Google’s Photos Assistant is an amazing tool, most of the time. For starters, it’s not too obtrusive, offering users a different take on their photos and videos when it thinks it has spotted a special event or image that is worth a little extra effort. But there’s also a problem.
Most of the offers are in the form of stylised photos:
The moments are well chosen, and well executed to the point where Photos Assistant is welcome on my phone any time.
Until yesterday, when it offered me a “Then and Now” photo for the first time:
Obviously, that’s a collage of two pics of two of my kids at different moments in time. Which is nice, because it’s something I’d never think of searching around for.
But I spent too much time thinking about what it took for Google’s AI bot to find it, and I just can’t get comfortable with it.
The pictures were taken about 10 months apart. The one on the right was taken just hours before Photos Assistant offered me the collage.
So after the one on the right was taken, Google’s AI immediately scanned eight months of photos (i.e. hundreds) and identified both my children in a similar pose.
That’s impressive, given they’re not exactly staring down the lens. Although having covered AI developments for years, I’m also aware it’s probably a fair way down the scale of achievements compared to say, where SenseTime is up to in China right now.
But, whatever. It’s still an AI program that can almost instantly identify facial features from hundreds of images, and it’s right in the palm of my, and millions of other hands right now.
And that’s a step too far for me right now.