Many “AI” companies are cheating everyone by not doing a good job of remembering essential details
Roger Grosse @RogerGrosse Amortized variational inference is neither amortized nor variational nor inference. x.com/jpillowtime/st…
Jonathan Pillow @jpillowtime Amortize means to “to pay off a debt with regular payments”
But in amortized inference you pay a big up-front cost to train an inference network, then inference is cheap per datapoint. Isn’t that the opposite of amortization?
Shouldn’t we call it Big Down Payment inference?
Replying to @RogerGrosse
The inferences now are “cheap” in the sense of “not worth much”, “not reliable”, “not storing the original data and context and environmental clues”. It tries to “snapshot” a reality that is intrinsically infinite and eternal – constantly changing, but within human ability to survive. What we have now is sham, a scheme, where the costs of hidden and inaccessible knowledge – that could be shared – is forced onto users, buyers, and future generations.
Filed as (Many AI companies are cheating everyone by not doing a good job of remembering essential details)
Richard Collins, The Internet Foundation
CC: @jpillowtime
