Arora is a great speaker, the clarity of explanation in this talk is so underrated
@brian.josephson7 жыл бұрын
"About 10^40 atoms in the human head" is clearly an overestimate. As it happens, there's a web page where this question is addressed (education.jlab.org/qa/mathatom_03.html ) and this gives the answer as 4.56*1026.
@brian.josephson7 жыл бұрын
That should be 4.56 x 10^26 of course (for the non-mathematicians, ^ is used to indicate a superscript or a power).
@codegeek983 жыл бұрын
You may Eliezer Yudkowsky's writings on _why_ the Chinese Room is fallacious
@rsinh37923 жыл бұрын
Sir reviewer of the manuscript asked me this question "Please compare your proposed machine learning model with other based method in term of time complexity? Calculate the time complexity of your algorithm", Can you please help me how to address this question?
@NeeleshSalpe6 жыл бұрын
I think the current computation model like CPU/GPU +Memory which mimic basic Neumann computer is the bottleneck in complex computation. Our all algorithms and assumptions are based on this basic computation model, we are bounded by CPU/ memory. If we have solved questions like NP-hard we need different computation model which I don't know now ... and we want to mimic brain whose computation model is completely different than ours one, we are trying to fit the square peg in round hole. That is why we will be always restricted. Once the new computation model comes beyond Neumann's limitations we could solve the problem.. Even though we are trying to mimic neurons with deep learning models, unfortunately, we can not mimic , as underlying base computation unit is not same!
@gJonii5 жыл бұрын
You can simulate neural networks in polynomial time using our computers, so there's no particular point in changing the architecture.
@qwerdbeta4 жыл бұрын
He shrugs off noncomputable programs as very rare. They are rare in applied comp sci today, but definable, formally, they are so common compared to comparable problems (lower infinity cardinality). This guy is so loose he makes errors!