Research fields come in two flavors: those where your career progresses when you actually get something right (reality-grounded), and those where it's enough to create the appearance of getting something right (belief-grounded). Deep learning research is somewhere in between
-
-
Show this thread
-
As a result, the deception factor in deep learning papers is often quite high
Show this thread
End of conversation
New conversation -
-
-
Yes! Another way to think about this is whether your work is being evaluated top down (e.g. by authority figures/gatekeepers) or bottom up (e.g. by customers). You can cheat when evaluation is top down, not bottom up.
-
Some bosses may have good criteria and technical knowledge Masses of "customers" may fall prey to delusions, e.g. Ponzi schemes But you are more right than wrong, in a statistical sense For me question is: Do you eat your own dog food? Related:https://twitter.com/nntalebbot/status/1151559733788205056 …
End of conversation
New conversation -
-
-
is that what objective reality is?
-
Yeah that's kind of a narrow definition, eh?
End of conversation
New conversation -
-
-
Agree 100%. How do you change if your success is determined by the second one?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
This is somewhat conflicting, on one hand, yes, metric-based evaluation is better, but doesn't it lead to very focused tunnel vision?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
User adoption is an objective statistic, but the means of achieving it can be through deception. Oftentimes deception is a better means to achieving user adoption. There is no difference between the two
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Might be your best tweet ever, or perhaps I'm reading too much in it For incentives we have darwinism and reinforcement learning We may need better (actionable) models for mass delusions and [epistemic] bubbles TBH: what to do about/in them?https://twitter.com/trylks/status/1374768620396548096 …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.
incentives
Subjective
Objective