I have added some discussion of how these results relate some earlier observatins of mine in a longform piece at Medium::https://link.medium.com/2FUz9lTsiS
Show this thread
Folks, let’s stop pretending that the problem of object recognition is solved. Deep learning is part of the solution, but we are obviously still missing something important. Terrific new examples of how much is still be solved here: #AIisHarderThanYouThinkhttps://twitter.com/MaxALittle/status/1068089091714097152 …
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Nobody is pretending it is solved. However, some people are claiming that people are pretending it is solved. Name me one researcher who is pretending?
-
Go back to Lecun, Bengio and Hinton’s 9 page Nature paper in 2015 and show me one hint there that this kind of error was possible. Or recall initial dismissive reaction to https://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Nguyen_Deep_Neural_Networks_2015_CVPR_paper.pdf …
- 8 more replies
New conversation -
-
-
Reminds me of the (possibly apocryphal) story of autonomous test vehicles in Australia colliding with kangaroos because they disappeared during hops (no longer ground connected!)
-
Bit of a base rate problem, too :) (ie a lot of obstacles to avoid)
End of conversation
New conversation -
-
-
If I may respectfully beg to differ, cost effective, excellent ROI for object recognition has been profitably deployed for 30+ years. It is "perfect"? No. Does it need to be, to be viable? No-
#SafetyFirst#machinelearning#BigData#aiethics#AI#MachineVision#DeepLearning -
Depends on the task. I am struck by your
#safetyfirst hashtag, though, because I think a lack of safety is exactly these results point by Nguyen and collaborators point to. You should read the paper. - 1 more reply
New conversation -
-
-
Good news for Jamaica! (Everything is a bobsled!)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
It's great that the authors made their code available:https://github.com/airalcorn2/strike-with-a-pose …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Regarding Moravec's paradox and human imagination isn't Deep Learning just meant to be a _decentralized intelligence_ extension? Still wondering why human is the benchmark? for replacing? instead of human+?
#MultiplicityThanks. 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.