Go Back  Bike Forums > The Lounge > Foo
Reload this Page >

Will computers become spiritual?

Notices
Foo Off-Topic chit chat with no general subject.

Will computers become spiritual?

Old 11-19-19, 04:47 PM
  #76  
wphamilton
Senior Member
 
wphamilton's Avatar
 
Join Date: Apr 2011
Location: Alpharetta, GA
Posts: 14,804

Bikes: Nashbar Road

Mentioned: 70 Post(s)
Tagged: 0 Thread(s)
Quoted: 2671 Post(s)
Liked 121 Times in 84 Posts
Originally Posted by tyrion View Post
Theoretically speaking, there's nothing a quantum computer can do that a traditional computer can't, given enough time.
It's known that Turning-complete computation models can describe quantum computers, as you imply, but "given enough time" is a little weak as far as the difference goes. Algorithms which cannot be solved in polynomial time (Turing machines, ie computers) can theoretically be computed by a quantum machine.

But I am personally a little uneasy with the whole paradigm of equivalence, because a quantum machine isn't really going to be computing what are essentially boolean logic algorithms. Which is basically all that any computer program is, a giant convoluted boolean expression.

There's too much space between what quantum architecture might accomplish and what we know is true about traditional computation models. Of course, I am speculating wildly about things in the mostly unknown regions so don't take any of this very seriously.
wphamilton is offline  
Old 11-19-19, 06:42 PM
  #77  
RubeRad
Keepin it Wheel
 
RubeRad's Avatar
 
Join Date: Aug 2011
Location: San Diego
Posts: 8,775

Bikes: Surly CrossCheck, Moto Fantom29 ProSL hardtail

Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked 205 Times in 159 Posts
Originally Posted by Seattle Forrest View Post
I'm enjoying this conversation, and I think you're doing an excellent job providing a sanity check. 🙂
Likewise. Let's continue to be like Fonzie and stay out of P&R

I don't see an obvious parallel among humans for unzipping a file, or compiling source code into binaries. Appreciate the attempt to keep my feet on the ground though.
My point was, the training data for the classifier is the input, so different input/different results, not surprising.

But I think the compiler analogy is more apt. What you get when you provide the training data, is not output; what you get is a classifier, an executable that behaves a certain way -- produces certain outputs given certain inputs. And differently trained classifiers will produce different outputs for the same inputs. Very similar to when you provide source code to a compiler you get an executable; different source code-->different behavior on the same inputs.

Now you're right, we can't pinpoint inside the neural network, or even in the training data, 'this is why it's classifying this dog as a cat' -- not the same way we can step through source code line-by-line in a debugger and understand what it's doing. But we can influence the behavior by introducing new training data; especially by taking operational false-positives and false-negatives and using them with manually corrected classifications.

Here's a thought though. A static neural network is completely deterministic (AFAIK deep learning doesn't involve pseudo-randomness). If you ask a human the same question on different days, you may get different answers. But that's because the 'inputs' are not really constant, the human is assessing as input data all of his past experiences, and new experiences may cause answers to the same questions to change.

But a neural network classifier can be like that, if in place of 'new experiences' you add 'augmented training data'. So it can 'grow/develop'
RubeRad is offline  
Old 11-19-19, 06:44 PM
  #78  
RubeRad
Keepin it Wheel
 
RubeRad's Avatar
 
Join Date: Aug 2011
Location: San Diego
Posts: 8,775

Bikes: Surly CrossCheck, Moto Fantom29 ProSL hardtail

Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked 205 Times in 159 Posts
Originally Posted by tyrion View Post
That wasn't AI, that was relatively straightforward data analysis combined with psychographics.
My understanding was, they took a relatively small sample, trained a classifier to be able to predict persuadability based on social media attributes (used AI), then applied that classifier on gigantic scales, especially in battleground states, to identify likely persuadable individuals to target with advertising.
RubeRad is offline  
Old 11-19-19, 08:09 PM
  #79  
tyrion
Senior Member
 
tyrion's Avatar
 
Join Date: Oct 2015
Location: San Diego, California
Posts: 2,457

Bikes: Breezer Radar

Mentioned: 22 Post(s)
Tagged: 0 Thread(s)
Quoted: 1313 Post(s)
Liked 183 Times in 118 Posts
Originally Posted by RubeRad View Post
My understanding was, they took a relatively small sample, trained a classifier to be able to predict persuadability based on social media attributes (used AI), then applied that classifier on gigantic scales, especially in battleground states, to identify likely persuadable individuals to target with advertising.
Maybe I'm mistaken. I hadn't heard of Cambridge Analytica using AI classifiers.
tyrion is offline  
Old 11-20-19, 10:20 AM
  #80  
RubeRad
Keepin it Wheel
 
RubeRad's Avatar
 
Join Date: Aug 2011
Location: San Diego
Posts: 8,775

Bikes: Surly CrossCheck, Moto Fantom29 ProSL hardtail

Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked 205 Times in 159 Posts
I also might be misremembering. They certainly 'learned/trained' from a smaller dataset, and were thus able to predict for the general population, based on publicly available indicators, who was 'persuadable'. Maybe that's an AI classifier, maybe it's just 'Big Data'.
RubeRad is offline  
Old 11-20-19, 12:19 PM
  #81  
Seattle Forrest
Senior Member
Thread Starter
 
Seattle Forrest's Avatar
 
Join Date: Mar 2010
Location: Seattle, WA
Posts: 19,871
Mentioned: 53 Post(s)
Tagged: 0 Thread(s)
Quoted: 9483 Post(s)
Liked 970 Times in 590 Posts
That's kind of analogous to people and other animals. Anyone who has had pets has seen how done experiences color their perceptions for the rest of their lives.
Seattle Forrest is offline  
Old 11-20-19, 02:26 PM
  #82  
Hondo Gravel
Of Worms and Ruins
 
Join Date: Jan 2018
Location: Hondo,Texas
Posts: 1,157

Bikes: Too many Motobecanes

Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 706 Post(s)
Liked 261 Times in 189 Posts
My I-phone is a paranoid schizophrenic but I blame it on having to live with one bar. And I even have a booster. They built a cell tower out here but for some reason they haven’t activated it! ??
Hondo Gravel is offline  
Old 11-20-19, 04:00 PM
  #83  
tyrion
Senior Member
 
tyrion's Avatar
 
Join Date: Oct 2015
Location: San Diego, California
Posts: 2,457

Bikes: Breezer Radar

Mentioned: 22 Post(s)
Tagged: 0 Thread(s)
Quoted: 1313 Post(s)
Liked 183 Times in 118 Posts
Originally Posted by Hondo Gravel View Post
My I-phone is a paranoid schizophrenic but I blame it on having to live with one bar. And I even have a booster. They built a cell tower out here but for some reason they havenít activated it! ??
Your phone isn't paranoid. That "cell tower" is probably a CIA surveillance unit.
tyrion is offline  
Likes For tyrion:
Old 11-20-19, 04:35 PM
  #84  
Hondo Gravel
Of Worms and Ruins
 
Join Date: Jan 2018
Location: Hondo,Texas
Posts: 1,157

Bikes: Too many Motobecanes

Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 706 Post(s)
Liked 261 Times in 189 Posts
Originally Posted by tyrion View Post
Your phone isn't paranoid. That "cell tower" is probably a CIA surveillance unit.
I was thinking it was the KGB and their pesky e-mails they send me trying to get me to drink vodka.
Hondo Gravel is offline  

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off


Thread Tools
Search this Thread

Contact Us - Archive - Advertising - Cookie Policy - Privacy Statement - Terms of Service

Copyright © 2018 MH Sub I, LLC dba Internet Brands. All rights reserved. Use of this site indicates your consent to the Terms of Use.