SoylentNews is people
SoylentNews
Google's new DIY AI kits could help shape the future
Google just announced two new "AIY" (it's like DIY, but for artificial intelligence) kits that build upon the ideas the company set forth with its first-generation kits. This time around, however, the new kits ship with everything a student might need to build AI solutions, including a Raspberry Pi Zero WH board.
"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 by launching a new version of our AIY kits," Billy Rutledge, Director of AIY Projects at Google, wrote in a blog post. "The Voice Kit lets you build a voice controlled speaker, while the Vision Kit lets you build a camera that learns to recognize people and objects. The new kits make getting started a little easier with clearer instructions, a new app and all the parts in one box."
Also at The Verge.
(Score: 5, Informative) by leftover on Tuesday April 17 2018, @11:56PM (8 children)
From what little I could see, all these kits do is make live audio and video feeds to Google. Any AI is at Google, done as a side effect of harvesting even more of the user's lives.
Bent, folded, spindled, and mutilated.
(Score: 3, Touché) by takyon on Wednesday April 18 2018, @12:04AM (1 child)
It's ok, kids don't have a right to privacy.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by c0lo on Wednesday April 18 2018, @06:15AM
But can they be registered as sex offenders if they are sexting to AIY?
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 0) by Anonymous Coward on Wednesday April 18 2018, @03:53AM
Thanks... saved me the time of fishing through it.
I was sure hoping all the AI was done ON THE RASPBERRY.
I would sure like to play around with some gesture-sensing stuff I could train for taking cues from me for various things... like calling the police ( and putting ambient room audio on the line, giving no indication the call was in progress, other than making noises a clock might make, just to let me know the call is in progress), calling me on the phone, causing a sprinkler to spray my front stoop ( discouraging JW's and other unwanted door-to-door salesmen ), without giving it away that I was the one who caused the mayhem.
Just to help me get away from awkward social situations. You know it goes... phones get priority over someone face-to-face with you. And, once called, its OK if you tell the guy face to face with you "Sorry, I'm late! Gotta Go!", cause everybody accepts the machine, even a phone, as outranking them. If you want to say NO to someone, have a machine do it. People will accept rudeness from a machine far more than they will take it from another human.
( I still remember a friend who had his big Doberman trained to pee on people's leg on a hand signal... he used it for door-to-door people who could not take "no" for an answer. He would give them two chances to leave without being peed on. The third failed attempt to make them go away resulted in his dog blessing them with a stream of warm water. )
(Score: 2) by qzm on Wednesday April 18 2018, @04:19AM (1 child)
Do they? (sorry, real question, I would like to know).
In some of the descriptions it includes things like 'download different recognition routines from google', which is rather open ended but does imply somethings.
So, does anyone know, does the machine learning routines (not the training of course) actually run on the PI, or are these just streaming data to google?
If they are streaming to google, and of course are targeted at children, there are other questions to be considered.
If they do run things onboard, then they are at least somewhat interesting
(Score: 0) by Anonymous Coward on Wednesday April 18 2018, @07:22AM
I watched their video. It says it works without a cloud and without a net connection, so the processing is local. However all the voice kit does is STT and the Vision kit sounded like it comes with a pre-trained NN which knows a 1000 common objects. It sort of sounded like you could train a new NN, but there was no indication of a UI to do so. The devices are enclosed in a cardboard box. These are close to expensive junk. A nice mini-book on how to make these systems with standard parts rather than the more expensive, Google branded parts would have been better. You aren't learning any AI. All the AI is done for you. Think of these as just sensors which can't plug into your project's existing hardware but instead require another power source, motherboard, and cardboard enclosure.
If they're more than that then they really need to work on their marketing. The article didn't even link to the project's home page (I'm assuming it has one). They included the marketing video, a link to a blog post announcing the devices, and a link to Target to buy them.
I see this project as a result of a 'designed by committe' process. For the future I predect Google will either toss a ton of money at it and you'll be able to buy a lot of expensive sensors in a poor imintation of Lego Mindstorms and actual electronic modules, or it'll quietly disappear in two years.
(Score: 2) by Snospar on Wednesday April 18 2018, @07:56AM (2 children)
In the video it states that all processing is done locally on the device without an internet connection and not in the Cloud. So No, not just another Google spy feed.
Huge thanks to all the Soylent volunteers without whom this community (and this post) would not be possible.
(Score: 2) by leftover on Wednesday April 18 2018, @03:02PM
Yes, I am trying to find more about the "Vision Bonnet". Will probably just go buy one to to see.
Bent, folded, spindled, and mutilated.
(Score: 2) by leftover on Thursday April 19 2018, @01:47AM
OK, have one and I still don't know. The kit is very tightly wrapped, even has a custom pushbutton switch. My current impression is that the Vision Bonnet has the capacity to execute a NN but not to train it. I haven't even looked at the Android app host yet.
Bent, folded, spindled, and mutilated.