Around this time last year Google and Raspberry Pi did something rather interesting. Together they packaged machine learning—the ability for your Raspberry Pi to think and reason—as a kit, and made it available free on the cover of a magazine. The magazine sold out in hours.
The AIY Projects Voice Kit lets you add voice interaction to your Raspberry Pi project, and if you didn’t manage to get your hands on one of the kits that came with issue 57 of the MagPi, it was a long wait until there was news that more kits were on the way and were available for pre-order.
The new Voice Kits arrived on shelves in October, and close on their heels came the news of the second AIY Projects Kit, the Vision Kit. However only a limited number of these kits made it to shelves before Christmas, so not many people have managed to go hands-on with the Vision Kit hardware yet.
That is, until now.
As Google has just announced updates to the Voice and Vision Kits, and the newly updated hardware will be on shelves in Target stores across the United States later today.
The most obvious change to both kits is what’s inside the boxes. The Voice and Vision Kits now come with a Raspberry Pi, while the Vision Kit includes a Raspberry Pi Camera Module as well. Which means that the two kits contain everything you need to get started, except a USB power supply.
After a rough initial launch for the Vision Kit, Google has obviously been listening to feedback, and the AIY Projects site has also been updated with much improved build documentation for both it, and the Voice Kit. Alongside that, Google has released a companion Android application to help simplify getting the project kits connected to your wireless network.The Voice Kit
The biggest hardware changes are to the Voice Kit. The original kit was built around a voice HAT board along with a microphone add-on daughter board.
A HAT has been designed to be used with a ‘standard’ Raspberry Pi, however opening the new Voice Kit box shows us that things have changed. Instead of a full HAT the new Vision Kit is a pHAT — more commonly know, thanks to Lady Ada — as a “Bonnet.”
Also included in the box is a Raspberry Pi Zero WH, the familiar 3W speaker from the previous incarnation of the kit, an arcade button, and a micro SD card pre-provisioned with the AIY Projects disk image.
Finally, there’s a cardboard case for the project build, which after Google Cardboard, has become almost synonymous with Google’s own prototyping efforts.
While switching away from a full-size Raspberry Pi HAT to a smaller Bonnet form factor hasn’t changed the size of the cardboard case all that much, as the size of the Voice Kit box is determined by the size of the speaker that comes with the kit, but it does mean that projects built around the kit can be smaller.
“…with the Pi Zero form factor, we had the opportunity to shrink down our accessory boards which is compelling to makers exploring innovative smart device ideas, not to mention lowers costs. We’re excited to bundle the Raspberry Pi components in our updated kits and look forward to seeing what creative ideas makers share out soon!” — Billy Rutledge, Director of AIY Projects at Google
The new Bonnet is built around a Microchip SAMD09U microcontroller, which is being supported by a Realtek ALC5645 audio chip. The smallest member of the SAM D family, the SAMD09U is a low-power Arm Cortex-M0+ microcontroller — the little brother of the faster SAM D21 used in the Arduino Zero, MKR1000, and MKRZero. The Microchip MCU handles the buttons, LEDs, and buzzers in the kit and provides the processing behind the analog I/O headers on the other side of the board — the chip has a 10 channel 12-bit ADC onboard.
Rather than having a separate daughter board, the new Bonnet also incorporates the two MEMS microphones—positioned at either end of the Bonnet—directly on to the board, with the two InvenSense ICS-43434 microphones in the original kit having been replaced with two Knowles SPH1642HT5H-1 microphones.The Vision Kit
Already based around a Bonnet, rather than a full HAT, there have been far fewer hardware changes to the updated Vision Kit.
The kit includes a big arcade button, a privacy LED, a piezo buzzer, some spacers, a tripod mounting nut, and a cardboard case. Gone is the lens assembly, present in the original kit, presumably no longer needed.
Labelling on the ribbon cables, both the longer cable connecting the Raspberry Pi Camera Module to the AIY Projects Vision Bonnet, and the short cable that connects the Bonnet to the Raspberry Pi Zero has been updated. After the problems with the original kit, the labelling — showing which direction you should plug the cables in — is now printed directly onto the ribbon cables which should help reduce confusion.
Yet unlike the Voice Kit, the Vision Kit is designed to run the all the machine learning locally — on the device — rather than talk to the cloud. Whereas it was possible to run TensorFlow locally on your Raspberry Pi with the Voice Kit, the Vision Kit is designed from the ground up to run do all its image processing locally. In theory at least, depending on your project, you can use the kit without any network connection at all.Artificial Intelligence in STEM
It’s interesting to see the focus in today’s announcement shift slightly towards education. There seems to have been a strong demand from teachers for the kits and Google have decided to address that.
“We’re taking the first of many steps to help educators integrate AIY into STEM lesson plans and help prepare students for the challenges of the future… The new kits make getting started a little easier with clearer instructions, a new app and all the parts in one box.” — Billy Rutledge, Director of AIY Projects at Google
Putting everything you need to get started into one box is actually a bigger help than you might expect for educators, but it’s going to be interesting to see what resources Google will be offering to help use the kits inside existing STEM lesson plans.Artificial Intelligence on the Edge
For me however, the really interesting thing about the Google AIY Project Kits is what it tells us about how Google views the way machine learning should be used. The kits give us the ability to run trained networks “at the edge” nearer the raw data — without the cloud support that seems necessary to almost every task these days, and even without a network connection — this could help reduce barriers to developing, tuning, and deploying machine learning applications out in the world.
It could potentially help make “smart objects” actually smart, rather than just network connected thin clients for machine learning algorithms running in remote data centres. It could, in fact, be the start of a sea change about how we think about machine learning and how the Internet of Things might be built. Because now there is — at least the potential — to allow us to put the smarts on the smart device, rather than in the cloud.
I think all of this is a sign of the next big swing for computing, back, away from the cloud. Towards building an “edge ecosystem” for artificial intelligence and, interestingly perhaps, the refreshed site has a Model Exchange where Google is showcasing a collection of neural network models designed to work with the AIY kits. Almost as if they’re thinking along the same lines?Kit Availability
The updated AIY Projects Kits are available now at Target, both in store and online. The Voice Kit costs $49.95, while the Vision Kit costs $89.95, and both kits now come with everything you need to get started—including a Raspberry Pi Zero WH, and for the Vision Kit, the Raspberry Pi Camera Module. Unfortunately the updated kits aren’t yet available outside the United States, but Google tells us they’re working on that, and that you should sign up to their mailing list to be notified when they do become available.