OpenAI, the company behind ChatGPT, might have picked up most attention this week with a former employee's vague warnings about the chances of AI wiping out humanity.
However, despite the calls to slow down development, it hasn't halted work on its camera tools, acquiring a photo-processing company that might even go toward the company's long-anticipated AI gadget.
Glass Imaging is an AI tool that helps improve the quality of images taken with the small image sensors and lenses on camera phones, and create results more like those on devices with full-frame sensors, like DSLRs or mirrorless cameras.
There is an extensive gallery of examples on the company site, largely shot on the iPhone 15 and iPhone 16 cameras in imperfect light. I was especially interested to see that the AI wasn't above altering the pose of the subject; look at the boy in the example, whose head has turned to face the camera.
Glass A.I. processing sample
An original image shot on an iPhone and the result of Glass A.I. processing
Glass Imaging was a venture-backed company that claimed it could "Unlock the full potential of imaging systems," and the technology used is called 'Glass AI'. It claims to be a "neural ISP" – which presumably means image signal processing.
Image signal processing is the next step on a digital camera's circuits after capturing the light with the sensor. The signal, perhaps from a CMOS sensor, passes via analog-to-digital conversion to an a ISP chip (or a combined system-on-a-chip).
This is where the color science and processing takes place that results in an image file that can be handed off to the storage (such as a memory card).
I understand that adding better technology (under the umbrella of AI) to the ISP can result in theoretically better images than otherwise expected from the same chip.
Essentially, this is what Apple calls "computational photography" – and all camera companies, to a greater or lesser extent, attempt already. Indeed the Honor 600 phone has already incorporated this very tech. Glass Imaging is just openly promising to use the latest thing in tech – AI – and suggesting it'll do it better.
Where that looks a bit strange is in the examples that change the pose rather than just the texture and quality of the image. Now that is the kind of thing that might result from generative AI, but (at least to my way of thinking) isn't really just improving the lens and sensor to their theoretical maximum!
Eliminating grain and making it look like you used 'better glass'? That's great. But changing the pose? That is a different matter.
Not that OpenAI has specified its plans for the software, but the company has been investing a great deal in bringing its own AI-driven device to market – and many are suggesting there is a connection.
This spend is pocket change compared to the $6.5 billion (around £4.9 billion / AU$9.1 billion) OpenAI spent on Io Products, the company Jony Ive founded. Ive is formerly Apple's chief designer, with the iPhone in his list of credits, and has already been connected to smart devices with cameras.
OpenAI is also still thought to be preparing a stock offering, though that is likely to wait until 2027 because of the AI safety fears (or, more precisely, the fear that the IPO won't work as well as hoped in the present climate).
You might also like…
If you can't wait, why not just check out the best iPhone for photography?
