Why Google Gave Up on Boston Dynamics

In 2017, Google did something that

surprised everyone. They sold Boston

Dynamics. They were the most famous

robotics company on Earth. Viral videos,

millions of fans, and decades ahead of

the competition.

But behind the hype, they were a company

in crisis. They'd built the most

advanced machines ever, and nobody could

figure out what they were actually

useful for. So, why did Google give up

on them? To understand why the world's

best robots almost failed, we have to go

back to the very beginning.

Meet our main character, Marc Raibert.

In 1980, Raibert was an MIT professor

who realized that the way we were

building robots was fundamentally

broken. At the time, robots worked by

basically following a very precise

script. They could follow instructions,

but they couldn't adapt. Raibert's

breakthrough was something called

dynamic locomotion.

Instead of pre-programming every move,

he built robots that behaved more like

animals. They didn't think about

walking, they just reacted to it. Every

leg had sensors measuring lean, speed,

and force. If the robot starts to tip,

it didn't need to wait for a command

from a computer. It would just adjust

and stay balanced on its own. And

suddenly, these robots didn't look stiff

or controlled anymore. They looked

alive. In 1992, Raibert developed this

research into his own company, Boston

Dynamics. But you might be surprised to

hear that for the first decade, they

really weren't a robotics company at

all. They made their money through

government contracts, building software

like DI-Guy, which simulated human

movement for military training.

The robots were still just a side

project in the back of the lab. The

technology was finally there, but

Raibert still faced the one problem that

would haunt the company for the next 20

years.

What do you actually do with a robot

that can walk?

Fast forward to 2003. DARPA, the branch

of the Pentagon that funds high-risk

futuristic technology, decided it was

time to take robots out of the lab. They

had a specific problem. More than half

of the Earth's landmass is impossible to

reach with a wheeled or tracked vehicle.

If you want to carry gear over a steep

mountain or through a thick forest, you

need legs.

At the time, Boston Dynamics had the

most advanced robotic legs in the world.

So, they got to work. The result was Big

Dog. If you were on YouTube around 2008,

you probably remember this thing.

Instead of falling over like every other

robot would, Big Dog was able to recover

and stay standing. It didn't look like a

machine anymore, it looked like an

animal.

DARPA was impressed enough to give

Boston Dynamics a $33 million contract

to build a battlefield version called

the LS3. It was a beast. It could carry

400 lbs of gear and hike through mud or

snow at an incline.

On paper, it was exactly what the

military wanted. But in the real world,

it had one massive problem. To get that

much power and range, they had to use a

gas engine. And that engine was loud.

In a combat [snorts] zone where you're

trying to stay hidden, the LS3 was

basically a giant alarm bell for the

enemy.

So, the military eventually walked away,

saying they couldn't use a robot that

gave away their position from half a

mile away.

After years of development and millions

of dollars, Boston Dynamics had built a

masterpiece that they couldn't sell.

They had the technology, but still no

customer. So, instead of trying to fix

the noise problem, they decided to try

something even harder.

Building a humanoid.

Now, building a robot with four legs is

hard, but building one with two that

walks like a human is even worse.

With four legs, you have a stable base.

But when we walk, we're basically in a

constant state of falling forward and

catching ourselves. Their first attempt

was Petman.

It was a high-tech test dummy used to

see how military chemical suits and

camouflage held up while moving.

It could walk on a treadmill, but it had

to be held up by a huge overhead crane

and a bunch of cables. It proved they

could mimic human movement in a robot,

but it wasn't a standalone product yet.

That changed in 2013 with Atlas.

Atlas, once again, was built for a DARPA

challenge. This time, a challenge based

on the Fukushima nuclear disaster. The

idea was to build a robot that could go

into areas too dangerous for people, but

still use human tools, climb stairs, and

open doors.

>> [music]

>> And Atlas could actually do it. It was 6

ft tall and weighed over 300 lbs, and

there was a catch. While the viral

videos made it look like the future had

arrived, the reality was still pretty

messy. Atlas was incredibly expensive.

It used high-pressure hydraulics that

would leak, and it still needed a

massive leash or power cable to stay

running for more than a few minutes.

Boston Dynamics didn't even win the

DARPA challenge. Another team with a

slower, more practical robot did.

But once again, Marc Raibert had pushed

the boundaries of what was possible. But

he still hadn't built a product.

And this is where Google entered the

picture. Google, the internet search

engine, is entering a new frontier. It's

bought a cutting-edge robot technology

company. In 2013, Google started buying

up robotics companies. They were

building a new division led by Andy

Rubin, the guy who created Android.

Robots could eventually be the future of

labor. So, Google went on a spending

spree, buying Boston Dynamics and seven

other robotic startups all at once.

Google had money, and Boston Dynamics

had the best hardware on Earth.

But in reality, the two companies had

very different goals.

Google is a software company. They're

used to building apps and code that can

be updated in a day. They wanted a

product that could actually ship and be

put to work in a year or two. But Boston

Dynamics was not there yet. They were

still focused on research.

They were building one-of-a-kind

prototypes that cost millions of dollars

each and needed a full team of engineers

just to stay running.

By 2016, that conflict was clear. Andy

Rubin had left Google, and the company

realized they were pouring money into a

project that wasn't going to make a

profit anytime soon.

So, in 2017, they decided to cut their

losses and put Boston Dynamics up for

sale. It was a huge blow. If the biggest

tech company in the world couldn't

figure out how to make money with these

robots, who could?

The writing was on the wall. Marc

Raibert and his crew might soon be out

of a job.

But then, a Japanese company called

SoftBank stepped in. SoftBank wanted

products, and they knew how to get

Boston Dynamics there.

No more building prototypes just for the

DARPA challenges or potential government

contracts. Now, they had to build

something people would actually want to

use.

That shift had already quietly started a

couple years earlier, when in 2015,

Boston Dynamics introduced a fully

electric Spot.

It might not seem like it, but this was

a huge turning point. By switching from

gas engines and hydraulics to electric

motors and batteries, the robot became

much quieter and easier to manage. For

the first time, they had a machine that