Creating artificial intelligence

Advancements in technology now afford scientists greater opportunities to explore space. In 2006, New Horizons spacecraft was launched on a mission to take images of Pluto. Flyover images from the spacecraft reveal a new perspective on Pluto.


Scientists at Caltech in Pasadena, California, are discovering new ways to give robots and drones brains of their own with the hope of helping humans in their everyday lives.

Here's the story.

Imagine a world where robots and drones are all around us, interacting with us, helping us, accompanying us, a world where robots and drones are not remotely controlled.

They work autonomously.

That's what scientists like Dr. Aaron Ames at the California Institute of Technology are working towards.

Caltech in Pasadena has just opened its brand-new Center for Autonomous Systems and Technologies, also known as CAST.

Dr. Ames and his fellow science brainiacs say their goal is to get robots and drones to partner with humans.

For example, what if you could have a robot take care of your aging parents who have health problems and live alone?

What if you could have robots and drones respond to natural disasters, like fires or hurricanes, to save people so you don't endanger the lives of rescuers in certain situations?

What if you could have a robotic prosthetic that works on its own if you are an amputee, or what if you're simply a couch potato and want a robot or drone to do all of your shopping for you?

This is really cool.

Yes, it is.

This is Cassie.

It's a walking robot built by Agility Robotics and newly arrived at Caltech, and the hope is on this robot to test autonomous-related concepts.

When you say autonomy, I mean, you're kind of saying this thing is going to have its own brain and work on its own.

That's right, all of engineering in one package.

You have to think about how it's designed mechanically, how the linkages come together to get the kind of behavior you want.

You have motors, right?

But those motors have to be controlled by processors, which then have to be controlled by other processors.

So at the core of all that, there's algorithms, and the algorithms represent the map that describes what walking is.

So you have mathematics, algorithms, computer science, electrical engineering, driving all the motors, how you design the motors and mechanical engineering the actual, you know, construction of the system, so it really brings together all these different disciplines of engineering in one single package.

It's going to be able to make its own decisions.

So we have to understand what that means by decisions, though.

You know, we don't mean that it's going to decide where to shop or not or whether to, you know, buy shoes from Amazon or not.

We mean just local decisions about how will it go forward, how does it handle the terrain that it's walking on, right?

You have robots that can do repetitive tasks very well but have difficulty with unstructured things.

You have artificial intelligence that can deal with structured things very well, like chess and Go.

How do we bring this together?

So how do we take capable bodies like this and put capable minds on them?

And importantly, how do we have those minds be able to reason in unstructured and uncertain environments?

And it's that that will be the key to autonomy down the line.

Taking the reins and walking the Caltech campus with Cassie was not what I expected.

Can I be honest with you?


It seriously freaks me out, though.

Why is that?

I don't know.

Just because it's something that's not human.

I don't really have to guide her.

I mean, she's practically walking herself.


She wants to pull away from you, right?

Well, that's the thing.

That's right.

It's as if she has a mind of her own, and she's walking on her own.

That's right.

And that's kind of freaky.

It is kind of freaky, and people, they think this is a result of artificial intelligence sometimes, and they think that it has a life of its own, but it doesn't.

This is a result of very clever and interesting mathematics' being implemented on the robot.

I design the brains behind the robots here at CAST.

Dr. Anima Anandkumar is a computer science professor at Caltech.

Anandkumar specializes in artificial intelligence and machine learning.

She helps program Cassie.

For me, like, artificial intelligence and machine learning is asking how we can take in all the data of the world and create knowledge.

In other words, help give Cassie intelligence, help get her her brain.

Anandkumar explains that Cassie is able to walk around the university with us because she has basically been given instructions on how to walk.

These instructions are the AI that has been programmed into the robot.

This artificial intelligence allows Cassie to know how to walk on certain kinds of surfaces, like the concrete here.

However, you see her struggle when we try to walk with her on uneven pavement or grass.

Scientists are working towards a day when they won't have to program specific instructions.

With improved processors and sensors, robots like Cassie could be able to take information from their environment and learn how to adapt and react on their own.

Full autonomy is indeed the challenge, right?

So that's why we don't aim for full autonomy right in the beginning.

What we do is, can we augment human and manual control with AI?

So in the beginning, they AI system is, you know, absorbing what the humans are doing, how they're controlling, and over time, hopefully, it learns better and better.

Scientists and researchers at Caltech are trying to give drones those same capabilities.

This is the largest drone-testing facility in the United States.

What you're seeing here at the lab is a giant wind tunnel.

This wind tunnel simulates all different types of weather patterns, things like wind gusts, high turbulence, even tornadoes.

The goal is to design a state-of-the-art drone system to operate autonomously in any type of weather condition.

Once the mission is defined, they have to solve every problem in between.

It means if they have to go from point A to point B, they have to learn and also resolve issues like going around the buildings or avoid people or making sure that they deliver the objects or parts that they're designed for.

One of the big problems with flying a drone is that when they encounter storms, like high wind around building, it has to have a brain.

Otherwise, it won't be able to reason and solve problems.

Drones are going to play a big role in the future of our society.

Think of drones that they have to go scout in dangerous areas, or they have to go save people or find people that need help.

In that respect, operators cannot be there all the time.

They have to be able to think.

It sounds like you're trying to turn a drone into a human.

Well, we are trying to make the drone to a partner for humans.

Partners for humans?

How funny is that, though?

What's your dog's name?


You have Dash.

We have Cassie.

There you go.