There's no shortage of important and largely untouched problems in AI. We have no energy to waste on complaining about academic trivialities
-
What do you care about most: credit assignment for a finding no one will remember in 5 yrs, or building the most important tech of our time
Put differently: shouldn't the research & publication system prioritize *the speed of progress* rather than I would call "academic drama"?
-
-
Improving the ratio signal / noise matters in nowadays academia. Too much energy wasted there.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
1) Many times, it takes far longer to debunk bad examples than to do new research.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
2) E.g. Now the veracity of the entire field of social psychology is being questioned because old research had questionable eval.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
3) Most DL works these days resemble experimental sciences (more than oldschool ML) in that they analyze complex blackbox-like systems.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Progress towards what? A noncomprehensible world where no human understands what's going on? Local minimum of AI with humans as data-slaves?
-
Unless we refine the "academic drama" protocols, humans cannot meaningfully communicate with each other and review the progress.
End of conversation
New conversation -
-
-
4)
@yoavgo's post about more rigor and better eval is important lest we end up in the same crises as many other experimental fields. -
The many practical applications may help avert a real "replication crisis". "Does your product work?" displacing mere "academic drama".
- Show replies
New conversation -
-
-
An important point of
@yoavgo's post is that MPUs w/ poor eval and overblown claims doesn't speed up progress, and in fact gets in the way. -
Furthermore there is little cost to flooding the system. Unlike, say, a mathematician who regularly posts proofs eventually shown incorrect.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.