Not long ago, Google introduced Clips, an AI-powered camera trained to capture the best moments of your life. It has no LCD screen and there’s only a shutter button, which is completely optional. Google Clips uses artificial intelligence to recognize and save your “perfect moments” itself. But how is it possible? According to Google, it’s because they hired “a documentary filmmaker, a photojournalist, and a fine arts photographer” to help train the camera’s neural network.
In a recent blog post, UX Designer Josh Lovejoy explains the process of training the Google Clips’ AI to recognize a photo worth saving. As he explains, “if a human can’t perform the task, then neither can an AI.” Therefore, the team behind Google Clips focused on the human-centered design.
First of all, they focused on the real human needs. For example, they thought of new parents who watch their children grow up through their phone’s camera, trying not to miss anything. One of the applications of Google Clips is to enable the parents to be more involved, while the camera captures the right moments on their behalf.
Another way to train the AI is “guide the intelligence” and do it the right way. So, Google gathered footage from people on their team and tried answering the question “what makes a memorable moment.” But, the professionals in the creative field also helped them get the answer. They hired a documentary filmmaker, a photojournalist, and a fine arts photographer to help them develop the algorithm.
The team fed the algorithm a bunch of examples, both of good and bad photos and footage. However, Google admits that training AI isn’t a precise process. They didn’t lose sight of the human capability of using common sense and AI’s lack of capability to do so. As Lovejoy writes, “in the context of subjectivity and personalization, perfection simply isn’t possible.” But he adds that it really shouldn’t even be a goal. After all, if everything was perfect with AI, we would have hilarious photos like this.
Still, I find it interesting that Google hired professional photographers and a filmmaker to help them train AI. Even if the artificial intelligence is still imperfect, I’m sure they contributed to the higher accuracy of the algorithm.