2021 BRN Discussion, page-25064

Thanks Chapman, interesting find.
My initial comments about it are that:
-it's coincidental timing. They are fully aware Brainchip is about to release Akida so are clearly trying to steal their thunder and say they've got a viable competitor (even though they don't)
-It's still a research chip. It still sounds years away from a commercial product from all the comments. I think Mike Davies recently still said its still 5 years or more away. I suspect the new CEO Pat was involved here and basically said do we throwaway all these expensive years of research or spend a little bit more, throw it all together and see what we get.
-They are in the process of making their software compatible with Tensorflow and other main types. Brainchip already has Tensorflow compatible software.
-They are using preproduction Intel 4, an equivalent 7nm node. No doubt this is to indicate they have better power performance numbers when comparisons to Akidas 28nm chip will shortly be made
-I don't think their on chip learning is the equivalent of what Brainchip has achieved. They keep going on about how with Loihi 2 they can finally do backpropagation... True learning by humans is feedforward without any type of backprop. Plus Akida has been able to do backprop for years if required.
https://www.google.com/amp/s/finance.yahoo.com/amphtml/news/intel-advances-neuromorphic-loihi-2-150000964.html
New approaches for continual and associative learning: Loihi 2 improves support for advanced learning methods, including variations of backpropagation, the workhorse algorithm of deep learning. This expands the scope of adaptation and data efficient learning algorithms that can be supported by low-power form factors operating in online settings


It definitely sounds like they are still years behind. It wouldn't surprise me if Pat will give them a deadline which if they can't make, they license the relevant parts of Akida just to bring their chip to market.

Pure speculation, DYOR