-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
I probably won't able to change Mr.LeCun's mind here but it's high time to abandon datasets. An algorithm should get its input from the messy real world, process continuous input in real time, its output should be a prediction corrected by the real life feedback. Plus no labels.
- Još 1 odgovor
Novi razgovor -
-
-
Learning representations is precisely why deep learning sucks. Unlike DNNs, the brain can instantly see a complex object it has never seen before. A DNN cannot detect anything unless it already has a representation of it in memory. This is why all current AV projects are doomed.
-
Doomed is an incorrect assumption. Perhaps a sideways movement, reconstruction or feedback is needed. In the meantime, they beg for funds by claiming progress in papers, miniscule as it may be.
- Još 5 drugih odgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.