Natural language benchmarks don’t measure AI models’ general knowledge well

Re:Natural language benchmarks don’t measure AI models’ gene... by sudosysgen on Sunday August 23, 2020 @10:10PM (#24256988)

For loops in AI are actually only there to implement continuous mathematical operations such as matrix multiplications in a single core. Modern AI chips and implementations use very little to no for loops.

[ Parent | on HN ]

Re:Natural language benchmarks don’t measure AI models’ gene... by LdSGSgvupDV on Monday August 24, 2020 @12:06AM (#24257445)

Sorry, maybe it is a wrong analogy. What I want to emphasize is that there is no calculating time in nature. At least, we could not feel "lag" in our "Nature Server".

For the part of chips, did you mean that it runs the "for loop" as parallel in each micro-chip (or gate or whatever tiny structure it is)?

[ Parent | on HN ]

Re:Natural language benchmarks don’t measure AI models’ gene... by sudosysgen on Monday August 24, 2020 @01:15AM (#24257691)

>Sorry, maybe it is a wrong analogy. What I want to emphasize is that there is no calculating time in nature. At least, we could not feel "lag" in our "Nature Server".

Yes, but the reality is that there is lag in our projection of what nature is and that we just ignore it. Different parts of our perception actually don't exist at the same instant.

>For the part of chips, did you mean that it runs the "for loop" as parallel in each micro-chip (or gate or whatever tiny structure it is)?

Pretty much, the for-loop is unfolded and ran in parallel.

[ Parent | on HN ]