4 ms·

His answer on why he does not believe in Computer Science is pretty convincing. What I would love to have happened is a professor telling me that what I am goin

by why-el 15y ago

His answer on why he does not believe in Computer Science is pretty convincing. What I would love to have happened is a professor telling me that what I am going to be learning is not exactly science. Nobody did, so I spent a couple of years trying to scientify things.

mistercow 15y ago
I've heard others give the same argument, but I don't find it very compelling. There's no rule that says that science cannot be applied to purely human endeavors. We would not say that sociology is not a science, even though it is a study of a purely human concept. Science is simply a method of increasing your knowledge of a system. It doesn't matter if that system is the bare physical universe, the chemical interactions within a cell, or the flow of data and logic in a computer program. One unambiguous application of the scientific method to computing is in debugging. Your program behaves unexpectedly – aha! a new phenomenon. The lateral thinking hardware in your brain engages as you tinker around, reproducing the bug and noticing the conditions in which it happens. You gather all of this together with your existing knowledge of the system, and come up with an explanation for the bug – a hypothesis! Then you try to falsify your hypothesis. You might try to falsify it by watching a value in the debugger, for example. Or you might modify your program and see if the bug persists or changes its behavior. You aren't usually recording all of this on paper, because in this case you don't care much about presenting your findings to others. Others will be satisfied if the bug is gone. And of course, you probably aren't thinking of the words "hypothesis" and "experiment". But that's what you did, and that's why you figured out the bug. (On the other hand, some programmers try to debug their programs by poking and picking randomly at their code until they get it working. I would urge them to try adopting a scientific approach. It's not just a way to satisfy a teacher in a chemistry lab, as it turns out.)
- zerohp 15y ago
  Actually, people still say that sociology, and other social "sciences" are not science. You're assuming that the scientific method is science, but its more than that. Many modern fields of endeavor call themselves sciences mostly because of the success of science. They only superficially emulate science. This includes Computer Science, which is really mathematics and engineering. Feynman himself speaks about it: http://www.youtube.com/watch?v=IaO69CF5mbY http://www.youtube.com/watch?v=IaO69CF5mbY
  - mistercow 15y ago
    Feynman is talking there about things which call themselves science but do not actually correctly follow the scientific method. The distinction between "the scientific method" and "science" is meaningless, and the claim that CS does not qualify as a science because it only deals with something manmade is little more than elitism.
karpathy 15y ago
Computer Science as a field has evolved since then, and I doubt he would insist on the same view today. It is very common for Computer Scientists to work in close collaborations with e.g. biologists, physicists, etc. to study natural phenomena. Some branches of Computer Science, I'd agree, are more artificial (Compilers, Programming Languages, Networking, etc... ) and less directly about nature. These branches were mostly synonymous with Computer Science in those times, and I believe these are what Feynman was also thinking about.
- dlo 15y ago
  In the situation you are describing, computers are used as tools in the sciences of biology and physics. Just because something is used to aid another science does not make itself a science.
  - karpathy 15y ago
    That's false. For example, I've done some work on computational models of the visual cortex. Researchers in this area have in the past made testable predictions about the structure of the visual cortex regarding for example, distributions of synapses, types of neurons and their function, etc. All this done based on computational arguments about time/space complexity, etc. The brain is basically a computing device, it makes perfect sense that it should be analyzed as such.
tikhonj 15y ago
That's funny because one of the first things my first CS professor said (paraphrased): computer science is very poorly named--it isn't a science and it isn't about computers.
- why-el 15y ago
  I was not that lucky. But I later watched a couple of videos from SICP and it made sense. :)

by why-el 15y ago

mistercow 15y ago

I've heard others give the same argument, but I don't find it very compelling. There's no rule that says that science cannot be applied to purely human endeavors. We would not say that sociology is not a science, even though it is a study of a purely human concept. Science is simply a method of increasing your knowledge of a system. It doesn't matter if that system is the bare physical universe, the chemical interactions within a cell, or the flow of data and logic in a computer program. One unambiguous application of the scientific method to computing is in debugging. Your program behaves unexpectedly – aha! a new phenomenon. The lateral thinking hardware in your brain engages as you tinker around, reproducing the bug and noticing the conditions in which it happens. You gather all of this together with your existing knowledge of the system, and come up with an explanation for the bug – a hypothesis! Then you try to falsify your hypothesis. You might try to falsify it by watching a value in the debugger, for example. Or you might modify your program and see if the bug persists or changes its behavior. You aren't usually recording all of this on paper, because in this case you don't care much about presenting your findings to others. Others will be satisfied if the bug is gone. And of course, you probably aren't thinking of the words "hypothesis" and "experiment". But that's what you did, and that's why you figured out the bug. (On the other hand, some programmers try to debug their programs by poking and picking randomly at their code until they get it working. I would urge them to try adopting a scientific approach. It's not just a way to satisfy a teacher in a chemistry lab, as it turns out.)

zerohp 15y ago
Actually, people still say that sociology, and other social "sciences" are not science. You're assuming that the scientific method is science, but its more than that. Many modern fields of endeavor call themselves sciences mostly because of the success of science. They only superficially emulate science. This includes Computer Science, which is really mathematics and engineering. Feynman himself speaks about it: http://www.youtube.com/watch?v=IaO69CF5mbY http://www.youtube.com/watch?v=IaO69CF5mbY
- mistercow 15y ago
  Feynman is talking there about things which call themselves science but do not actually correctly follow the scientific method. The distinction between "the scientific method" and "science" is meaningless, and the claim that CS does not qualify as a science because it only deals with something manmade is little more than elitism.

karpathy 15y ago

Computer Science as a field has evolved since then, and I doubt he would insist on the same view today. It is very common for Computer Scientists to work in close collaborations with e.g. biologists, physicists, etc. to study natural phenomena. Some branches of Computer Science, I'd agree, are more artificial (Compilers, Programming Languages, Networking, etc... ) and less directly about nature. These branches were mostly synonymous with Computer Science in those times, and I believe these are what Feynman was also thinking about.

dlo 15y ago
In the situation you are describing, computers are used as tools in the sciences of biology and physics. Just because something is used to aid another science does not make itself a science.
- karpathy 15y ago
  That's false. For example, I've done some work on computational models of the visual cortex. Researchers in this area have in the past made testable predictions about the structure of the visual cortex regarding for example, distributions of synapses, types of neurons and their function, etc. All this done based on computational arguments about time/space complexity, etc. The brain is basically a computing device, it makes perfect sense that it should be analyzed as such.

tikhonj 15y ago

That's funny because one of the first things my first CS professor said (paraphrased): computer science is very poorly named--it isn't a science and it isn't about computers.

why-el 15y ago
I was not that lucky. But I later watched a couple of videos from SICP and it made sense. :)