Main content
Course: AI for education > Unit 1
Lesson 2: Video series: What is AI? | Code.orgHow neural networks work
Did you know that the networks that support the development of artificial intelligence are based off of the human nervous system? In this video, you'll learn how AI scientists built artificial neural networks that can gather information from various sources and synthesize them into an insight, and how additional inputs can be used to train an AI neural network.
Start learning at http://code.org/
Stay in touch with us!
• on Twitter https://twitter.com/codeorg
• on Facebook https://www.facebook.com/Code.org
• on Instagram https://instagram.com/codeorg
• on Tumblr https://blog.code.org
• on LinkedIn https://www.linkedin.com/company/code-org
• on Google+ https://google.com/+codeorg. Created by Code.org.
Start learning at http://code.org/
Stay in touch with us!
• on Twitter https://twitter.com/codeorg
• on Facebook https://www.facebook.com/Code.org
• on Instagram https://instagram.com/codeorg
• on Tumblr https://blog.code.org
• on LinkedIn https://www.linkedin.com/company/code-org
• on Google+ https://google.com/+codeorg. Created by Code.org.
Want to join the conversation?
- Man, I love this guy. #sofunny(2 votes)
- How does it know how much weight to add at first(1 vote)
- so the neural network actually is tremendos numbers of codes?(0 votes)
- What is the name of this website(0 votes)
- Do you mean the one which created these vids? In that case, it's code.org.(1 vote)
Video transcript
Hi! I'm Dion. I'm one of the creators
of Forethought AI. At Forethought, we build artificially intelligent tools that
people can use at work to be more productive. To make a learning machine, early computer
scientists looked for clues by studying other things that are good at learning,
and it turns out that nothing is better at learning than the human brain! Our brains
are made up of special cells called neurons. A neuron has two ends: input
signals enter in on one end, they're combined together inside the neuron,
and leave out the other end as a single output. All of the billions of neurons in your brain
are connected to each other, in what's called a biological neural network. It's how your brain
processes information and recognizes patterns. Early AI scientists decided to mimic human neurons
by making their own simple artificial neurons in software. Nothing fancy, just multiple signals
going in as inputs, passing through the neuron, and getting combined and processed by some
simple math into a new signal going out. It's a good start, but one
neuron alone doesn't do much. The full potential of this idea is only
unleashed when the artificial neurons are connected together to make an artificial
neural network. This is what allows computers to recognize images, drive cars,
and make some truly weird art. To see how a neuron works, let's build a movie
recommendation system, that uses critics reviews to guess how much you'll like a movie. Then,
we'll use your feedback to make the system better! Here are three movie critics: Ali, Bowie,
and Casey. Each one rates a movie anywhere from one to five stars. Now, let's
build a single artificial neuron. Each of the critics ratings enters on this side
as input, some calculations are done in here, and we get a single output. In
this case, it's a movie rating. Here's the first movie. Ali gives it one star,
Bowie gives it five, and Casey gives it a four star review. At first, the critics opinions all
carry the same weight, and are counted equally. The inputs enter, there's some basic
math, and out comes a recommendation. Now, let's watch the movie so
we can give it our own rating! Uh, okay. That was weird! Let's let's pretend you
really liked it, and gave it a five star rating. The rating you just provided is now used to
train the neuron. Based on your rating, the weight of each critic's opinion is recalculated.
Your rating is closer to that of Bowie and Casey, so their opinions get more weight. You
didn't agree with Ali's single star review, so that weight goes down. Now
let's train the neuron again. Here's another movie, and here are new
ratings from our critics. And this time, the neuron will give more weight to these two
ratings when calculating its recommendation. And here's the output! Now let's give it a watch. Well, at least that was
short! Let's give it a rating. Our new rating adjusts the weights again.
This process repeats over and over, until we've trained a system to know our preferences,
and recommend movies that we'll probably enjoy. In this example, there's just one neuron.
That's far more simplistic than most systems. Powerful neural networks have millions
of neurons arranged in layers. There are input layers, any number
of hidden layers, and output layers. The output of one layer of neurons, becomes
the input to the next layer, and so on. Many real world media music and shopping
recommendation systems work like this, using ratings for millions of everyday users in those neural networks. Everyone
has a hand in modifying the weights. Neural networks have so many other uses.
They're working behind the scenes on big problems, like growing healthier food,
predicting floods and forest fires, aiding wildlife conservation, and
even detecting and curing disease.