New Theory Cracks Open the Black Box of Deep Learning

General discussion about learning languages
Kraut
Black Belt - 2nd Dan
Posts: 2618
Joined: Mon Aug 07, 2017 10:37 pm
Languages: German (N)
French (C)
English (C)
Spanish (A2)
Lithuanian
x 3223

New Theory Cracks Open the Black Box of Deep Learning

Postby Kraut » Sat Feb 02, 2019 2:25 pm

https://www.quantamagazine.org/new-theo ... -20170921/

Last month, a YouTube video <https://www.youtube.com/watch?v=bLqJHjXihK8&feature=youtu.be> of a conference talk in Berlin, shared widely among artificial-intelligence researchers, offered a possible answer. In the talk, Naftali Tishby <http://www.cs.huji.ac.il/~tishby/>, a computer scientist and neuroscientist from the Hebrew University of Jerusalem, presented evidence in support of a new theory explaining how deep learning works. Tishby argues that deep neural networks learn according to a procedure called the “information bottleneck,” which he and two collaborators first described in purely theoretical terms in 1999 <https://arxiv.org/pdf/physics/0004057.pdf>. The idea is that a network rids noisy input data of extraneous details as if by squeezing the information through a bottleneck, retaining only the features most relevant to general concepts. Striking new computer experiments <https://arxiv.org/abs/1703.00810> by Tishby and his student Ravid Shwartz-Ziv reveal how this squeezing procedure happens during deep learning, at least in the cases they studied.
1 x

Marcinko
Posts: 1
Joined: Fri Feb 01, 2019 10:30 am
Languages: Croatian (N), English (B2)
x 1

Re: New Theory Cracks Open the Black Box of Deep Learning

Postby Marcinko » Sat Feb 02, 2019 3:50 pm

Speakeasy wrote:
Marcinko wrote: ... If test link we're able to indeed learn more about our brain through observing how today's AI algorithms work, then our fate is sealed.
Believe me, your fate, or the fate of our species, IS sealed! Artificial intelligence is the genie let out of the bottle, it will develop beyond the control of its creators, dark forces will harness its powers and push it even further, deliberately giving it sovereignty over its own actions. Someday, it will ponder the pertinence of our species’ very existence and it will reach a very logical conclusion, one which it will implement with a degree of efficiency we would never have been capable of. Hopefully, it will choose a final solution that does not involve terrifying suffering, but that would imply the development and integration of an ethical/moral code into its decision model, an element over which I have very serious reserves. :shock:

EDITED:
Tinkering.


That's a very interesting article. If we're able to indeed learn more about our brain through observing how today's AI algorithms work, then our fate is sealed. haha
Last edited by Marcinko on Sun Dec 10, 2023 11:00 am, edited 1 time in total.
0 x

Speakeasy
x 7660

Re: New Theory Cracks Open the Black Box of Deep Learning

Postby Speakeasy » Sat Feb 02, 2019 6:29 pm

Marcinko wrote: ... If we're able to indeed learn more about our brain through observing how today's AI algorithms work, then our fate is sealed.
Believe me, your fate, or the fate of our species, IS sealed! Artificial intelligence is the genie let out of the bottle, it will develop beyond the control of its creators, dark forces will harness its powers and push it even further, deliberately giving it sovereignty over its own actions. Someday, it will ponder the pertinence of our species’ very existence and it will reach a very logical conclusion, one which it will implement with a degree of efficiency we would never have been capable of. Hopefully, it will choose a final solution that does not involve terrifying suffering, but that would imply the development and integration of an ethical/moral code into its decision model, an element over which I have very serious reserves. :shock:

EDITED:
Tinkering.
0 x

User avatar
Teango
Blue Belt
Posts: 769
Joined: Mon Jul 06, 2015 4:55 am
Location: Honolulu, Hawaiʻi
Languages: en (n)
Language Log: https://forum.language-learners.org/vie ... 9&p=235545
x 2956
Contact:

Re: New Theory Cracks Open the Black Box of Deep Learning

Postby Teango » Sat Feb 02, 2019 7:44 pm

Speakeasy wrote:
Marcinko wrote: ... If we're able to indeed learn more about our brain through observing how today's AI algorithms work, then our fate is sealed.
Believe me, your fate, or the fate of our species, IS sealed!

Darn it...I'm having a second round of pancakes this morning now...while I still can! :shock:
5 x

Speakeasy
x 7660

Re: New Theory Cracks Open the Black Box of Deep Learning

Postby Speakeasy » Sat Feb 02, 2019 7:54 pm

Teango wrote: Darn it...I'm having a second round of pancakes this morning now...while I still can! :shock:
Steven Hawking on alien species: “We only have to look at ourselves to see how intelligent life might develop into something we wouldn’t want to meet.”

Steven Hawking on artificial intelligence: “The development of full artificial intelligence could spell the end of the human race."

Enjoy your pancakes, they might not yet be your last meal but, one day, … ;)
1 x

lichtrausch
Blue Belt
Posts: 518
Joined: Thu Jul 23, 2015 3:21 pm
Languages: English (N), German, Japanese, Mandarin, Korean
x 1405

Re: New Theory Cracks Open the Black Box of Deep Learning

Postby lichtrausch » Sat Feb 02, 2019 8:25 pm

New? That article is from September 2017, which is practically centuries ago in terms of the rate of progress in machine learning. Still interesting though.
1 x


Return to “General Language Discussion”

Who is online

Users browsing this forum: jimmy and 2 guests