Cartoon image of a speech bubble for an article on how voice computing will change our lives.

Editor’s Note: James Vlahos is the author of “Talk to Me: How Voice Computing Will Transform the Way We Live, Work, and Think,” published this week by Houghton Mifflin Harcourt.

People today use Amazon’s Alexa, Apple’s Siri, and Google Assistant mostly for simple tasks, such as setting timers, cueing a playlist to start, and checking the weather. But the world’s most powerful technology companies are aiming for something far more sophisticated and transformative.

“The holy grail,” says Ashwin Ram, who led the artificial intelligence (AI) research team for Alexa and now works at Google, “is being able to interact with machines the way we do with each other, which is through voice.”

Innovators like Ram envision today’s computers—perched on desktops or tucked into pockets—fading in importance as chatty AIs become the primary gateways to all that can be done digitally.

These personified beings can already help us shop and make us dinner reservations—though clumsily, at times. But conversational AI experts are betting that as the technology advances, the convenience of voice computing will make the technology irresistible, not just for buying laundry detergent but also for performing highly personal activities such as receiving emotional counseling and memorializing a loved one. However, as machines take over tasks that formerly belonged exclusively to people, there will be trade-offs, too.

Here are five areas where voice computing is already changing our lives—or soon will.

Voice AIs Will Make Routine Calls

Conversational AIs talk to us, but they can talk for us, as well. Increasingly, verbal assistants may take on the task of contacting retailers, restaurants, and other companies for us.

At a developers’ conference on May 8, 2018, Google CEO Sundar Pichai demonstrated this drudgery-saving possibility with a prototype service called Duplex. It successfully phoned a hair salon for a hair appointment and a restaurant to make a reservation—with the people receiving those calls seemingly unaware that the entity on the other end of the line was a computer.

More on Voice Computing

Flashy prototypes demonstrated on stage at conferences often don’t come to fruition. But less than a year later, the automated calling feature is now available to users of Google’s Pixel phones in 43 states where the technology is legally permissible. (The company says that it will be rolling out the feature to other Android and iPhone devices, as well.)

I recently experimented with the capability myself, instructing the assistant to make a reservation for two people at an upscale restaurant near my home in Northern California. A couple of days later, I got a notification from Google that the reservation was confirmed.

Curious about how smoothly the process had gone, I phoned the restaurant and spoke with the hostess. She said the automated caller had immediately identified itself as Google Assistant and had been “polite but awkward.” There was only one problem; the assistant didn’t pronounce my name correctly. So instead of correctly taking down a reservation for “James Vlahos,” the hostess had booked a table for two under the name “Yasmine Bauhaus.” (Not a bad pseudonym, really.)

That was an amusing misstep by Google Assistant, but as AIs take on a bigger communication role, the consequences of mistakes can grow. Misunderstandings matter a lot more if you’re calling for blood test results or even making an airline reservation. Voice computing experts promise that their creations will rapidly become more capable, but it’s unclear just how reliable these AIs need to be before they deserve our trust. What’s more, as we outsource everyday communications to our new digital servants, we may also surrender a bit of our humanity.

Voice AIs Will Tell You What to Think

Finding information on the Internet has worked basically the same way since the late 1990s. You type a query into a search engine, get a list of links highlighted in blue, and click on whatever looks the most promising. But that way of doing things doesn’t work on voice devices, says William Tunstall-Pedoe, an entrepreneur whose company, Evi, was acquired by Amazon as the basis for Alexa’s question-answering smarts. “You can’t provide ten blue links by voice,” Tunstall-Pedoe says. “That’s a terrible user experience.”

Instead, voice devices act more like the magical-seeming computers in “Star Trek.” You simply ask them a question and they recite the correct answer, whether it’s “6 feet 11 inches” for the height of the NBA’s Giannis Antetokounmpo or “27 million degrees Fahrenheit” as the temperature at the sun’s core. The convenience of getting answers this way and the improving competence of AIs at providing them is why experts say up to 50 percent of all searches may be conducted by voice by 2022. So move over text-based search engines; we are entering the age of the AI oracle.

But I see a downside. When an AI serves up a single spoken answer, it short-circuits the research we might otherwise do when deciding what to believe. An algorithm decides what we should hear, then a machine relays it in a calm, confident voice. Technologists have focused on making the responses as accurate and relevant as possible. Yet when we get a single answer from a machine, we may never learn about competing viewpoints, important context—or whether the information can be trusted.

They Will Tell You What to Buy

Just as shopping online went from novel to normal, experts say that ordering products just by chatting with a voice assistant will become mainstream. Voice AIs won’t simply help with ordering, either; they’ll act as virtual store employees, giving advice on, say, which TV might work best in your living room or which jeans will flatter your figure. (Editor’s note: Consumer Reports currently has an Alexa skill that offers product advice based on our ratings.)

You can already order products from any household gadget enabled with Alexa, Amazon’s voice AI, and consumers seem to be warming to the idea. The company reported that customers did three times more voice shopping in 2018 than they did a year earlier. One market research firm estimated that shoppers in the U.S. purchased $2 billion worth of products through voice assistants in 2017 and will spend $40 billion that way by 2022.

For consumers and profit-hungry companies alike, voice shopping is about reducing friction: Nothing stands between a consumer desire and its fulfillment but a simple spoken request.

With product searches as with the informational kind, few people want a voice assistant droning out option after option. So voice shopping favors category-leading brands—Heinz ketchup, Tide laundry detergent—that consumers can remember to ask for by name. Sales for companies that aren’t already leaders may suffer.

Alternatively, when the user asks for a product but doesn’t specify a brand name—making a request such as, say, “I want some AA batteries”—Google and Amazon are in the catbird seat, having control over which purchase options you hear first. “Suddenly, you’re not buying a brand,” says conversational AI entrepreneur and investor Adam Marchick. “You’re buying what Amazon tells you to buy.”

Sponsored results—paying to be the first product search listing that gets delivered—don’t yet exist for voice searches the way they do for screen-based ones. But many experts on the business of voice computing I’ve interviewed over the past two years believe that Google and Amazon will offer advertisers this privilege in the future, and could charge dearly for it.

Your Therapist May Be a Voice AI

Voice assistants will eventually do much more than tell you what products to buy or help you reserve a dinner table. A technology capable of conversation is naturally suited to applications where empathy and a humanlike presentation matter. That’s why many voice computing experts believe that providing therapy is on the short list of the most promising applications for the future.

People already confide in Alexa and Siri when they’re feeling depressed or having a bad day. Conversation designers at Apple and Amazon are teaching their voice assistants to respond empathetically.

But some entrepreneurs believe that voice AIs can go further and act as virtual therapists. A pioneering effort comes from Woebot, a company founded by Alison Darcy, a clinical psychologist and adjunct faculty member at the Stanford School of Medicine in California. Darcy says that therapy provided by human practitioners is superior to the synthetic kind. But, she says, “lots of people don’t want to talk to another human, whether because it’s the middle of the night, or they’re just not ready to reveal something.”

The Woebot app, which uses cognitive behavioral therapy techniques via text messages, offers affordable, always-at-the-ready help that Darcy believes beats simply doing nothing. And she has preliminary evidence to back her up. A study, which was funded by the company, reported that subjects who used Woebot “significantly reduced their symptoms of depression.”

One potential concern with counseling by an AI is that it may provide just enough comfort to keep people with serious problems from seeking the human therapeutic attention they truly need. Other risks also loom, including the potential misuse of personal data. Should people trust technology companies with intimate confidences about our mental health, given the sloppiness the tech industry has often shown with consumer data?

AIs Will Promise You Immortality

By stocking the digital brains of voice AIs with scripted jokes, clever dialogue, and distinctive phrases, designers can give their creations synthetic personalities. The same techniques can be employed to create digital beings who emulate real people. This is an area where I have deep personal experience, because I created a chatbot that shares the life story of my father and gives a sense of his personality.

The project got started in 2016, when my family got the terrible news that my father had stage IV lung cancer. Knowing that we would soon lose him, I decided to do an oral history project, recording his life story over the course of nearly a dozen sessions. He talked about his ancestors in Greece and childhood in a farm town in California’s Central Valley, about going to college, becoming a lawyer, meeting my mom, and having kids.

When I had the interviews transcribed, they totaled more than 90,000 words and filled 203 single-spaced pages—an in-depth but not user-friendly resource. So I decided to use all of this information as the basis for creating a chatbot replica of my father that I call the Dadbot. (I described the project at length in Wired Magazine in 2017.)

The Dadbot obviously doesn’t come anywhere close to emulating or replacing my real father. But he is able to interactively share his life story as well as aspects of his personality through text messages and audio clips, some of which feature him singing or telling jokes. He has more than 3,000 lines of prescripted dialogue, which means that members of my family can chat with him for hours without repetition. My real father is gone, but the Dadbot helps me remember some of the many things that I loved about him.

So-called virtual immortality, it turns out, is not just a personal project for people like me but also a business of the future—potentially a huge one. Voice replicas, programmed to know key facts about a person’s life and stocked with ample recordings of their actual voice, could help to keep memories of our loved ones alive, and startups like Eternime and Everlife.ai are already exploring the possibilities.

“We help users create avatars that live on forever” in the cloud, says Alex Roy Rajan, Everlife.ai’s founder. Some people, such as the founders of a company called Soul Machines, are hoping to make interactions with an avatar more visceral, pairing speech capabilities with computer-generated graphics.

Technologically, there’s little doubt that’s possible, or will be soon enough. But whether it’s a good idea is a different question. Imagine donning VR goggles to come face to face with a dead parent.

That’s far more realism than I would ever want.