2nd West Coast Computer Faire March 3-5, 1978, I published my thoughts
The forecast made in the paper was made into a visual presented to the HP Micro Computer Interest Group Aug. 15, 1978. That 39 year old chart has been marked with where I think we are now, at right.
The vertical scale is power of 10 in bytes of high-speed dram memory (M=6=megabyte, B=9=gigabyte, T=12=byte). In terms of memory originally envisioned, we are at 7 (just under 10 gigabyte) not between 11 and 12. 64 bit addressing with caches and virtual memory do provide the software environment for what I envisioned. Thus, I take the present memory to be 500 GB, although 1 TB is plausible as well.
The horizontal scale is marked at Sep. 2017, when this is being written.
The lines diverging to right represent growth projections for computer capability into the (then) future. The information capacity of human brain is taken to be the horizontal lines such that overlap is shown as the bounded are marked in yellow.
If the brain capacity metric makes any sense, then we are already at the lower estimate now.
The paper concludes 12-120 million man years are needed to program a brain computer. This is way off. It speculates 3 megabyte will yield an intelligent program. Incidently, the total file byte size of my TelephoneMessagePad service and related code is about 2.5 megabyte.
Assume a 500 gigabyte hard drive, 32 bit addressing, and 3 GB of ram. We can use PL/D as engine and Python for supervision.
Let the Python supervisor handle caching and the PL/D inner loop handle processing. I guess the PL/D inner loop can handle maybe 100 thousand neurons/second. At that rate, it would take 256,000 seconds (just under 3 days) to sweep the entire hard drive.
Let neurons compete for running in the inner loop. There may be room for 1 gig of neurons but only the key 100, 000 are run. There is a competition. There are different teams of neurons under control of emotions which set global limits and set points.