[ExI] Neural networks score higher than humans in reading and comprehension test
interzone at gmail.com
Tue Jan 16 22:45:29 UTC 2018
I'm not trying to take anything away from the accomplishments, but current
deep learning architectures are nowhere close to strong AI, and in my
humble opinion are likely a dead end in that regard.
I will quote the inventor of Keras which is a heavily used deep learning
library layered on top of Tensorflow (or other) architecture to sum up my
"Neural networks" are a sad misnomer. They're neither neural nor even
networks. They're chains of differentiable, parameterized geometric
functions, trained with gradient descent (with gradients obtained via the
chain rule). A small set of highschool-level ideas put together
"I'd say ML is both overhyped and underrated. People overestimate the
intelligence & generalization power of ML systems (ML as a magic wand), but
underestimate how much can be achieved with relatively crude systems, when
applied systematically (ML as the steam power of our era)"
On Tue, Jan 16, 2018 at 5:35 PM, John Clark <johnkclark at gmail.com> wrote:
> On Tue, Jan 16, 2018 at 1:31 PM, Dylan Distasio <interzone at gmail.com>
>> There is zero comprehension.
> If that is true and if neural networks can can now answer questions about
> what they just read better than a human can (and they can) then I don't
> know what "comprehension" means,
> whatever it means it can't be anything very important.
>> The nets are dead behind the eyes.
> If that is true then being "
> dead behind the eyes
> " is an advantage because it makes you more intelligent. Regardless of
> how you want to spin it the fact remains that as of today a computer can
> answer questions about what it just read better than a human can, and
> the implications of that fact are momentous.
> John K Clark
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat