These days AI developers seem to have forgotten the eternal truism of "garbage in, garbage out".

Data science? More like data alchemy: putting basic data through poorly-understood process and expecting gold.

@DissidentKitty AI devs have their own eternal truism tho:

"The more utterly incomprehensible a process the easier it is to sell to defence contractors and venture capitalists.

Doubly so if that process outsources institutionalized racism to an AI so they don't have to feel guilty about it or whatever."


As someone who works in this space, I think what many practitioners forget is to do the actual "science". Make testable hypothesis, design controlled experiments, try to eliminate confounders, LOOK AT YOUR FUCKING DATA, etc

@zxv Yeah. Like I said, too many people treat it like it's alchemy: just plug data into it and expect good results.

@DissidentKitty I mean programmers have long been guilty of claiming to measure one phenomenon when actually they are measuring another. And then over-optimising that measure.

Now that the technology has more impact they are hurting more people because of it.

@DissidentKitty this isn't a big deal if you know this since almost nothing important can be directly measured, but they usually don't bother.

Sign in to participate in the conversation
Radical Town

A cool and chill place for cool and chill people.