Several points here:
- Spike rate or spike time
- Existing algorithms only partly take advantage of neural technology (MB working on algorithms to take advantage of neuromorphic computing)
- Still need automotive-grade neuromorphic chips before this tech is common in cars.
- Extend vehicle range
- Increase the number of AI functions
- 2020 Joined Intel Nueromorphic Research Community
- Expanding collaboratons with research partners and unis
Spike rate was shown by Thorpe, referring to Adrian's 1920's paper, to be redundant. Thorpe invented spike rank coding (N-of-M coding, enabling insignificant spikes to be ignored. Maybe the reference to spike rate is a red herring for competitors, but their DD would be deficient if they don't know about MB + Akida and that Akida uses rate coding.
New neuromorphic-aware algorithms may produce further time/energy savings.
We know MB is developing an MB.OS processor with Nvidia and some friends, so this will be automotive grade (probably similar to RadHard for NASA, and Vorago has already done that for Akida).
Hi ILL,
I see several points here:
- Spike rate or spike time
- Existing algorithms only partly take advantage of neural technology (MB working on algorithms to take advantage of neuromorphic computing)
- Still need automotive-grade neuromorphic chips before this tech is common in cars.
- Extend vehicle range
- Increase the number of AI functions
- 2020 Joined Intel Nueromorphic Research Community
- Expanding collaboratons with research partners and unis
Although we have seen a couple of articles on spike rate recently, spike rate was shown by Thorpe, referring to Adrian's 1920's paper, to be redundant. Thorpe invented spike rank coding (spike time coding) (N-of-M coding), enabling insignificant spikes to be ignored. Maybe the reference to spike rate is a red herring for competitors, but their DD would be deficient if they don't know about MB + Akida and that Akida uses rate coding. So really lip service to the NDAs, or a nod to a couple of decades of waster research.
New neuromorphic-aware algorithms may produce further time/energy savings.
We know MB is developing an MB.OS processor with Nvidia and some friends, so this will be automotive grade (probably similar to RadHard for NASA, and Vorago has already done that for Akida).
Increasing the number of AI functions can only be good for Akida.
Extend vehicle range ties in with increasing the number of AI functions plus advantages from neuromorphic-aware algorithms.
... and then there's the ecosystem - we're in with MB, Nvidia, and probably everyone else on MB's dance card.