/sci/ - Science & Math » Thread #12974903

42KiB, 800x500, computation7.jpg

View Same Google iqdb SauceNAO

How does intelligence arise from computation?

Anonymous Wed 14 Apr 17:39:51 2021 No.12974903 View Reply Original Report

Quoted By: >>12974939 >>12975207 >>12975386 >>12975416

Some days ago I've read an interview with OpenAI founder which said they are not interested in more data but in more raw computation power because that will provide the intelligence. How does intelligence/information stems from computation? And Do you have any videos or
books about that?

Anonymous

Anonymous Wed 14 Apr 2021 17:47:15 No.12974939 Report

Quoted By: >>12975084 >>12975363

>>12974903
Because nobody can program for real anymore. The vast majority of people who studied after 2000 or so are the k8nd of people who would go to finance and aim for walstreet, looking to rake in easy money. Most of the recent stuff is just throwing more computing power at the problem and hoping it will find something that works.

Anonymous

Anonymous Wed 14 Apr 2021 18:27:21 No.12975084 Report

Quoted By: >>12975087 >>12975150

>>12974939
Name some real programmers

Anonymous

Anonymous Wed 14 Apr 2021 18:28:27 No.12975087 Report

Quoted By: >>12975169

>>12975084
your mom, anon

Anonymous

Anonymous Wed 14 Apr 2021 18:41:05 No.12975150 Report

Quoted By: >>12975169

>>12975084
Terry A. Davis

Anonymous

Anonymous Wed 14 Apr 2021 18:44:23 No.12975169 Report

Quoted By:

>>12975087
>>12975150
So it was all a lie. many such cases

Anonymous

View Same Google iqdb SauceNAO 1200px-Neural_network_example.svg.png, 99KiB, 1200x1600

Anonymous Wed 14 Apr 2021 18:50:45 No.12975207 Report

Quoted By: >>12975280 >>12975346 >>12975378 >>12975386 >>12978255

>>12974903
We actually don't know how deep learning neural networks produce their outputs (AKA their "intelligence"). It's a mystery. We only know that more computational power = better performance
https://www.youtube.com/watch?v=GiWK4PREXiU

Anonymous

Anonymous Wed 14 Apr 2021 19:07:14 No.12975280 Report

Quoted By:

>>12975207
Only good answer of the thread so far, thanks anon

Anonymous

Anonymous Wed 14 Apr 2021 19:24:35 No.12975346 Report

Quoted By: >>12975389

>>12975207
surely that's not true? all these things have been carefully mathematically defined. we know how they learn because we came up with how they could do so

Anonymous

Anonymous Wed 14 Apr 2021 19:27:07 No.12975363 Report

Quoted By:

>>12974939
>waaaah people cant program anymore
The existence of shitty programmers now that the bar has lowered does not mean good programmers don’t exist.
Regardless, AI problems go beyond “just programming better.”

Anonymous

Anonymous Wed 14 Apr 2021 19:31:11 No.12975378 Report

Quoted By: >>12975396

>>12975207
>we brute force a formula but we don't know what the formula is actually "thinking" when it's giving us an output thus the formula is a mystery
ok

Anonymous

Anonymous Wed 14 Apr 2021 19:32:44 No.12975386 Report

Quoted By:

>>12975207
That's not really true, you could in fact open up a trained neural network and figure out pretty easily how it does which thing and where do the outputs come from. That thing which is impossible is creating the algorithm by hand ahead of the fact.

>>12974903
Intelligence is an emergent property, the more calculations you can do the more it will seem like and by extension be intelligence. More over the current pathways to AI like neural networks require immense computational power and the more you have available the better networks you can train

Anonymous

Anonymous Wed 14 Apr 2021 19:33:09 No.12975389 Report

Quoted By: >>12975430

>>12975346
We understand them to some extent, and the theory for ML is getting more attention than before, but being able to develop a underlying theory of ML (ie, not just fragmenting statistical learning, how fast can we converge, and computational learning, what can be in principle leaner, together) is still active research. ML is simultaneously very lucrative ground for the theoretical computer scientist yet hard to break into. There’s a huge order of magnitude more of mathematical maturity in breaking into ML theory than empirical ML right now.

Anonymous

Anonymous Wed 14 Apr 2021 19:34:55 No.12975396 Report

Quoted By: >>12975411

>>12975378
It’s definitely more clever than brute forcing. Anon we get it, you know jackshit about the subject, but no not everything is as simple as your reductive view.
People like you are part of the problem, since stuff like this pedals ideas that all of CS is brute forcing answers out

Anonymous

Anonymous Wed 14 Apr 2021 19:38:50 No.12975411 Report

Quoted By:

>>12975396
>It’s definitely more clever than brute forcing
I actually wanted to use the term "smart force" but I didn't expect anyone to get that so I settled with brute force.

Anonymous

Anonymous Wed 14 Apr 2021 19:40:28 No.12975416 Report

Quoted By: >>12977460

>>12974903
Define intelligence. Because current AI is swarm intelligence, not human intelligence.

Anonymous

Anonymous Wed 14 Apr 2021 19:45:37 No.12975430 Report

Quoted By: >>12975572

>>12975389
ah right, I understand. i shall now dedicate my time to proving the incompleteness of machine learning and become the next gödel :)

Anonymous

Anonymous Wed 14 Apr 2021 20:28:02 No.12975572 Report

Quoted By: >>12976214

>>12975430
>anti-theory fags in ML yet again
There’s a lot more to theory than classification theorems, but sure whatever makes you sleep better at night

Anonymous

Anonymous Wed 14 Apr 2021 23:01:49 No.12976214 Report

Quoted By: >>12978391

>>12975572
i was joking. it was a joke. i'm not going to prove the incompleteness of machine learning.

Anonymous

Anonymous Wed 14 Apr 2021 23:04:56 No.12976227 Report

Quoted By:

More powah = more INT.

Anonymous

Anonymous Thu 15 Apr 2021 05:46:44 No.12977448 Report

Quoted By:

bump

Anonymous

Anonymous Thu 15 Apr 2021 05:55:23 No.12977460 Report

Quoted By:

>>12975416
This is the problem, people conflate intelligence and consciousness and all that stuff.
To me intelligence would be accomplishing the given task as well as a human or better. To people in AI that are grounded in reality (i.e. not talking to some pop-sci interviewer) this is what they mean as well.

Anonymous

Anonymous Thu 15 Apr 2021 11:26:17 No.12978255 Report

Quoted By:

>>12975207
>mathematical mysteries of deep neural networks
>never talks or even touches statistical learning theory
I hate every single one of you faggots who push this shit.

Anonymous

Anonymous Thu 15 Apr 2021 12:19:21 No.12978370 Report

Quoted By:

AI is a meme

Anonymous

View Same Google iqdb SauceNAO thumb_beboldi-am-one-of-simple-m (...).png, 88KiB, 200x493

Anonymous Thu 15 Apr 2021 12:27:48 No.12978391 Report

Quoted By: >>12978584

>>12976214

Anonymous

Anonymous Thu 15 Apr 2021 13:23:36 No.12978584 Report

Quoted By:

>>12978391
kek this is me

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches