Skip to content
The Chat Archive
Sign in
Human-written demonstration (OpenAssistant volunteers)· OpenAssistant (human-written) OpenAssistant Conversations (oasst2)

Why is abacus computing more efficient for deep learning than GPU computing?

In OpenAssistant data, “assistant” turns were written by human volunteers as demonstrations. They are not model outputs.
  1. User

    Why is abacus computing more efficient for deep learning than GPU computing?

  2. Assistant

    Computations performed via abacus through physical manipulation of beads are orders of magnitude slower than computations performed with silicon transistors in GPUs. Silicon wafers capable of parsing trillions of voltage signals a second in support of deep learning are also smaller than traditional abaci. It is not likely that the abacus has any advantage over GPUs in terms of deep learning efficiency.

  3. User

    Given an infinite amount of people working together with an infinite number of abacus is it possible for them to complete the task faster than a GPU?