A set of smart vending machines at the University of Waterloo is expected to be removed from campus after students raised privacy concerns about their software.
The machines have M&M artwork on them and sell chocolate and other candy. They are located throughout campus, including in the Modern Languages building and Hagey Hall.
Earlier this month, a student noticed an error message on one of the machines in the Modern Languages building. It appeared to indicate there was a problem with a facial recognition application.
“We wouldn’t have known if it weren’t for the application error. There’s no warning here,” said River Stanley, a fourth-year student, who investigated the machines for an article in the university publication, mathNEWS.
I feel like it’d be tough to find a chip powerful enough to capture demographic attributes while also cheap enough to ship in vending machines? But admittedly I’ve little context on embedded systems and their capabilities
While I have no idea how much a computerized vending machine costs, I found this article about a age/gender classifier that runs on a Raspberry Pi 4.
Looking at the machine’s big touchscreen, I think this classifier would fit on the SBC or require a relatively small upgrade.
The error message says “.exe” and looks like a dot net namespace.
Yikes, smh… Yep that’ll do it. I hate this timeline.
Same Raspberry Pi foundation that hired a cop with a background in surveillance tech as their “resident maker”?
Doritos are probably plenty powerful enough
There’s a vending machine in a co-working space I use sometimes that has a full on fridge and oven, and when you order off the touchscreen…something happens inside and sometimes a hot cooked thing comes out. I have no idea how it works and have not used it myself, because it seems possibly kinda gross.
Would it be significantly more costly than some of the features vending machines already have, such as card readers? I think these things are pretty costly already, but the profit margin on snacks and soft drinks is extremely high, so I’d imagine they’d recoup their cost pretty quickly.
Well I thought so, but apparently we have good enough software that can run on a rasp pi now, so clearly the hardware requirements are much much lower than I understood.
Geez, I remember needing to use cloud services just for simple OCR not that long ago…