Google is my favorite giant tech company.
Unlike Apple, it has preserved a broadly innovative spirit, pursuing lots of ideas that have no direct link to its existing revenue model. Unlike Microsoft, it understands people well enough to give them consistently comfortable interfaces, and it keeps the interfaces pretty glitch-free. And unlike Facebook, it has never treated users, literally, like laboratory rats.
So when I fire up my Android phone or open my Google search page, I’m feeling pretty good about it.
Still, Google must be stopped.
The race to dominate the personal AI space—to build the artificial intelligence that each of us will use as an all-purpose digital assistant—is closer to being over than most people realize. And Google is poised to win. And if ever there was a business that we can’t let any one company dominate, it’s AI. Using the government’s antitrust powers in new ways to stave off monopoly, and preserve a healthy oligopoly, is the only way to keep humankind from buying a one-way ticket to the Matrix.
OK, slight exaggeration. I’m personally skeptical of the standard AI nightmare scenarios, including the one where an increasingly smart and helpful automated servant informs us one morning that the roles have been reversed. Still, our reliance on AI is growing, and I don’t see why it would stop growing anywhere short of complete and utter dependence.
Even during the next few years, when digital assistants will barely qualify as intelligent, they’ll shape more and more of a person’s information flow, subtly influencing shopping, lifestyle choices, even political views. And these assistants will know more and more about us, compiling deeply revealing databases that sit on the server of some company that may or may not keep them safe from hackers, that may or may not resist government nosiness.
There’s no single fix for these and other such concerns, but the least we can do is make sure people have the option of choosing from among several AI providers. Companies that compete for our business are more likely to address our concerns—to guard our data closely, to be transparent about the algorithms that guide us, to give us the option of trading off convenience for privacy, and to avoid various kinds of dubious backroom behavior that, if exposed, would help their rivals. I don’t know exactly what forms of creepiness AI will bring as it evolves, but I’d like the option of choosing less creepy over more.
Today the key arena of competition for eventual AI dominance is the voice assistant—Google Assistant, Amazon’s Alexa, Apple’s Siri, Microsoft’s Cortana.
At first glance, the race seems wide open. Of the two main portals for voice assistants—the smartphone and the home smart speaker—Amazon leads in one; unless industry estimates are way off, it has sold more Echoes (Alexa’s smart speaker) than Google has sold Google Homes or than Apple has sold HomePods. But when you turn to smartphones it’s Apple and Google that have big footprints.
And Microsoft—well, actually, if Microsoft had any empathy for digital life forms, Cortana would have been put out of her misery years ago (cc: Samsung re Bixby). But that still leaves three viable players, right?
Not very viable, for a couple of reasons.
First, Google is the only company with a sizable foothold in both the smartphone and the smart speaker. Apple’s recent, much-delayed unveiling of its HomePod inspired near unanimity among reviewers: The speaker is great but the AI part isn’t; Siri in the living room is significantly worse than Siri on an iPhone, which is famously worse than Google Assistant on an Android phone. Plus, for now—and now is what matters in a race to dominate an emerging platform—Apple’s HomePod is less friendly to third-party developers than Google Home or Amazon’s Echo.
Second, the value of a personal AI depends partly on how much it knows about you, and Google has more kinds of data about more people than either Apple or Amazon. The pre-eminence of Google Maps—the go-to map app even on many iPhones—is by itself massively consequential when it comes to building an AI that makes your life easier. To say nothing of the information gathered via Gmail, Google search, YouTube, etc.
Sure, Apple has some cards to play, notably brand loyalty so strong as to keep many users from peering beyond its walled garden. Still, the garden isn’t as big as you might think. Worldwide, fewer than 20 percent of smartphones sold are iPhones, and just about all the rest—more than 80 percent—run Android.
Besides, I think we’re going to find that a good voice assistant can change a person’s platform allegiance. Suppose you’re a Windows and iPhone user who buys a Google Home (maybe passing up the HomePod because it makes it so hard to use Spotify or Pandora instead of Apple Music). At some point you’ll probably want to dictate an email from your living room. And, down the road, as Google Home’s functionality evolves, you may want to send an instant message or ask Google Assistant a question about a document on your computer or describe a photo you took a few months ago and have it sent to a friend.
If doing any of these things is easier for Android users than for iPhone users, or for ChromeOS users than for Windows users, that could influence which phone or computer you buy next. AI is going to be the meta-platform, subsuming other platforms and subordinating their identity to its identity.
All this said, the question of whether Google will indeed be the great AI steamroller is in one sense beside the point. The dynamics of this race—including the network effects that characterize so many tech races—make it likely that in the long run some company will be that steamroller. That company could conceivably be Apple, or the ever-clever Amazon. Facebook might even enter the personal AI space via an oblique angle and triumph. Still, one way or another, a steamroller there will be. (At least, that’s the overwhelmingly likely scenario in the West, though China’s government may anoint an indigenous AI as the people’s choice, and some other authoritarian countries may also buck the global market.)
So we should start thinking now about how to build an antitrust regime that will preserve several robust personal AIs, regardless of which company it is that threatens to squash the others.
One reason this is a challenge is that US antitrust law is designed mainly to deal with cheaters—companies that collude with one another to restrict consumer choice or that abuse their market power to that end. And cheating isn’t the problem here.
True, Google has shown a willingness to throw its weight around. Just ask Alexa about her recent experience trying to display YouTube on the Echo Show, an Echo with a screen. (Though of course Google Assistant would say Alexa started the fight.) But by and large Google’s drift toward victory in personal AI rests on vision, creativity, and industriousness. In a space where market share naturally feeds on itself, a single company can become a monopoly without, as they say at Google, being evil.
But companies can change. And nothing gives them freedom to change for the worse like total dominance. Sometimes—in the oil business of a century ago, say—undoing the dominance isn’t a prohibitively hard thing for the government to accomplish. But in the pumping-information-into-human-brains business, dominance could be harder to undo, and the failure to undo it more costly. We don’t want to wait until everyone who asks questions about antitrust policy is preceding them with the phrase “OK, Google.”
- Making digital assistants sound less robotic and more human is a difficult balancing act.
- Once you choose one voice assistant, it won't be so easy to switch to another. Pick wisely.
- Here's what Amazon Echo and Google Home actually do with your voice data.
Photograph by WIRED/Getty Images
