That's really sounds great, actually this technological development is called mixed mode designs terminologically, mixed mode controllers are very much in use in microcontroller based designs, this may help in this subject area as well.
Analog emulations will be able to function as neurons but it needs to be correctly morphed and second thing is the susceptible of analog devices with temperature and its nonlinearity to handle it and control it again digital domain will be required.
I think one of Carver Mead's often overlooked insights was that many analog functions can be emulated by building similar structures in silicon--such as his famous silicon cochlea which spawned the field of neuromorphic engineering. Yes, you can simulate its function with an algorithmic digital solutions, but only if you truly understand all the dynamics. Whereas an analog emulation might exhibit dynamics that are not well understood, but which nevertheless operate similarly to the way the real biological systems operate.
A single transistor synapse is a huge and probably necessary step, but the monumental interconnect problem seems almost insurmountable. But there should be little doubt that an analog approach will be a lower power solution than a digital one.
At very small geometries I wonder if error correction is needed to protect small analog signals from errors due to radiation... The need for error correction might require duplication of 'logic' and other inefficiencies...
Digital simulation of sufficient range should be able to be indistinguishable from analog. Step functions with small enough steps (note the lack of rigorous definition here) should be able to emulate continuous functions.
Oddly enough, it seems like digital simulation of sufficient resolution actually map into reality the same way that analog does. Drop that analog circuitry to the quantum level and it starts looking very digital. The difference is that digital simulation bits maintains state rather than going to the statistical weirdness of the quantum world. My biggest concern is that there may be real differences as a result.
I have been writing about artificial neural networks and neuromorphic systems since the 1990s when Carver Mead invented the artificial retina based on the human eye. Mead emphasized the need to use analog not only to save power, but to truly emulate the continuous analog functions of the human nervous system rather than just simulate them with step-time digital functions. This roadmap empasizes the same need for analog functions in neuromorphic systems. What's your opinion on analog emulation versus digital simulations?
What are the engineering and design challenges in creating successful IoT devices? These devices are usually small, resource-constrained electronics designed to sense, collect, send, and/or interpret data. Some of the devices need to be smart enough to act upon data in real time, 24/7. Specifically the guests will discuss sensors, security, and lessons from IoT deployments.