What is your timeframe for deploying self driving cars at scale ?
The more you experiment at the margins, the more you realize that the stories we tell ourselves about why specific deep learning techniques work (e.g. residual connections, self-attention, BN...) are unreliable narratives that don't match the mathematical reality on the ground.
-
-
-
By 2200 or more, we should have a reliable level-5 self driving cars. Don't bother to thank me for such wonderful prophecy.
- Show replies
New conversation -
-
-
Cactus. Can you explain this?
-
I think this is a pattern that is pretty common in ML: There is a task we want to solve. There is an intuitive way to solve it, which we try to break down into components. We figure out some matrix operators that produces a similar output to each component and put it together.
- Show replies
New conversation -
-
-
Successful connectomes express the products of learning through reduction. Reduction erases the tracks of the connectome’s learning process from itself.
-
And then there are evolution’s tracks to uncover too
End of conversation
New conversation -
-
-
"The more you experiment at the margins, the more you realize that the stories we tell ourselves…are unreliable narratives that don't match the…reality on the ground." I was able to generalize this. That makes it more powerful, right?
-
Just like Newton dynamics broke when we started looking at extreme scale planetary dynamics. Those were the margins of Newtonian Theory.
- Show replies
New conversation -
-
-
Like those who see early stopping as a fabulous free lunch and those who deplore it? We know who is wrong in this case!
-
What's not to like about early stopping? I suspect somebody somewhere is chained to a rock and being eaten by eagles every day anew for it.
End of conversation
New conversation -
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.