Galileo, a computer model that is able to estimate how heavy the objects are, so it can predict what they'll do in other situations. Credits: news.mit.edu
Galileo, a new program, can predict how things will move, just like we do.
Yes, we humans are pretty good at predicting the future. Not really the future, but small things: like how something will move in different conditions under some forces.
Researchers are trying to give computers that ability.
What are we talking about?
Imagine you have a heavy block and a rubber ball at the top of a ramp.
What happens if you just let go?
Of course, you know the ball will roll down the hill and, even if the block will start moving, it won’t be as faster as the rubber ball. You know that round things roll, and things with edges generally don’t.
That’s just physics, but you don’t have to have a physics background to make the guess, you just know.
How do you know? Just because of your experiences: probably when you was young, you played with balls and block.
But computers generally haven’t play a lot with balls and block, so they don’t learn how objects interact with the world. Until now.
A group of researchers at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) have developed a computer model called Galileo that is able to watch videos of different objects interacting and to estimate how heavy the objects are, so it can predict what they’ll do in other situations.
“This is just the first step in imbuing computers with a deeper understanding of dynamic scenes as they unfold ” postdoctoral researcher Ilker Yildirim, co-author of the research, said.
How they did it
Yildirim and his co-authors first showed Galileo 150 videos of different objects in different situations so it can store a complete database with all of the information given in the videos, then added some human intuition too. Or, rather, computer intuition.
They linked Galileo with Bullet, computer software used by video games and movies as a ‘physics engine’ capable of making animated graphics look incredibly real by simulating how physics works in the real world.
Then they added deep learning that allowed Galilleo to learn from its previous experiences, just like humans.
Human vs. Galileo
They tested it against people comparing the human predictions and Galileo’s predictions.
They found that humans and the computer had very similar predictions.
You can test it in this website created by CSAIL.
Then?
Next Hop will be work with Galileo on more complicated predictions involving fluids or springs, and eventually getting to a point where it can make predictions in the natural world even faster than we can.
“Imagine a robot that can readily adapt to an extreme physical event like a tornado or an earthquake. Ultimately, our goal is to create flexible models that can assist humans in settings like that, where there is significant uncertainty.”
