Hi DingoBorat,
I agree that the Dev Kits are not targeted at hobbyists - however, the Mini PCIe Board, which is only ~ 1/10th of the price of a Raspberry Pi Development Kit, but still sets you back US$499, was intended “for anyone from hobbyists to companies who wants [sic!] to just plug 10 boards in for specific applications, all the way through to licensing IP”, according to Rob Telson in the January 2022 interview with Sally Ward-Foxton:
View attachment 53624
Erm, yes, it does. He literally writes “We build and test a technically simple prototype of the proposed physical RC system employing an inexpensive Arduino microcontroller”. How else could he have written his paper?
So he obviously knows exactly how much money it cost them to build the prototype.
The way I see it, this research is not about commercialising a product and pitting it against the “competition” such as Brainchip (which was just an example to put the prototype’s inexpensiveness in relation to commercially available neuromorphic hardware), but a life hack for researchers on a tight budget, so to speak.
If a US$50 AKD1000 USB stick
had been available, he might have gladly picked that one instead of the Arduino microcontroller, for all we know.
Lots of academic research is fundamental research, which aims to improve humans’ understanding of the natural world - primarily the pursuit of knowledge for more knowledge.
But maybe my cursory glance at the paper missed the part where he expressed his intention to commercialise the prototype. Did you spot any evidence of that?
Note that he doesn’t consider his prototype a panacea: “… we argue that
in certain practical situations the efficacy of the physical RC system may exceed the one of optimised machine learning software run on a high-performance workstation.”
And while it is admittedly embarrassing for the preprint’s author holding a PhD in Electrical Engineering to also have gotten the Akida Dev Kit’s power consumption wrong (thank’s for explaining,
@Diogenese!), to me this error still doesn’t alter the result of his cost comparison (see my prior post on that matter), as his prototype is very low-power itself - the entire experimental setup apparently consumes less than 1 W.
View attachment 53627
Yep, looks like it.