Google's artificial intelligence engineers like to boast that the company, best known for its Web search innovations, has always been interested in AI and machine learning.
A bevy of new AI cloud services for businesses harnesses some code that was written nine years ago, Google VP and Engineering Fellow Fernando Pereira excitedly announced today at the company's office in San Francisco. That's the equivalent of a few millennium in Silicon Valley time.
But despite the long gestation period, Google's investment in AI follows a familiar pattern to the rest of its products: it can charge businesses to use some of it, but much more will eventually be given away to consumers for free.
So to help pique consumers' interest, Google also launched its A.I. Experiments website today, the brainchild of a handful of mostly millennial coders who work in its Creative Lab in New York City. The site offers five demonstrations intended to convince people that AI isn't just for robots and people with PhDs.
The first is Giorgio Cam, which simply asks users to "take a picture to make music with the computer." It works best on an Android device, but you can also use it on a laptop or desktop, too, assuming you haven't taped over your webcam. You take five snapshots, which are sent to the cloud for some image processing so the website can identify them and set them to music.
When I tried it on a few objects the engineers had on hand, it recognized a photo of an orange perfectly, but struggled to identify an iPhone depicted on the cover of a copy of Life magazine, labeling it a "gadget" with 76 percent confidence.
Next up was A.I. Duets, a delightful glimpse of a future in which computers join virtuosos on stage. You can play a few notes on a virtual keyboard and the computer will continue riffing, sort of like a call-and-response song you might have learned during summer camp or a grade school assembly. It's reminiscent of Project Magenta, Google's project to teach machines to play music. That project, along with the A.I. Experiments website, uses an open-source tool called TensorFlow that Google launched last year and is already catching on among third-party developers.
Finally, I tried the Bird Sounds demo, which was exactly what it sounds like, albeit with a searchable matrix that organizes the wavelengths of the calls of more than 14,000 birds. A search for "sparrow," for instance, reveals the quirky warbling of the house sparrow towards the far left of the matrix, while the song sparrow's rich melodies put it in a category all by itself at the far right.
Google's intention with its AI demos are clear: it wants to prove that artificial intelligence can be applied to pretty much any human pursuit, from bird watching to making music.
As Diane Greene, the head of Google Cloud, put it: machine learning is the "the value that's being added to almost any digital thing people do."
And if that's too theoretical for you, you're certain to have some boxes of ancient family photos in the basement that could easily benefit from some automatic touch-ups, which the new PhotoScan app for iOS and Android (also announced today) can help you with.