Comments by "Keit Hammleter" (@keithammleter3824) on "Future Computers Will Be Radically Different (Analog Computing)" video.

  1. Is this Veritasium guy clueless or just fast with his facts? He makes mistakes at detail level and and overall level. For instance, he is wrong about the development of computing: Analogue computers took off during and after World War 2 as they were found to be effective in solving all sorts of dynamic and control systems (eg missile guidance, auto-pilots) problems. Digital computers essentially began during WW2 (in the USA for ballistic calculations; in Germany for airplane stress calculations) and took off in late 1950's due to accounting, stock control and other business application development. Analogue computers continued until the 1980's, when good applications running on cheap personal computers became available. Neither came significantly before the other. In comparing digital computation with analogue, he's got it very wrong. It's not a case that to add 2 numbers in digital you need 50 transistors and for analogue you just need a wire connection as he claimed. Those of us who have worked on actual analogue computers know that the basic computing unit is a thing called an operational amplifier - which typically needs about 30-50 transistors (early, less accurate analogue computers used vacuum tubes in lower numbers). And just as much power, What killed analogue computers was that with the development of good applications in the 1980's allowing fast problem set-up, digital became much cheaper and easier. He's confused distributed (also termed parallel) processing with analogue processing. They are two different concepts. You can have distributed digital processing or serial processing. You can also have serial (termed cascaded) or distributed analogue processing. Using MOSFETS as simple multipliers as he described is not new. But it is good only in a few niche applications due to two fundamental features of MSOFETS: 1) the current is NOT a simple product of voltage - they are not linear, and 2) it's darn hard to make them consistent - ie each MOSFET giving the same result as another. That's why general purpose analogue computers used those carefully engineered 30-transistor operational amplifiers, and not just one transistor per computing element.
    1