2020 BRN Discussion, page-3523

I didn't know about Akida two weeks ago, but was fairly quickly convinced that this one had potential.

One indicator that Akida is on the right path, is that both IBM and Intel also develops SNN's.

There is a heap of companies developing "AI" chips that are not really AI, most of them packing loads of cores together with separate memory each and claiming each their own revolution. Some are employing new tricks like early exit, to minimize computation. They might be very good at accelerating AI and doing other kinds of massive computation. They might also have some advantages over Akida in various applications.

What I'm noticing is that Brainchip has very little direct competition. There are energy efficient AI accelerators, but not quite as much as Akida and most are targeted for datacenters.

IBM's chip is not really in competition at the current price level and appears to be weaker than Akida. Intel's chip seems to be behind too and it's pricier. Akida might not be as precise as TrueNorth, but this might not matter on most edge devices, especially the price taken into consideration.

Brainchip also seems very focused on integration in various ways, like; accelerating existing DNN's, implementing AV conversion in their SOC, having their own development environment and selling IP blocks.

I see a relatively cheap product, with low power consumption, that can do a lot of magic, that is also easy to implement, that might be without significant competition in some market segments.

I can even speculate that some will find ways to decrease overall power consumption of mobile phones with Akida, by offloading tasks that are hard to do for traditional processors.