Tip:
Highlight text to annotate it
X
I started as an experimental psychologist
and I was not interested in robots, per say.
I was interested in the brain.
Over the years though I came to the realization that
I wanted to get a better grasp
and better understand how biological intelligence works
is to synthesize that intelligence in computer programs.
And look how it behaves.
We believe that that will give us intuition
of actually how the brain works.
So NASA came to us with a very tough problem.
NASA wants to create robots that are increasingly autonomous
that can be sent to space
they don't require a lot of sensors
but they can still perform efficient navigation, for instance,
without requiring GPS.
So rather than telling the Mars rover where to go
exactly, all the time,
NASA and their affiliates are trying to get the rover to
increasingly drive itself
so that they can give high level instructions.
This is especially useful for kind of driving into
caves where you can't always send instructions to the rover.
You want it to explore these lava tubes or caves
where you can't communicate with it and you want it to come back up afterwards.
So it turns out the biological intelligence is really, really good
at performing all these operations without the need of fancy sensors or GPS signals.
And what we did was successfully demonstrated that
a robot can navigate, orient itself,
recognize things in the environment
without needing a lot of expensive sensors.
So the way a camera sees if very different than what we see.
It's just a flat, set of numbers,
just like this grid of numbers,
and somehow you have to turn that grid of numbers into something intelligent.
We have developed an object recognition system called Cog Eye.
And basically what it does is
if you give it an image
it will look at the important objects in the image
and look at different views of those objects
and it will learn to recognize them
so that if given new views
it can classify them properly.
In a sense you feel like you are trying to recreate a life
or at least a very, very narrow slice of life,
which is biological intelligence,
and it's a gargantuan effort even to simulate something really small,
say a small network of neuron individual system.
It makes you really appreciate
the complexity of biological intelligence.