The biggest challenge with language benchmarks is how quickly they get obsoleted.

Jun 3, 2022 · 6:16 PM UTC

12
4
1
70
Replying to @gdb
Haha, nice.
Replying to @gdb
Well said! This is the greatest challenge in the majority of all benchmarks, with time they get obsolete. This might be as a result of rapid developments in the field.
2
Replying to @gdb
That’s the challenge of all benchmarks. After all, it’s a science and an art to discover and define good north stars in science.
2
2
55
Replying to @gdb
@gdb @drfeifei Great communication deduction but what are the terms/words(English) that’ll persevere for centuries same way Latin did with Newton, Descartes for modern day invention relationships?
Replying to @gdb
Is this a reaction to gpt-4chan?
Replying to @gdb
Isn't that a good thing?
1
Replying to @gdb
Maybe it needs an out of the box thinking!
Replying to @gdb
I wonder if Algorithms will influence language one day, rather than just the other way around. That could stabilize the “benchmarks”.
Replying to @gdb
Yeah so impressed…The racism scales: wired.com/story/efforts-make…

The Efforts to Make Text-Based AI Less Racist and Terrible

Language models like GPT-3 can write poetry, but they often amplify negative stereotypes. Researchers are trying different approaches to address the problem.

wired.com
Replying to @gdb
And how everybody keep pretending that they are still relevant
Load more