/sci/ - Science & Math » Thread #11294870

1MiB, 2560x1440, Network Earth.jpg

View Same Google iqdb SauceNAO

Anonymous Sat 11 Jan 22:41:02 2020 No.11294870 View Reply Original Report

Quoted By: >>11294877 >>11294981 >>11294982 >>11295053 >>11295071

Can someone explain in simple terms what the hype is around Neural Networks & Deep learning? I mean "wow, you can create node layers to recognise inputs and give outputs. Great. So what?"

Anonymous

Anonymous Sat 11 Jan 2020 22:43:16 No.11294877 Report

Quoted By: >>11294892

>>11294870
The answer is "yes". In fact it is just exactly what you said yourself.

Anonymous

Anonymous Sat 11 Jan 2020 22:47:04 No.11294892 Report

Quoted By: >>11294936

>>11294877
I know what it does. I just don't understand why people talk about it being the next great thing. What am I missing?

Anonymous

Anonymous Sat 11 Jan 2020 22:59:15 No.11294936 Report

Quoted By: >>11294973

>>11294892
It was the next best thing 10 years ago. Basically, they can learn any sufficiently continuous mapping between inputs and outputs and even the dumbest retards can use them. That's the main selling point.

Anonymous

Anonymous Sat 11 Jan 2020 23:11:02 No.11294973 Report

Quoted By: >>11294995

>>11294936
Thanks. Why haven't neural networks solved Captchas yet?

Anonymous

Anonymous Sat 11 Jan 2020 23:13:41 No.11294981 Report

Quoted By:

>>11294870
the hype is that while you can create node layers to recognize inputs and give outputs, you havent. So what you say? The fact that only few companies managed to actually create something so simple that also produces results and you havent is the hype.

Anonymous

Anonymous Sat 11 Jan 2020 23:14:45 No.11294982 Report

Quoted By:

>>11294870
the hype is that it is here to stay because it saves time

Anonymous

Anonymous Sat 11 Jan 2020 23:18:31 No.11294995 Report

Quoted By:

>>11294973
They have, several times already. Google is putting more and more noise in them for that reason. Traffic lights in the far distance and Gaussian noise etc.
Their plan worked well. Either captcha is proof you're a human, or someone develops a solution to the underlying problem.

Anonymous

Anonymous Sat 11 Jan 2020 23:36:32 No.11295053 Report

Quoted By: >>11295099

>>11294870
Normies think it's actually learning the way animals or people do. They will never listen to you if you tell them it's glorified curve fitting.

> With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.
~John von Neumann

Anonymous

Anonymous Sat 11 Jan 2020 23:43:26 No.11295071 Report

Quoted By: >>11295099

>>11294870
>I mean "wow, you can create node layers to recognise inputs and give outputs. Great. So what?"
You can apply that same statement to biological intelligence. So what?

Anonymous

Anonymous Sat 11 Jan 2020 23:51:40 No.11295099 Report

Quoted By:

>>11295053
Maybe animals and humans also do glorified curve fitting - by trying to find the optimal survival function. Not denying that it is on a completely different scale, but why could it not be based on a mathematically similar framework?

>>11295071
exactly

Anonymous

Anonymous Sun 12 Jan 2020 18:22:04 No.11296630 Report

Quoted By:

The hype is that it allows the machine to get really good at something. Stockfish (Chess engine) was destroyed by Alpha go Zero (AI-based) in chess not long ago.

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