Presentación
Vídeo
Transcripción
Extracto de la transcripción automática del vídeo realizada por YouTube.
alright good morning everyone so welcome to a first talk of the day please give a hand for our speaker Jessica Jessica Hamrick who will be talking about games for science creating interactive psychology experiments in Python with pandas 3d thank you one moment
okay where it is my mouth it is so so yes I'm a graduate student at UC Berkeley and I'm gonna be talking to you about some research that I've done with my colleague Peter Battaglia who is a research scientist at MIT and unfortunately couldn't
be here at pycon but has been like a really integral part of this research he was my mentor while I was an undergraduate at MIT and actually i'll be talking about some experiments that he's done since i left MIT so to start out with first let's
talk a little bit about games so video games have become increasingly popular recently as of May 2012 apparently angry birds had been played a total of two hundred thousand yeah 200,000 years with an average of three hundred million minutes of playing time
every single day which is a pretty big number TV sense of how big it is 300 million minutes is approximately equal to five hundred and seventy years so that's five hundred and seventy years of playing time just for angry birds every single day that's
a huge amount of human effort and time that's being invested into this one game so wouldn't it be cool if we could leverage that popularity in order to do science like we could do so much science if we had that much manpower so that's the idea
behind this talk is doing you know creating fun games in order to do interesting science and to give you an outline of what I'll be talking to you about I'm first going to go over the research question that we're interested in I'll then tell
you what our hypothesis is that we think is going on and then I'll tell you about the experiments that we've run in order to test this hypothesis and the tools that we used to make those experiments and you know the the tools that we do use to make
these experiments they are not too difficult to use so I think anyone in this room would be able to pick them up fairly easily we have some demo code that's online you can go and look at it so you know as I'm going through the talk if this is interesting
to you you could probably do something pretty similar if you wanted um but ok so first of all what's the question that we're interested in so we are fundamentally interested in how people reason about the world and in particular how people reason about
the physical world so from the time that you get up in the morning until you go into work you've already interacted with the physical world in ways that no computer can even begin to rival so you've gone up you made breakfast you got dressed you put
things into your bag you drove to work a computer can't do any of those things well I guess maybe they can you know drive now with the Google self-driving car but a computer definitely can't do all of those things and definitely can't do it as
fluidly and as flexibly as people can and so you might think well maybe that's just because we have lots of experience but the world you know if you get up and make breakfast every morning so you just you know it's easy because you do it every day
but I think that our ability to reason about physics is much more flexible and much more powerful than that and to try to convince you of that I could show you a table with some blocks on it you've probably never seen this table before or this particular
configuration of blocks yet if I told you to imagine that you know maybe there's some kids running around and they bump into the table some blocks fall off you probably have a pretty strong idea of what color block is more likely to fall off so just shout
out what you think do you think it's red or yellow is more likely to fall off red okay right how about this one right okay this one all right and read this one maybe maybe yellow uh-huh how about this one yellow ok this one red right so you know just from
these few scenes you've never seen these before get it you all have a pretty you know strong opinion of what might happen and in general you agree with each other you know the level of agreement you know can vary from scene to scene but it seems like there's
pretty there's something pretty systematic and some really powerful flexible reasoning going on and so this is exactly the thing that we're interested in is how is it that we can reason about these types of scenes I think that's really interesting
so that's the question is how people reason about the physical world so then now I'm going to tell you about what our hypothesis is that's going on so the field that I man is called computational cognitive science and in computational cognitive
science we view the brain as a computer and we're interested in how the brain generates thought and behavior and in particular in computational cognitive science we are interested in how the brains software works so how the software produces behavior our
colleagues in computational neuroscience are interested in the hardware and that's more like the individual neurons and so our scientific approach is to come up with theories of how the minds software produces thought and behavior and then instantiate
those theories in a computer model so that might be a robot or an artificial intelligence but it's something that can produce behavior just like people can produce behavior and then we can compare the behavior that the model produces with the behavior
that people produce in order to test our theories and fine-tune them and come to a better understanding of how people think so how do we create a model of how people reason about physics well luckily for us there's already lots of software that does physics
so for example the physics engine in a video game like Angry Birds and moreover you know if we take a look at physics games this seems to offer us a really nice platform for studying how people reason about physics because it requires the same you know types
of reasoning faculties that you need to interact with the everyday world so this is sort of you know where we started what we were thinking about and we took this you know pretty far and what we hypothesize is that you actually have a physics engine in your
[ ... ]
Nota: se han omitido las otras 3.109 palabras de la transcripción completa para cumplir con las normas de «uso razonable» de YouTube.