Is Samsung’s AI Image Generator Too Good to Be True?

Blurred Bees and Fake Pirates: How AI is Changing Photography and Our Reality

The pirate ship in Elliott Bay was one thing, but it was a blurry bee that sent me over the edge. Samsung, in its eagerness to tout the AI prowess of its new phones, particularly the Galaxy Fold 6, has included a feature called "sketch to image". Draw a rough sketch on a photo or a blank note, and the phone uses generative AI to transform your doodle into a convincing, detailed image. Initially, I dismissed it as just another AI gimmick, but the results are far more impactful than I initially anticipated – and leave me with a growing sense of unease.

Using sketch to image within a note app is relatively harmless. Draw something, highlight it, and choose from styles like "3D cartoon" or "illustration" to enhance your doodle. The image is sent to the cloud, processed, and presented as a few options. The results are often cute and fun; I drew with my two-year-old, creating goofy dump trucks and school buses. Occasionally, you get a teddy bear with too many arms, but nothing truly concerning.

However, things get truly unsettling when applying sketch to image to photographs. I, a terrible artist, discovered that even my rudimentary sketches were transformed into photorealistic images. The AI-generated elements seamlessly blend into the photos, scaled and matched to the surroundings, making them exceptionally difficult to spot as fakes.

This is where the blurry bee incident comes into play. I captured a photo of flowers near a dock in Seattle, with the focus set on the distance, leaving the foreground slightly blurred. I drew a ridiculously crude bee on one of the flowers, expecting the AI to insert a clear image of a bee, thus revealing its artificial origin. Wrong! The AI bee was just as blurred as the flower it was landing on. Without knowing its backstory, I’d never suspect it wasn’t real. It looked like a natural, perfectly timed capture, the result of skill and patience. In fact, I’m not sure I’d even spot the "AI-generated content" watermark in the image’s corner.

"It’s convincing at a glance – but if you look for more than a second, you’ll notice that something is off."

I spent a week experimenting with sketch to image, and the results are inconsistent. Sometimes the AI-generated images have telltale signs, including alien-looking text or strange textures that look unnatural. The images are believable at a glance, but with a closer look, imperfections become apparent. Other times, the content itself gives away the manipulation, such as my photo of a giant orange cat in a Seattle intersection or a massive pirate ship docked in Elliott Bay. Even these obviously fabricated scenarios appear incredibly realistic.

While large objects generally look fake, it’s easy to add a car to a busy road or a sailboat in the distance, leaving most viewers none the wiser. Beyond the watermark, which can be easily cropped out, there is almost no way to detect that the image has been manipulated. This is unsettling, even disconcerting.

"Out-of-focus bee isn’t going to undo the fabric of our society."

This technology is currently optional and may not be widely used, but its implications are significant. While a blurry bee won’t shatter society, we’re reaching a point where AI is blurring the lines between reality and fiction more than ever. The ability to manipulate images within the very device used to capture and share them is a game-changer. Generative AI tools are rapidly evolving, surpassing our collective understanding of what might be genuine and what might be fabricated.

The impact transcends photography and extends to our overall perception of truth. This is particularly concerning given the current political climate. Fake news and misinformation are already widespread, and AI tools like sketch to image have the potential to exacerbate these problems.

For me, the most unsettling aspect is introducing this technology to my toddler. He’ll grow up knowing a simple button can transform a rough sketch into a polished image. He’ll witness the effortless manipulation of images, adding a train to a photo of train tracks with minimal effort. This drastically alters the understanding of artistic creation I had as a child and may lead to a different perspective on the meaning of visual content.

Despite the looming concerns, sketch to image is undeniably entertaining. There’s an endearing, almost comical quality to the AI’s output. I tried adding a green monster poking its head out of Puget Sound, only to see a giant green polar bear with rippling muscles appear on the shore. A stick figure sketch transformed into a life-size stick figure with a shadow beneath it. The absurdity of these creations is somewhat disarming.

"Is the definition of photography changing in front of our eyes? Is our understanding of truth in images transforming at an unbelievably precarious time for our democracy?"

While the implications are serious, I can’t deny the fun I’ve had with sketch to image. I took a picture of a rabbit and the AI put a tiny top hat on its head. What a time to be alive!

Sketch to image is currently available on the Galaxy Z Fold 6 and Z Flip 6. Samsung hasn’t publicly announced whether it will be rolled out to other Galaxy phones, but given their history of aggressively expanding AI features to previous model generations, it seems highly probable. Samsung has also committed to bringing AI features to 200 million phones this year alone. If the blurry bee is any indication, we’re likely in for a wild ride as this technology becomes more accessible. It’s a ride that will inevitably force us to confront the implications of AI on our perception of reality and truth. Perhaps that blurry bee wasn’t just a bug, but a harbinger of a new, uncertain future.

Article Reference