What the Tech?!? #4 - Digital Brains for a digital age
Ever since man created the first computer, we've striven to make them better, stronger, faster, and smarter.
Source: Pixabay
One way of making them better, first proposed over 70 years ago, is to recreate the architecture of the brain, creating connections that act much like neurons, in an attempt to replicate the way organic brains process information.
"Neural Networks", as they've been dubbed, have fallen in and out of the computer science spotlight since their inception, but over the last decade or so, have seen a golden age of advancement and utility. We've succeeded in creating a digital system that can "learn" new things by being "trained" instead of "told". This article will attempt to explain what a neural network is, but I won't go too far into how they work, as even expert computer scientists have a hard time determining what exactly is happening to data between input and output.
Simple layout, not so simple method
Source:Wikimedia Commons
Neural networks are built by layering nodes that usually send a signal in one direction from layer to layer, changing the data just a bit each time, until it reaches the output layer. Each node acts as a neuron in the brain, and is connected to every node in the layer above and below it. Each node has what's called a "weight" that determines it's importance when passing data into the next layer. This starts to create a very complicated system that is able to "learn" new things.
The video below, from the Youtube channel 3Blue1Brown, will explain the architecture and function of neural networks in much more detail.
"Machine Learning"
As explained in the video, neural networks can be "trained" to "learn" new tasks by repetition, instead of having a rigid set of instructions. By running data through the system, then tweaking the weights of each input as the signal "flows", these digital brains are able to become increasingly accurate at performing the task at hand. Youtube user CGP Grey, with the very informative and comical video below, talks about how these machines gain their intelligence, if we aren't directly programming the actions and "thoughts" they express.
AI is everywhere!
Source: Pixabay
So we have machines that pay attention to the videos we watch, and what sort of interactions we have online. We have bots that can recognize images, and trade stocks. Soon, cars will be driving themselves, and your laptop will be smarter than you are. I don't think these are inherently bad things, though many would likely argue that point with me.
Whether you like it or not, machine learning and artificial intelligence have become an integral part of society, and we will start using AI to not only automate many mundane tasks, but allow us to experience an intelligence we may never attain ourselves.
Thanks for reading! If you learned something new, consider Resteeming this post so your followers might learn something, too!
We built robots to do so many things, yet we haven't yet built ones to mine for us. I mean literally mine, as in precious metals and stuff, and then make stuff out of that stuff, so we don't have to do any work that we don't want to do. And also bots that will fix our bodies so we'll never die. So there's a lot to be done yet, I wouldn't start worrying about machine overlords just yet, as far as I'm concerned machines and algorithms can't do any of the important stuff yet, like cure cancer. When they do that, instead of serving me inane youtube videos, then I'll be impressed.
Btw, nice youtube videos!
Very interesting. I wrote a short blog in my early days on Steemit about the results of using a neural network to encode the film Blade Runner. I found out ebay uses neural networks and machine learning to scan posts for images of items that aren't allowed to be sold on the site. There are ways to tweak an image to fool the neural net though. Fascinating stuff!
It truly is, and yes, you can easily fool even the smartest of neural nets at the moment.
That second video shows a bot that's sure it sees a bee and not a three, but it actually just sees a dog in a bee suit :D Pretty soon though, computers are going to be WAY better at recognizing things like that then even we are, and we'll have no idea how they do it.
Skynet? ;-P
They got a ways at least before they are “smarter” then we are. Sure they can do things faster but when they can’t determine a cat from a dog or a cloud from a semi-truck they got some learning to do.
I know a couple of people working on AI. You would think it be all existing and fun. Heck no it sounds very boring and frustrating. Everyone wants to cheat beating the Turning test. I’ll be little more concerned when they beat it without any tricks involved repeatedly and there a few different ones of them doing it.
Your Post Has Been Featured on @Resteemable!
Feature any Steemit post using resteemit.com!
How It Works:
1. Take Any Steemit URL
2. Erase
https://3. Type
reGet Featured Instantly � Featured Posts are voted every 2.4hrs
Join the Curation Team Here | Vote Resteemable for Witness
This post has received a 0.98 % upvote from @morwhale thanks to: @thatsweeneyguy.