Samsung’s new image-generating AI tool is almost too realistic

Samsung’s latest innovation, the AI-powered "sketch to image" tool, is making waves with its impressive and slightly unsettling capabilities. Available on the new Galaxy Z Fold 6 and Z Flip 6, this tool allows users to draw simple sketches on photos or note pages, and then transform those doodles into detailed images using generative AI. When Samsung first introduced the tool at its Unpacked event, it seemed like another run-of-the-mill AI feature. However, after some hands-on testing, it becomes clear just how powerful and transformative this technology can be.

The "sketch to image" tool works in a fairly straightforward manner. Users draw a sketch, highlight it, and then select from various styles such as “3D cartoon” or “illustration.” The sketch is sent to the cloud, and within moments, users receive several AI-generated options to choose from. The results are often fun and whimsical, making it an enjoyable feature to use. For example, drawing with my two-year-old produced goofy, yet endearing images of dump trucks and school buses. Occasionally, the AI-generated images can be a bit off, like a teddy bear with too many arms, but overall, the tool’s outputs are charming and entertaining.

However, the real intrigue begins when using this tool on photographs. Even as a self-proclaimed terrible artist, my rudimentary sketches were transformed into photorealistic images seamlessly integrated into the original photos. The AI adjusts the added elements to match the surrounding environment, making them difficult to distinguish as fakes. This brings us to the curious case of the bee. I took a photo with flowers in the foreground, which were slightly blurred due to the camera’s focus on the distant background. Drawing a crude bee sketch on one of these flowers, I expected the AI to insert a sharply focused bee, revealing it as an artificial addition. Instead, the AI produced a blurred bee that perfectly matched the blurred flowers, creating an almost indistinguishable fake.

The implications of this are significant. If I hadn’t known the bee was AI-generated, I would have assumed it was a lucky shot, captured by a patient photographer. The AI’s ability to create such convincing images raises questions about the authenticity of what we see online. Even though there are sometimes telltale signs of generative AI, such as strange textures or nonsensical text, the technology is advancing rapidly. Adding subtle elements to photos, like an extra car on a busy street or a distant sailboat, can easily go unnoticed, especially when the AI watermark can be cropped out.

While this may sound alarming, it's important to keep things in perspective. The use of the "sketch to image" tool is entirely optional, and many people might never discover it in the gallery app. However, it does highlight how AI capabilities are outpacing our collective understanding of image authenticity. The convenience of having such a powerful tool on the same device used to capture and share photos blurs the lines between reality and digital manipulation.

Introducing this technology to younger generations also presents a unique challenge. My toddler, for instance, will grow up in a world where transforming a simple sketch into a polished image is a matter of pressing a button. This is a stark contrast to how artistic creation was perceived when I was a child. The ease of enhancing a photo with a simple addition, like a train on empty tracks, could alter how we value and interpret artistic endeavors.

Despite these concerns, the "sketch to image" tool is undeniably fun to use. There’s an endearing quality to the AI’s interpretations, such as when a green monster sketch turned into a massive, muscular green polar bear, or when a stick figure doodle became a life-size stick figure with a shadow. The playful nature of these creations showcases the lighthearted potential of the technology.

As we witness these advancements, we must also consider the broader implications. The definition of photography is evolving before our eyes, and our understanding of visual truth is being challenged, especially in an era where misinformation can have significant consequences. Yet, the ability to add a tiny top hat to a rabbit in a photo with a few clicks highlights the whimsical side of these developments.

Samsung has not yet announced whether the "sketch to image" feature will be available on other Galaxy phones, but given their history of rolling out AI features to older models, it seems likely. Samsung aims to introduce AI capabilities to 200 million phones this year alone, suggesting that these tools will become more widespread. As we navigate this increasingly AI-driven landscape, it’s clear that things are set to get a little weird, and perhaps, a lot more interesting.