What Is a Camera? Exploring the Evolution with Google’s Pixel 10 Pro

At The Verge, we often ponder the question, “What is a photo?” as we dissect the line between the real and unreal—especially concerning images captured by our smartphones. Now, let’s expand this inquiry with a new question: what is a camera? The answer has become increasingly complex with the launch of the Google Pixel 10 Pro and Pro XL, as generative AI technology is now deeply integrated into the camera itself.

Understanding Pro Res Zoom

One groundbreaking feature of this new camera system is Pro Res Zoom. This should not be confused with Apple’s ProRes video format or Google’s previous Super Res Zoom. Pro Res Zoom is activated when you zoom beyond 30x, reaching up to a staggering 100x digital zoom. In typical scenarios, cameras struggle to provide clarity at such high zoom levels, often resulting in grainy, unusable images. The camera employs sophisticated algorithms to fill in the gaps created by upscaling these minute portions of a photo, a process that has historically yielded disappointing results, particularly at higher magnifications.

What sets Pro Res Zoom apart is its ambition to provide a usable image where other cameras have faltered. This is achieved through a latent diffusion model, a technique described by Google’s Pixel camera product manager, Isaac Reynolds. While he regards it as an evolution rather than a brand-new technology, it stands out for its ability to minimize common artifacts associated with photo processing.

The Technology Behind the Scenes

The latent diffusion model is described as “pretty good at killing the artifacts” common in previous algorithms. In demos, Pro Res Zoom exhibited the capability to significantly enhance even heavily zoomed-in images, delivering impressive results that one might not expect from such magnification levels.

When the feature was initially being developed, the processing time was about a minute. Remarkably, the team reduced this down to just four or five seconds. Once the image is processed, the original version is preserved alongside the enhanced one, allowing users to compare the two.

Designing with Ethics in Mind

An essential aspect of Pro Res Zoom is its built-in constraints: it doesn’t operate on images that contain people. When a human figure is detected, the algorithm efficiently enhances the background while leaving the individual untouched, a necessary measure to eliminate potential ethical concerns regarding AI manipulation of facial features. This acknowledgment of privacy reflects a broader cultural consideration in the design of these AI features.

Beyond this, Google has integrated responsible tagging through C2PA content credentials, marking photos edited with AI tools. This tagging system applies not only to images enhanced via Pro Res Zoom but also labels any photo taken with the Pixel 10 camera, including info on whether AI played a role in capturing or editing the image.

Addressing the “Implied Truth Effect”

The rationale behind tagging is rooted in combating the “implied truth effect.” If only AI-generated images receive labels, photos without such tags could misleadingly convey authenticity. Given the pervasive availability of image manipulation and generation tools, it’s critical to understand the true origins of what we see. The immutable nature of C2PA credentials provides a layer of certainty that can positively identify an image as camera-generated.

Acknowledging Potential Misuse

With the rise of such advanced imaging technology, there are concerns. Misapplications of AI could foster distrust in visual media, affecting societal institutions and individual relationships. Isaac Reynolds acknowledges that we are in a period of education concerning the implications of these technologies, which could unfortunately create a gap between our current understanding and the potential realities of the future.

The Philosophical Dilemma of Photographic Integrity

As we delve deeper into this conversation about technology’s impact on photography, a fundamental question arises: What is an image captured using Pro Res Zoom? Is it merely a snapshot of a moment, or is it the camera’s algorithm offering its best guess at reality? When capturing something iconic, like the Statue of Liberty, is it genuinely a photo taken by the user or a product of algorithmic manipulation?

Reynolds argues that Pro Res Zoom is “tuned very carefully to just be a picture." In other words, the essence of what we anticipate from photos remains intact, according to Google’s design philosophy. While these advancements inevitably raise questions about authenticity in visual media, they also illustrate the exciting possibilities that lie ahead in photography and technology.

As we ponder whether a camera that utilizes AI still qualifies as a traditional camera, it’s clear that we’re on the threshold of something new—an opportunity to redefine what it means to capture the world around us.