even on highways, driverless cars are still risky, video captioning is not reliable yet, robot grasping still has a long way to go, and AFAIK no actual species has yet been saved. my quote, taken out of context, wasn’t about AI’s potential, it was about what had been deliveredhttps://twitter.com/ylecun/status/1064781702202036224 …
-
FYI Mobileye system did not use convnets in 2015. They were looking at adding them to the chip at that time. The rest is good old computer vision. Apparently still on par with recent conv-net based AP 2.0.
aside from the fact that you may have overclaimed on the convnets (?), i don’t understand how the FB use (?) of videos addresses the quality of current video captioning in a scientific way.
