Comments by "Luiz Devil" (@luizdevil6855) on "Welch Labs" channel.

  1.  @T3sl4  turns out big-o doesn't matter, all algorithms are almost linear scaling if you have infinite money to pay for GPUs. all problems are polynomial. want to solve the travelling salesman problem with N=10, no problem, use 1 CPU want to solve it for N=1000, no problem, just use 100.000 CPUs does it matter its exponential ? no
    2
  2. The real magic was Systolic array hardware, aka, CUDA/RTX became cheap.
    1
  3. 17:05 2 orders of magnitude to human level processing. it just 8 yottaflops to simulate a human
    1