"There is controversy about whether parrots are capable of using language, or merely mimic what they hear." https://en.wikipedia.org/wiki/Talking_bird#Cognition_controversy … Sounds like an apt comparison to me.
Calling transformers stochastic parrots is wrong and insulting, and reveals a basic enmity toward ML. Transformers certainly have their limitations, but they do much more than just regurgitate word sequences from the training corpus.
-
-
-
A controversy because it's hard to tell :p It's pretty clear what the connotation is. Check out the verb form from Google/Oxford Language:pic.twitter.com/GRgKpRojIm
Keskustelun loppu
Uusi keskustelu -
-
-
Tämä twiitti ei ole saatavilla.
-
Tämä twiitti ei ole saatavilla.
- Näytä vastaukset
-
-
-
I wish more papers about trained models released their training corpus in searchable form. It's hard to evaluate how far from regurgitation it is without that.
-
It seemed there's this conflicting goals of publishing future new papers on gathered data vs increasing current papers validity by sharing all components of the system.
Keskustelun loppu
Uusi keskustelu -
-
-
If you say that they do more the onus is on you to prove it not on the naysayers.
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
ok... but did your last paper not confirm that DL models are just smart parrots that memorize data and predict via similarity function? Without super dense sampling of text, transformers would fail right?
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.