. On the Web, he’s davidpogue.com. On Twitter, he’s @pogue. On email, he’s poguester@yahoo.com. You can sign up to get his stuff by email, here.

More from David Pogue: