Thanks IDD,
The need for 8 bits is a bit (no pun intended) surprising.
As you point out, the cloud is one option, and is urgent from a climate perspective.
Also, as you said, compatibility with the big boys could be useful.
PvdM said in "4 bits are enough"
https://brainchip.com/4-bits-are-enough/ :
With the focus on edge AI solutions that are extremely energy-sensitive and thermally constrained and require efficient real-time response, this advantage of 4-bit weights and activations is compelling and shows a strong trend in the coming years.
After mentioning that BrainChip first used 4 bits in 2015 (Studio?), he referred to some 2020 IBM research which found that 4 bits could be as accurate as 16 bit deep learning, only faster and more efficient, so maybe they need the accuracy of 32 bit or 64 bit weights and measures.
We've been working with Valeo for a few years, so it would be late in the day for the requirement to originate with them. However, we have only just learned of the Luminar/Mercedes tie up, so it may be that their foveated LiDaR needs higher precision.
ISL and the USAF radar may be another option. All those fighter planes look the same on radar.
Of course there are many things which we do not know ...