Cookies   I display ads to cover the expenses. See the privacy policy for more information. You can keep or reject the ads.

Video thumbnail
So my principal interest is understanding
what goes on inside our heads.
And I'm convinced that one of the defining features of human
intelligence is that we can understand stories in a way
that other primates can't.
So we start with fairy tales, and we end up
with the cases of law school and business
school and medical school, and all throughout our education,
we're dealing with stories.
So believing as I do that stories are extremely
important, it's natural, then, to try
to build systems that understand stories
and that shed light on what the story understanding
process is all about.
So let me demonstrate this by showing you
a little bit of the Genesis system at work.
I will start with just how Genesis understands
ordinary English by typing in some very
simple sentences about things happening
in the physical world.
So we can say, for example, so there's
a description of a trajectory, and the Genesis system has
a variety of experts that understand
that sentence in terms of it being a trajectory that
goes along a path and ends up in a particular place.
We can add a little complexity by saying
the bird flew to the top of a tree, oh,
because a cat appeared.
So now we have a transition and a cause
to complement the movement along our trajectory.
And of course, we are not limited
to things that happen in the physical world.
We can move into an abstract world.
That's movement just like a cat moving to the top of a tree,
but this time it's in a political space
moving toward an abstract destination rather
than a physical one.
But still, we see the trajectory apparatus light up,
and now we've added the goal.
I can modify that a little bit and say
the president asked Iraq to move toward democracy,
and this time, we light up the persuasion box.
And if I say forced instead, that lights up coercion.
So now, we're going to move on and have
a look at a very simply stated version of Shakespeare's
play, Macbeth.
So the system is reading away, and everything
that you see inside of a white box
is explicitly told in the story, but all this stuff
in the gray boxes has been inferred by common sense rules.
For example, there's one here that says
that Macbeth murders Duncan.
So Duncan naturally becomes dead,
but that's not all that we can do.
We can look for more extended patterns in this thing
we call the elaboration graph.
For example, we can look for the revenge pattern,
and we find it there.
Macbeth harms Macduff that means that Macbeth angers Macduff.
That leads to Macduff killing Macbeth, so Macduff in the end
harms Macbeth.
So when you put all that together, it's a revenge.
Now, we can zoom back out, and see even a more complicated
pattern there.
There's the empiric victory.
That's because Macbeth does something that initially leads
to him being happier, he becomes king,
but eventually, that leads to Macbeth
being very unhappy because he gets murdered.