goosed2: I am not smart enough to decipher Suresh answer vis a vis actual real world application of Poet involvement Selfishly hoped he could have commented specifically how the OI design compliments these neuron memory applications for AI

In simple terms here’s my take:

• You can do computations in photonics which is very fast and thus preferable. However, without photonic memory, you have to interface with traditional, electronic memory in order to read input parameters, store intermediate results, read intermediate results, and store your final results. With AI, all of this involves extremely huge amounts of data and consequently huge amounts of photonics/electronics conversions back and forth, which is an energy consuming performance bottleneck. Anyway, the POET optical interposer provides the infrastructure for that. Good for POET!
• The big advantage of photonics memory would be that you can get rid of the photonics/electronics conversions and do everything purely in photonics. Still you will have different photonic components, e.g., compute components and memory components, which have to interface with each other. The POET optical interposer provides the infrastructure for that. Good for POET!