GPT-2 v Wolfram Alpha Prediction is not understanding, and never will be. 1/2pic.twitter.com/KxC5xP6oac
Show this thread
Language understanding is ultimately about parsing sentences and paragraphs into intermediate representations & acting on them. Wolfram Alpha can do that (for some cases, not all see http://rebooting.AI ); GPT-2 can't. 2/2
-
-
Yep!
#Language understanding requires breaking down sentences into subject, predicate, object and the "parts of speech" roles of each of those components. See: https://www.slideshare.net/kidehen/understanding-29894555/55 …#Logic#SemanticWeb#Ontology#KnowledgeGraphpic.twitter.com/vscE3hEsbt
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Yup. I helped launch w|a (think i was technical first official employee, for launch). The language exists ontologically in many different ways at once. It’s symbolic and metaphorical. And in areas like quantity and math it’s much easier to reason in different symbolic frames
-
I’ve been coding up a demo of parsing a book into significant terms in order of appearance, pulling images for these terms, machine learning scene descriptions from those images and then having a GAN compose a new story from the ordered image descriptions.
- 1 more reply
New conversation -
-
-
The schoolboy in me can’t resist pointing out that GPT pronounced in French is “j’ai pété” (I have farted)! Pretty much captures it I think ;-)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
mmmm, chowder.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
More precisely, its about superimposition of most likely images corresponding to the word images or to the word sounds
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.