Future Computers Will Be Radically Different (Analog Computing)
VeritasiumYouTube ↗
Further reading 1
Caption notes
Analog devices trade precision and repeatability for fast, efficient continuous computation. Neural-network matrix multiplication is a candidate application because it tolerates some error.
Original passages · 6
you can add two currents, just by connecting two wires.most of the time and energy goes into fetching those weight valuesneural networks don't need the precision of digital computers.you convert it from the analog domain, back to the digital domain,but also single-purpose, non-repeatable, and inexactthe resulting current is the product of voltage times conductance