Month: August 2024

Magic Editor’s new tool helped us add the bike and car with nothing more than a text prompt. | Photo: The Verge

As it turns out, a rabbit wearing an AI-generated top hat was just the tip of the iceberg.
Google is the latest phone company this year to announce AI photo editing tools, following Samsung’s somewhat troubling, mostly delightful sketch-to-image feature and Apple’s much more seemingly tame Image Playground coming this fall. The Pixel 9’s answer is a new tool called “Reimagine,” and after using it for a week with a few of my colleagues, I’m more convinced than ever that none of us are ready for what’s coming.
Reimagine is a logical extension of last year’s Magic Editor tools, which let you select and erase parts of a scene or change the sky to look like a sunset. It was nothing shocking. But Reimagine doesn’t just take it a step further — it kicks the whole door down. You can select any nonhuman object or portion of a scene and type in a text prompt to generate something in that space. The results are often very convincing and even uncanny. The lighting, shadows, and perspective usually match the original photo. You can add fun stuff, sure, like wildflowers or rainbows or whatever. But that’s not the problem.
A couple of my colleagues helped me test the boundaries of Reimagine with their Pixel 9 and 9 Pro review units, and we got it to generate some very disturbing things. Some of this required some creative prompting to work around the obvious guardrails; if you choose your words carefully, you can get it to create a reasonably convincing body under a blood-stained sheet.

In our week of testing, we added car wrecks, smoking bombs in public places, sheets that appear to cover bloody corpses, and drug paraphernalia to images. That seems bad. As a reminder, this isn’t some piece of specialized software we went out of our way to use — it’s all built into a phone that my dad could walk into Verizon and buy.
When we asked Google for comment on the issue, company spokesperson Alex Moriconi responded with the following statement:
Pixel Studio and Magic Editor are helpful tools meant to unlock your creativity with text to image generation and advanced photo editing on Pixel 9 devices. We design our Generative AI tools to respect the intent of user prompts and that means they may create content that may offend when instructed by the user to do so. That said, it’s not anything goes. We have clear policies and Terms of Service on what kinds of content we allow and don’t allow, and build guardrails to prevent abuse. At times, some prompts can challenge these tools’ guardrails and we remain committed to continually enhancing and refining the safeguards we have in place.
To be sure, our creative prompting to work around filters is a clear violation of these policies. It’s also a violation of Safeway’s policies to ring up your organic peaches as conventionally grown at the self-checkout, not that I know anyone who would do that. And someone with the worst intentions isn’t concerned with Google’s terms and conditions, either. What’s most troubling about all of this is the lack of robust tools to identify this kind of content on the web. Our ability to make problematic images is running way ahead of our ability to identify them.
When you edit an image with Reimagine, there’s no watermark or any other obvious way to tell that the image is AI-generated — there’s just a tag in the metadata. That’s all well and good, but standard metadata is easily stripped from an image simply by taking a screenshot. Moriconi tells us that Google uses a more robust tagging system called SynthID for images created by Pixel Studio since they’re 100 percent synthetic. But images edited with Magic Editor don’t get those tags.

Photos: The Verge
To be sure, tampering with photos is nothing new. People have been adding weird and deceptive stuff to images since the beginning of photography. But the difference now is that it has never been this easy to add these things realistically to your photos. A year or two ago, adding a convincing car crash to an image would have taken time, expertise, an understanding of Photoshop layers, and access to expensive software. Those barriers are gone; all it now takes is a bit of text, a few moments, and a new Pixel phone.
It’s also never been easier to circulate misleading photos quickly. The tools to convincingly manipulate your photos exist right inside the same device you use to capture it and publish it for all the world to see. We uploaded one of our “Reimagined” images to an Instagram story as a test (and quickly took it down). Meta didn’t tag it automatically as AI-generated, and I’m sure nobody would have been the wiser if they’d seen it.
Who knows, maybe everyone will read and abide by Google’s AI policies and use Reimagine to put wildflowers and rainbows in their photos. That would be lovely! But just in case they don’t, it might be best to apply a little extra skepticism to photos you see online.

Magic Editor’s new tool helped us add the bike and car with nothing more than a text prompt. | Photo: The Verge

As it turns out, a rabbit wearing an AI-generated top hat was just the tip of the iceberg.

Google is the latest phone company this year to announce AI photo editing tools, following Samsung’s somewhat troubling, mostly delightful sketch-to-image feature and Apple’s much more seemingly tame Image Playground coming this fall. The Pixel 9’s answer is a new tool called “Reimagine,” and after using it for a week with a few of my colleagues, I’m more convinced than ever that none of us are ready for what’s coming.

Reimagine is a logical extension of last year’s Magic Editor tools, which let you select and erase parts of a scene or change the sky to look like a sunset. It was nothing shocking. But Reimagine doesn’t just take it a step further — it kicks the whole door down. You can select any nonhuman object or portion of a scene and type in a text prompt to generate something in that space. The results are often very convincing and even uncanny. The lighting, shadows, and perspective usually match the original photo. You can add fun stuff, sure, like wildflowers or rainbows or whatever. But that’s not the problem.

A couple of my colleagues helped me test the boundaries of Reimagine with their Pixel 9 and 9 Pro review units, and we got it to generate some very disturbing things. Some of this required some creative prompting to work around the obvious guardrails; if you choose your words carefully, you can get it to create a reasonably convincing body under a blood-stained sheet.

In our week of testing, we added car wrecks, smoking bombs in public places, sheets that appear to cover bloody corpses, and drug paraphernalia to images. That seems bad. As a reminder, this isn’t some piece of specialized software we went out of our way to use — it’s all built into a phone that my dad could walk into Verizon and buy.

When we asked Google for comment on the issue, company spokesperson Alex Moriconi responded with the following statement:

Pixel Studio and Magic Editor are helpful tools meant to unlock your creativity with text to image generation and advanced photo editing on Pixel 9 devices. We design our Generative AI tools to respect the intent of user prompts and that means they may create content that may offend when instructed by the user to do so. That said, it’s not anything goes. We have clear policies and Terms of Service on what kinds of content we allow and don’t allow, and build guardrails to prevent abuse. At times, some prompts can challenge these tools’ guardrails and we remain committed to continually enhancing and refining the safeguards we have in place.

To be sure, our creative prompting to work around filters is a clear violation of these policies. It’s also a violation of Safeway’s policies to ring up your organic peaches as conventionally grown at the self-checkout, not that I know anyone who would do that. And someone with the worst intentions isn’t concerned with Google’s terms and conditions, either. What’s most troubling about all of this is the lack of robust tools to identify this kind of content on the web. Our ability to make problematic images is running way ahead of our ability to identify them.

When you edit an image with Reimagine, there’s no watermark or any other obvious way to tell that the image is AI-generated — there’s just a tag in the metadata. That’s all well and good, but standard metadata is easily stripped from an image simply by taking a screenshot. Moriconi tells us that Google uses a more robust tagging system called SynthID for images created by Pixel Studio since they’re 100 percent synthetic. But images edited with Magic Editor don’t get those tags.

Photos: The Verge

To be sure, tampering with photos is nothing new. People have been adding weird and deceptive stuff to images since the beginning of photography. But the difference now is that it has never been this easy to add these things realistically to your photos. A year or two ago, adding a convincing car crash to an image would have taken time, expertise, an understanding of Photoshop layers, and access to expensive software. Those barriers are gone; all it now takes is a bit of text, a few moments, and a new Pixel phone.

It’s also never been easier to circulate misleading photos quickly. The tools to convincingly manipulate your photos exist right inside the same device you use to capture it and publish it for all the world to see. We uploaded one of our “Reimagined” images to an Instagram story as a test (and quickly took it down). Meta didn’t tag it automatically as AI-generated, and I’m sure nobody would have been the wiser if they’d seen it.

Who knows, maybe everyone will read and abide by Google’s AI policies and use Reimagine to put wildflowers and rainbows in their photos. That would be lovely! But just in case they don’t, it might be best to apply a little extra skepticism to photos you see online.

Read More 

The AI is inconsistent, but the hardware is oh so good. Google finally got the hardware right.
The Pixel 9 Pro, its bigger Pro XL sibling, and the standard Pixel 9 look and feel like the flagship phones Google has been trying to make since the Pixel 6 ushered in the visor camera bump era. They feel solid, the screens are bright, and the damn edges are finally flat. As far as I’m concerned, Google can hang up a “Mission Accomplished” banner.
The software is another thing. Some of it is promising, some of it seems like a party trick, and some of it is downright reckless. Google’s been rolling out generative AI features here and there over the past year, but this feels like the company’s first big swing at an AI phone. It’s kind of all over the place.
There’s a little sparkly AI icon in so many different corners of the UI, and these various assistants and systems don’t work well together yet. Do you want to have a conversation with AI? Or use AI to write an email? Or organize and refer to your screenshots with AI? Those features all exist on the Pixel 9 series, but they’re all in separate apps and interfaces. It’s starting to feel like I need AI to sort out all of the AI, and that’s not a great place to be. What’s worse is that they all work inconsistently, making it hard to rely on any of them. Thank God the hardware’s so good.

Google flattened out the edges of the phone and evened out the bezels around the screen. It’s an iPhone from the front, and I don’t think that’s a problem at all. The camera visor is a chunky pill that no longer connects to the phone’s side rails.
It still looks kind of weird, but it’s instantly recognizable as a Pixel. And despite this protrusion, the phone also sits steadily on a table when you tap the screen and doesn’t wobble back and forth — a problem that Samsung’s phones suffer from if you don’t put a case on them.
Along with a refreshed design, the Pixel 9 series gets a new Tensor G4 chip. Between the updated processor and a new vapor chamber, the Pixel no longer feels like it’s about to catch on fire when I use it as a Wi-Fi hotspot. Love it. I’m also a fan of the faster fingerprint scanner, which feels like the one Google should have been using all along.

Just look at those uniform bezels!

For the first time, Google is offering the Pro version in two sizes. They come with different-sized batteries, naturally, but both managed a full day of heavy use without needing a recharge. The Pixel 9 Pro is the size of the Pixel 8 (and the standard Pixel 9) with a 6.3-inch screen. The 9 Pro XL is equivalent to the Pixel 8 Pro in size with a 6.8-inch display.
The displays themselves are a bit brighter than the previous gen, going up to 2,000 nits for HDR content and up to 3,000 nits in peak brightness mode — the 8 Pro supported up to 1,600 nits and 2,400 nits, respectively. I can easily appreciate the difference in direct sunlight; it’s not Galaxy S24 Ultra good, but it’s a lot better.
But despite the difference in size, these two Pro 9 devices share the exact same camera hardware, including a 5x telephoto lens — something you don’t get on every “small” flagship phone. The main and telephoto cameras are unchanged from the 8 Pro, but the ultrawide has been updated with a faster lens that helps boost low-light performance.
There are a few AI features right inside the camera app, naturally. Unlike some of the other AI tools on these devices, these are pretty pedestrian. That includes Add Me, which lets you composite two photos into one group shot so that the person who took the first photo can get in the picture. The UI guides you through the process in which you take a photo and then swap with someone who was in the shot. You’ll see a ghostly overlay of the first image and some on-screen prompts to help you frame up the second photo properly, and afterward, you get one image with everyone included.

It works best when there’s plenty of light and your subjects stay in consistent poses between frames. When it’s good, it’s really good, and I’d have a hard time telling if anything was up if I didn’t know better. But even in the best examples, you can still zoom in and see some fuzzy edges around details like hair. I think I’d actually use this occasionally, not least of all because I hate asking a stranger to take my photo.
Video Boost, the AI tool that improves video, got a sizable update this time around, too. It processes faster once the file is uploaded, and there’s more detail in boosted Night Sight clips. The first time I tested Video Boost on the Pixel 8 Pro, it was a little underwhelming, but with these improvements, it’s a feature I’ve actually wanted to use more. It cleans up footage taken at higher zoom magnifications and smooths out transitions between lenses, so it’s a nice all-purpose tool if you’re doing something a little more technically challenging than just shooting a quick clip of your cat doing something funny.

AI tricks aren’t limited to the camera app — even if they’re some of my favorite use cases. As Google reminded us about a hundred times at its launch presentation, the Pixel 9 series is AI all the way down, from the Gemini Assistant — the default virtual assistant this time — to a daily AI summary in the revamped weather app.
AI is the thing in phones this year, and the Pixel 9 series represents our first look at some technologies that will likely trickle out across previous Pixel phones and parts of the Android ecosystem. A couple are exclusive to the Pixel 9 series, and Google is mostly vague about which features will be distributed to older phones. But altogether they’re the foundation of what Google wants us to think of as AI-first phones for the AI era.

They’re hit-and-miss, but one feature in particular is a little too good. That’s “reimagine,” a generative AI tool you’ll find in Magic Editor. Instead of just erasing or moving things around in your photo, you can select a part of your image and add something with a text prompt.
The results are uncanny — so good that they’re problematic. Without too much trouble, we had it add a range of nasty and extremely believable stuff to photos — everything from a cockroach on a plate of food to a snake in a flower display at Whole Foods.

Google’s examples of “reimagine” in use feature wildflowers and hot air balloons, which, sure. It can add those things to your photos. They usually look good and only sometimes look like a baked potato. But they’re only tagged as AI-generated by a line in the image metadata, which makes them really easy to pass off as real images.
Pixel Studio is less problematic. You use text prompts to dream up images in a handful of predetermined styles, including “3D cartoon” and “freestyle,” which is the more photorealistic option. My kid got a real kick out of making trucks of various shapes and sizes being operated by cats. If you ask it to generate an image with “poop” in it (toddlers think this is wildly funny), then you’ll get something more realistic than you probably wanted to see.
You can also play a fun game where you get it to generate IP that Google likely did not intend it to create. Here’s an incomplete list of the images I got it to make for me with these exact prompts:

Pikachu sticking a paper clip in an electrical outlet
Toad eating a banana
Thomas the Tank Engine chain smoking

It’s strange how easily you can make, like, PG-13-rated images, too. It faithfully generated a cartoon baby deer lighting a joint, and I don’t know, guys, maybe there’s a better use for all these supercomputers running AI. At the very least, it’s great if your idea of fun is responding to your spouse’s questions with obnoxious AI-generated art.
I had high hopes for Pixel Screenshots, a Pixel 9-series exclusive and potentially far more useful app. It’s a repository for all of your screenshots that uses AI to parse out information from them and saves it as metadata so you can search for it later — Airbnb door codes, Wi-Fi passwords, that kind of thing. It all stays on-device, so it’s relatively secure.
The thing is, it’s a whole separate app. You can’t ask Gemini to find your boarding zone; you have to open up the screenshots app and search. At that point, I’ll just open up the Delta Airlines app and look at my boarding pass. Besides, the Screenshots app told me I was in boarding group M3 — the pass it scanned clearly said group three.

The app automatically collects all your screenshots once you opt in.

I would legitimately use this for Pinterest-type stuff since I haven’t used Pinterest in about a decade.

And that’s the problem: it hallucinates and misinterprets. A lot of the metadata on my screenshots is right, but some of it is just off. I took a screenshot of a particularly gross “reimagine” creation when I prompted AI to fill a bowl with geoducks.
Reimagine made something I can best describe as a bowl of raw thumbs, which the Screenshots app labeled as “a green bowl of chicken” that might be “overcooked or undercooked.” Presumably scanning the text of the AI prompt I used for the photo, which appears in the corner of the screen, it states that the image is “from the Geoducks app, a food delivery service.” Makes me feel great about the future of AI trained on synthetic data.
I’m not ready to write off Screenshots just yet, though. It’s the kind of feature that makes sense when you use your phone for months or years, not weeks. It takes very little effort to use since screenshots are automatically saved there. And its best feature is that when you screenshot a page in Chrome, it’ll save the URL along with the image so you can get back to the page easily. When you think of Screenshots as a replacement for infinite Chrome tabs or a Pinterest board, it makes a lot more sense.
Gemini Assistant, which I’ve used on lots of other Android phones, is much more familiar and is now the default assistant. It can do a lot more basic assistant stuff than it could when it launched, but it still can’t play my dang Spotify playlists. The Pixel 9 Pro and Pro XL come with a free one-year trial of Gemini Advanced (a cool $20 per month after that!), which allows you to tap into newer language models and a brand-new feature: Gemini Live. It’s Google’s version of ChatGPT’s conversation mode, and incidentally, it feels a little like talking to a page of Google results.

An AI phone for a weird future.

The Pixel 9 Pros represent Google’s most advanced efforts in mobile AI, for better and worse. There’s a lot of promise in some of these tools, and at this point, I genuinely prefer asking Gemini some of my low-stakes questions than wading through Google Search. But we can’t keep ignoring the fact that AI just makes shit up sometimes, and it’s hard to trust a technology like that with the details of your day-to-day existence.
The feature pileup as Google rushes to ship new AI products is also getting a little confusing — not to mention that they’re all seemingly called some version of “Gemini” or “Gemma.” I can ask Gemini Assistant with the Workspace extension to check my inbox for important emails, but I can’t ask Gemini Live. I can also open Gemini inside the Gmail app to ask the same question and get a slightly different answer. I can take a screenshot of something on Amazon that I’m thinking about buying and save it to the Screenshots app, but I can’t automatically add a photo of something on a store shelf. It’s starting to feel a little like AI everything, everywhere, all at once.
It’s starting to feel a little like AI everything, everywhere, all at once
But the important thing is that behind all the flashy AI features, there’s a really good phone in the Pixel 9 Pro and the Pro XL. These are phones that I can finally hold up next to a Galaxy S24 Plus or an iPhone 15 Pro and think, yes, these are all top-of-the-line devices. These Pixels aren’t the budget-priced flagships that they used to be, and I think the higher prices are well justified by the hardware. Plus, when you’re getting seven years of OS updates, you can squeeze a whole lot of value out of your investment.
And even as a small-phone enthusiast, the 9 Pro feels like a reasonable size to me. It’s not small, but it’s not gargantuan, either, and I deeply appreciate not having to sacrifice camera features by choosing it over the big one. Pixel image quality remains reliable, and the battery will keep up to the end of the day. Whether we’re ready or not, a new era of AI phones and photos is here, and it’s messy as hell. But the hardware — if not my faith in an AI-everything future — is solid.
Photography by Allison Johnson / The Verge

The AI is inconsistent, but the hardware is oh so good.

Google finally got the hardware right.

The Pixel 9 Pro, its bigger Pro XL sibling, and the standard Pixel 9 look and feel like the flagship phones Google has been trying to make since the Pixel 6 ushered in the visor camera bump era. They feel solid, the screens are bright, and the damn edges are finally flat. As far as I’m concerned, Google can hang up a “Mission Accomplished” banner.

The software is another thing. Some of it is promising, some of it seems like a party trick, and some of it is downright reckless. Google’s been rolling out generative AI features here and there over the past year, but this feels like the company’s first big swing at an AI phone. It’s kind of all over the place.

There’s a little sparkly AI icon in so many different corners of the UI, and these various assistants and systems don’t work well together yet. Do you want to have a conversation with AI? Or use AI to write an email? Or organize and refer to your screenshots with AI? Those features all exist on the Pixel 9 series, but they’re all in separate apps and interfaces. It’s starting to feel like I need AI to sort out all of the AI, and that’s not a great place to be. What’s worse is that they all work inconsistently, making it hard to rely on any of them. Thank God the hardware’s so good.

Google flattened out the edges of the phone and evened out the bezels around the screen. It’s an iPhone from the front, and I don’t think that’s a problem at all. The camera visor is a chunky pill that no longer connects to the phone’s side rails.

It still looks kind of weird, but it’s instantly recognizable as a Pixel. And despite this protrusion, the phone also sits steadily on a table when you tap the screen and doesn’t wobble back and forth — a problem that Samsung’s phones suffer from if you don’t put a case on them.

Along with a refreshed design, the Pixel 9 series gets a new Tensor G4 chip. Between the updated processor and a new vapor chamber, the Pixel no longer feels like it’s about to catch on fire when I use it as a Wi-Fi hotspot. Love it. I’m also a fan of the faster fingerprint scanner, which feels like the one Google should have been using all along.

For the first time, Google is offering the Pro version in two sizes. They come with different-sized batteries, naturally, but both managed a full day of heavy use without needing a recharge. The Pixel 9 Pro is the size of the Pixel 8 (and the standard Pixel 9) with a 6.3-inch screen. The 9 Pro XL is equivalent to the Pixel 8 Pro in size with a 6.8-inch display.

The displays themselves are a bit brighter than the previous gen, going up to 2,000 nits for HDR content and up to 3,000 nits in peak brightness mode — the 8 Pro supported up to 1,600 nits and 2,400 nits, respectively. I can easily appreciate the difference in direct sunlight; it’s not Galaxy S24 Ultra good, but it’s a lot better.

But despite the difference in size, these two Pro 9 devices share the exact same camera hardware, including a 5x telephoto lens — something you don’t get on every “small” flagship phone. The main and telephoto cameras are unchanged from the 8 Pro, but the ultrawide has been updated with a faster lens that helps boost low-light performance.

There are a few AI features right inside the camera app, naturally. Unlike some of the other AI tools on these devices, these are pretty pedestrian. That includes Add Me, which lets you composite two photos into one group shot so that the person who took the first photo can get in the picture. The UI guides you through the process in which you take a photo and then swap with someone who was in the shot. You’ll see a ghostly overlay of the first image and some on-screen prompts to help you frame up the second photo properly, and afterward, you get one image with everyone included.

It works best when there’s plenty of light and your subjects stay in consistent poses between frames. When it’s good, it’s really good, and I’d have a hard time telling if anything was up if I didn’t know better. But even in the best examples, you can still zoom in and see some fuzzy edges around details like hair. I think I’d actually use this occasionally, not least of all because I hate asking a stranger to take my photo.

Video Boost, the AI tool that improves video, got a sizable update this time around, too. It processes faster once the file is uploaded, and there’s more detail in boosted Night Sight clips. The first time I tested Video Boost on the Pixel 8 Pro, it was a little underwhelming, but with these improvements, it’s a feature I’ve actually wanted to use more. It cleans up footage taken at higher zoom magnifications and smooths out transitions between lenses, so it’s a nice all-purpose tool if you’re doing something a little more technically challenging than just shooting a quick clip of your cat doing something funny.

AI tricks aren’t limited to the camera app — even if they’re some of my favorite use cases. As Google reminded us about a hundred times at its launch presentation, the Pixel 9 series is AI all the way down, from the Gemini Assistant — the default virtual assistant this time — to a daily AI summary in the revamped weather app.

AI is the thing in phones this year, and the Pixel 9 series represents our first look at some technologies that will likely trickle out across previous Pixel phones and parts of the Android ecosystem. A couple are exclusive to the Pixel 9 series, and Google is mostly vague about which features will be distributed to older phones. But altogether they’re the foundation of what Google wants us to think of as AI-first phones for the AI era.

They’re hit-and-miss, but one feature in particular is a little too good. That’s “reimagine,” a generative AI tool you’ll find in Magic Editor. Instead of just erasing or moving things around in your photo, you can select a part of your image and add something with a text prompt.

The results are uncanny — so good that they’re problematic. Without too much trouble, we had it add a range of nasty and extremely believable stuff to photos — everything from a cockroach on a plate of food to a snake in a flower display at Whole Foods.

Google’s examples of “reimagine” in use feature wildflowers and hot air balloons, which, sure. It can add those things to your photos. They usually look good and only sometimes look like a baked potato. But they’re only tagged as AI-generated by a line in the image metadata, which makes them really easy to pass off as real images.

Pixel Studio is less problematic. You use text prompts to dream up images in a handful of predetermined styles, including “3D cartoon” and “freestyle,” which is the more photorealistic option. My kid got a real kick out of making trucks of various shapes and sizes being operated by cats. If you ask it to generate an image with “poop” in it (toddlers think this is wildly funny), then you’ll get something more realistic than you probably wanted to see.

You can also play a fun game where you get it to generate IP that Google likely did not intend it to create. Here’s an incomplete list of the images I got it to make for me with these exact prompts:

Pikachu sticking a paper clip in an electrical outlet
Toad eating a banana
Thomas the Tank Engine chain smoking

It’s strange how easily you can make, like, PG-13-rated images, too. It faithfully generated a cartoon baby deer lighting a joint, and I don’t know, guys, maybe there’s a better use for all these supercomputers running AI. At the very least, it’s great if your idea of fun is responding to your spouse’s questions with obnoxious AI-generated art.

I had high hopes for Pixel Screenshots, a Pixel 9-series exclusive and potentially far more useful app. It’s a repository for all of your screenshots that uses AI to parse out information from them and saves it as metadata so you can search for it later — Airbnb door codes, Wi-Fi passwords, that kind of thing. It all stays on-device, so it’s relatively secure.

The thing is, it’s a whole separate app. You can’t ask Gemini to find your boarding zone; you have to open up the screenshots app and search. At that point, I’ll just open up the Delta Airlines app and look at my boarding pass. Besides, the Screenshots app told me I was in boarding group M3 — the pass it scanned clearly said group three.

And that’s the problem: it hallucinates and misinterprets. A lot of the metadata on my screenshots is right, but some of it is just off. I took a screenshot of a particularly gross “reimagine” creation when I prompted AI to fill a bowl with geoducks.

Reimagine made something I can best describe as a bowl of raw thumbs, which the Screenshots app labeled as “a green bowl of chicken” that might be “overcooked or undercooked.” Presumably scanning the text of the AI prompt I used for the photo, which appears in the corner of the screen, it states that the image is “from the Geoducks app, a food delivery service.” Makes me feel great about the future of AI trained on synthetic data.

I’m not ready to write off Screenshots just yet, though. It’s the kind of feature that makes sense when you use your phone for months or years, not weeks. It takes very little effort to use since screenshots are automatically saved there. And its best feature is that when you screenshot a page in Chrome, it’ll save the URL along with the image so you can get back to the page easily. When you think of Screenshots as a replacement for infinite Chrome tabs or a Pinterest board, it makes a lot more sense.

Gemini Assistant, which I’ve used on lots of other Android phones, is much more familiar and is now the default assistant. It can do a lot more basic assistant stuff than it could when it launched, but it still can’t play my dang Spotify playlists. The Pixel 9 Pro and Pro XL come with a free one-year trial of Gemini Advanced (a cool $20 per month after that!), which allows you to tap into newer language models and a brand-new feature: Gemini Live. It’s Google’s version of ChatGPT’s conversation mode, and incidentally, it feels a little like talking to a page of Google results.

The Pixel 9 Pros represent Google’s most advanced efforts in mobile AI, for better and worse. There’s a lot of promise in some of these tools, and at this point, I genuinely prefer asking Gemini some of my low-stakes questions than wading through Google Search. But we can’t keep ignoring the fact that AI just makes shit up sometimes, and it’s hard to trust a technology like that with the details of your day-to-day existence.

The feature pileup as Google rushes to ship new AI products is also getting a little confusing — not to mention that they’re all seemingly called some version of “Gemini” or “Gemma.” I can ask Gemini Assistant with the Workspace extension to check my inbox for important emails, but I can’t ask Gemini Live. I can also open Gemini inside the Gmail app to ask the same question and get a slightly different answer. I can take a screenshot of something on Amazon that I’m thinking about buying and save it to the Screenshots app, but I can’t automatically add a photo of something on a store shelf. It’s starting to feel a little like AI everything, everywhere, all at once.

But the important thing is that behind all the flashy AI features, there’s a really good phone in the Pixel 9 Pro and the Pro XL. These are phones that I can finally hold up next to a Galaxy S24 Plus or an iPhone 15 Pro and think, yes, these are all top-of-the-line devices. These Pixels aren’t the budget-priced flagships that they used to be, and I think the higher prices are well justified by the hardware. Plus, when you’re getting seven years of OS updates, you can squeeze a whole lot of value out of your investment.

And even as a small-phone enthusiast, the 9 Pro feels like a reasonable size to me. It’s not small, but it’s not gargantuan, either, and I deeply appreciate not having to sacrifice camera features by choosing it over the big one. Pixel image quality remains reliable, and the battery will keep up to the end of the day. Whether we’re ready or not, a new era of AI phones and photos is here, and it’s messy as hell. But the hardware — if not my faith in an AI-everything future — is solid.

Photography by Allison Johnson / The Verge

Read More