With a company like Weebit Nano, it is always about the science...

With a company like Weebit Nano, it is always about the science and technology.
Lets come back to the science.
Published in Nature Communications a few days ago, very high quality paper. The chip used in the paper was 130nm design (from Leti, I guess) and arrayed to build a synase-like matrix to perform in memory computing.
https://www.nature.com/articles/s41467-023-37097-5

The paper really proved that with this ReRam matrix, the memory can be build up to learn and accumulate 'experience' to direct the next move.
This device learns faster than the conventional software directed machine learning and consumes much less energy (this is only the 130nm chip, what if it is 22nm and below?)
Plus the memory in this chips can be transplant to another one and do the learning without any barrier.
And the working mode of this chip can be event driven, otherwise, it can remain in standby. Not like the big server AI, on working mode all the time to work. This saves a lot of energy and redundancy.

.........