Last week, a friend texted to ask when I would arrive in New Orleans for her wedding. My phone immediately pulled my flight itinerary from an email to respond to her with the arrival time.
Another friend recommended that I check out an Ethiopian restaurant nearby. My phone instantly loaded a map showing where it was.
And when I took a photo of my corgi, Max, my phone coached me to take a better one by framing him in the bottom-right corner.
The phone that did all this was Google’s new $1,000 Pixel 10 Pro, which arrives in stores this week. Google describes the device as an artificially intelligent phone, a new kind of smartphone that requires persistent access to users’ personal data, including their contacts, location, messages and email, to anticipate their needs. In other words, it’s a phone that essentially uses people’s apps to save them time.
But is this what people really want? I pondered this question as I tested the Pixel phone over the last week, surrendering my data to Google’s A.I. in exchange for some help.
I came away from the experiment with mixed feelings. Streamlining tedious tasks, like sharing my flight itinerary in a text message, was helpful. But it didn’t sit well that I had shared so much information just to save a few seconds.
The A.I. also made mistakes. In one instance, to help with a phone call to an airline, it pulled information from an email unrelated to my flight. (More on this later.)
Google’s Pixel phones have historically not been top sellers in the overall handset market, but they provide a glimpse into what we may get in the future in exchange for giving up data to our devices.
Google, which designs the Android software system running on a vast majority of the world’s phones, influences how other companies make their devices. Apple also plans to release an A.I. assistant capable of anticipating an iPhone user’s needs, but it has faced setbacks because of challenges developing the technology.
Here’s how my week testing Google’s A.I. phone went.
Meet ‘Magic Cue’
At the heart of the artificially intelligent Pixel is a new software tool called Magic Cue, which Google designed to anticipate your needs on the phone. To set it up, you grant the software access to apps including email, messages, notes, contacts, calendar and screenshots. In exchange, you can streamline tasks such as:
-
When you receive a text message asking for a friend’s phone number, the software will read that message and dig through your contacts to generate an automatic response with that person’s digits.
-
When you place a phone call to a business, such as an airline, the software will pull up relevant information from your flight itinerary so you can read it to the airline’s support representative.
-
If you create a calendar event for a hike on the weekend, the weather app will show the climate forecast for the location of the excursion.
Alex Moriconi, a Google spokesman, said the company took a “privacy-first approach” with Magic Cue. Users must opt in to sharing their data to use the tool. The feature runs on the Pixel phone’s hardware, he added, meaning that the data you share with it for help with various tasks is not scanned or processed on Google’s servers.
In my tests, Magic Cue was hit or miss.
In some instances, like receiving a text message asking for a friend’s phone number, Magic Cue pulled up the correct answer.
In other situations, it failed. When I called United Airlines to talk about my flight to New Orleans, the software tried loading a confirmation number from my itinerary. But Magic Cue showed a string of numbers — 138826 — that referred to the amount I had spent on the airfare: $1,388.26.
Here’s the bigger problem. I tapped a small arrow button to see where Magic Cue was finding these numbers. It pulled them from an email from Monarch, a personal finance service that I use to track my expenses, not from my United Airlines itinerary. In other words, Magic Cue scanned an email related to my finances.
It was a reminder of how invasive data sharing can be. When Magic Cue is set up, the software asks for access to the Gmail app, but it does not explicitly say it will scan all emails.
Google’s Mr. Moriconi said that Magic Cue searched through my emails to surface what it deemed relevant — looking, in this case, for emails with the word “United” — and that I could tap a thumbs-down button to tell the software that its suggestion was unhelpful.
I would be more forgiving if A.I. technology were still considered very new. It has been nearly three years since OpenAI debuted ChatGPT, upending the tech industry. Again and again, A.I. technology has been notorious for making these types of mistakes.
Camera Assistant
Google added software to the new Pixel to help people take and edit photos. One feature is a so-called Camera Coach that studies what you are photographing to give step-by-step advice to compose a better shot.
For example, when I was trying to photograph my daughter playing with bubbles, I pressed a button to summon the digital coach. It instructed me to zoom in closer to her and activate an effect to blur out the background. I was pleased with the results, but was certain I could have figured this out myself.
A feature that some people may find more useful is a photo-editing tool that lets them make complex changes to an image by simply typing in a prompt. In a photo I took of a beef entree at a Vietnamese restaurant, I asked the A.I. to remove a hand from the image. Within seconds, it removed the hand and replaced it with artificial imagery of the dinner table.
In a past column about A.I.-powered photo editing, I wrote about my ethical qualms with producing and sharing artificially aided photos because they contribute to the widespread issue of fakery on the web. But I get that people may want to remove distractions like photo bombers for their personal use.
The Big Picture
A.I. phones like the Pixel could redefine the social contract for what we get in exchange for what we give up in our personal technology.
In the past, the give and take felt reciprocal. To use a maps app for directions, you had to share your location, for example. But with an A.I. phone, we are sharing a whole lot about ourselves just to speed up a few tasks.
This future is far from inevitable. It is up to consumers to decide whether to use these tools and weigh the benefits. For me, an A.I. phone needs to be far more reliable and helpful before I would make the leap.
To Chris Gilliard, an independent privacy scholar, there is nothing magical about Magic Cue.
“I think the magic is surveillance, right?” he said.
Though companies including Apple and Google have promised to protect consumers’ data in their privacy policies, Mr. Gilliard added, “walls break down as we give systems complete access to our lives.”
Brian X. Chen is the lead consumer technology writer for The Times. He reviews products and writes Tech Fix, a column about the social implications of the tech we use.
The post Google Pixel 10 Pro Review: This A.I. Phone Can Save Time if You Surrender Your Data appeared first on New York Times.
