Neural nets using individual tubes as nodes? Although the current trend seems to be quantizing down to a minimal amount of bits to process more in parallel, in

Neural nets using individual tubes as nodes? Although the current trend seems to be quantizing down to a minimal amount of bits to process more in parallel, in an analogue system you could have a near "continuous" range of values.