Ok, first the full disclosure. I'm a grad student who worked with both John Canny (the advisor involved) and Dan Resnik (the builder of this cool device). http://www.cs.berkeley.edu/~jfc/
The objections that I'm seeing seem to be of the following:
* "It's not real--the videos are fake."
No, it's real. The basic principle is this: if you take a table and shake it, stuff moves a little. If you shake it in a different direction, stuff moves in a different direction. Dan has figured out a pretty cool way of quantifying all this. He does a vector addition of several different shakes, and is able to therefore target the shaking.
* Why is it cool?
Well, first, because it's not obvious that it works. This is tricky math.
* Yes, but why THIS?
Because robot arms are a pain, and only manipulate one thing at a time, and they need a lot of elbow room, and a lot of motors, and they have to touch things. This requires four motors that JUST pulse in and out. It works on a flat table. Nothing touches except the table surface. This would be perfect for carefully-controlled environments like clean rooms (where you want to minimize the amount of stuff in the room), hazardous materials, and delicate objects. There's very little complexity, and because it's just a bunch of (tuned) vibrations, you could slap up a new one against a floor and it works.
The chess demo is just showing that one can comfortably manipulate a large number of items.
* Does anyone need to sort multiple things?
All the time. Factory floors separate out rejects from working models. Recycling centers separate cans from bottles. Usually, they hire people to sort the stinking messes apart, and they use clever special-built machines to separate metal (use a magnet) from glass (heavier than plastic when crushed) from paper. This sorts on a smooth--therefore easily cleanable--surface.
Dan now works for Siemens TTB, who are, among other things, very interested in small motors.
* My toy did this.
Yes, it did. But it did it in an extremely constrained way, and it probably took a really practiced flip of your wrist. And it probably did it in one dimension. (Dan has a little plastic train set that works on this principle).
* This eliminates human jobs
Not necessarily. It could work well in conjunction with a human job. Why do that annoying RSI-inducing reach/grab/sort when you can sit behind a desk, look at a video camera, and tick off the items on a screen? After all, image recognition isn't too good yet. The machine is responsible for the reach/grab/sort, and you don't have to wear a bunny suit.
Well, no. "Adobe" is pretty darn generic _for_houses_. If I started a house building company called "Abobe Huts, Inc" and then sued every time a tour guide pointed someone to a quaint adobe hut, I would get lose my shirt.
Adobe is very unique for computer software. (How many pieces of software had you heard of called "Adobe" before this compnay showed up?)
Um. Except that article was in the future tense ("someday, in the desert, she will lift large--over two ton--bricks"); this article is in the past ("in the desert, she has lifted large bricks.")
Saying that it is currently on an exp. curve doesn't mean that it will someday take over the universe. If the portion of their curve under examination matches an exponential curve, this is a fact.
But I like the statistics!
No, not really. If you turn in "more then a few phrases" of your own work, you are being dishonest.
A friend of mine once had a student turn in the same essay twice: "but it fit both requirements," the student said. Nevertheless, the question was to ask students to do research and to demonstrate a variety of knowledge. I'm not at all convinced that turning in the same essay twice counts.
If a professors term-paper assignment costs a student 15% of his time, then a professor should devote 15% of his time to grading the papers he receives.
Hm. As a student, it takes me about one hour per page to turn out a reasonable-quality essay.
As a TA, it takes me about one minute per page to grade the same essay. Which balances out at the sixty-person class. Most classes are smaller, but then again, most classes require a lot of preparation besides the grading itself: students dropping by to ask questions, emailing, composing reasonable answers.
Look: we're looking for good, coherently expressed ideas. Do you have a meaningful argument? Do you express your ideas clearly? It takes under a minute to read a page of a journal article; how long should it take us to read your essay?
As a TA at a Large Public Institution, we also used a cheating detector. Check the software at
http://www.cs.berkeley.edu/~aiken/moss.html
(other variants are at
http://www.findsame.org
http://www.plagiarism.org)
That system would return us a list of all programs that it--working conservatively--thought may include some degree of plagiarism.
I manually went through all code that it found was a good match, and removed those that seemed (to me, subjectively) to be coincidences.
There were very few of those. If the system had fingered you, chances are you really had matched it.
It was actually pretty clear which pairs of students had even worked together: they'd tend to solve problems in pretty much the same way, and their code would look different (different variable names, different choices of what precisely went inside or out of loops), even though their ordering of functions was the same and their variable declarations were similarly placed.
In contrast, students who had copied-and-pasted were obvious. Their code would jump out at us. Most coders had pretty specific styles, and when it varies between capitalization, function names, comment styles, indentation styles, and even (sometimes) the language of comments and variable names...
... well, there's no question that someone has added a snippet of code from someone else. Since the system would show us the someone else, we pretty much knew the culprit.
That said, we accused no one of plagiarism. It's too much of a pain--at most universities--to deal with the long chain of accusations and defenses. It just isn't worth the effort. So we knock their GPA down a fraction of a point and go on with our lives.
Which is disturbing, actually, because we found some students who had been known--among other TAs and professors--as cheaters. Who had been confronted semester after semester. Who had always gotten a few points off and gone on with their lives.
Slashdot Mirror
Ok, first the full disclosure. I'm a grad student who worked with both John Canny (the advisor involved) and Dan Resnik (the builder of this cool device). http://www.cs.berkeley.edu/~jfc/
The objections that I'm seeing seem to be of the following:
* "It's not real--the videos are fake."
No, it's real. The basic principle is this: if you take a table and shake it, stuff moves a little. If you shake it in a different direction, stuff moves in a different direction. Dan has figured out a pretty cool way of quantifying all this. He does a vector addition of several different shakes, and is able to therefore target the shaking.
* Why is it cool?
Well, first, because it's not obvious that it works. This is tricky math.
* Yes, but why THIS?
Because robot arms are a pain, and only manipulate one thing at a time, and they need a lot of elbow room, and a lot of motors, and they have to touch things. This requires four motors that JUST pulse in and out. It works on a flat table. Nothing touches except the table surface. This would be perfect for carefully-controlled environments like clean rooms (where you want to minimize the amount of stuff in the room), hazardous materials, and delicate objects. There's very little complexity, and because it's just a bunch of (tuned) vibrations, you could slap up a new one against a floor and it works.
The chess demo is just showing that one can comfortably manipulate a large number of items.
* Does anyone need to sort multiple things?
All the time. Factory floors separate out rejects from working models. Recycling centers separate cans from bottles. Usually, they hire people to sort the stinking messes apart, and they use clever special-built machines to separate metal (use a magnet) from glass (heavier than plastic when crushed) from paper. This sorts on a smooth--therefore easily cleanable--surface.
Dan now works for Siemens TTB, who are, among other things, very interested in small motors.
* My toy did this.
Yes, it did. But it did it in an extremely constrained way, and it probably took a really practiced flip of your wrist. And it probably did it in one dimension. (Dan has a little plastic train set that works on this principle).
* This eliminates human jobs
Not necessarily. It could work well in conjunction with a human job. Why do that annoying RSI-inducing reach/grab/sort when you can sit behind a desk, look at a video camera, and tick off the items on a screen? After all, image recognition isn't too good yet. The machine is responsible for the reach/grab/sort, and you don't have to wear a bunny suit.
Well, no. "Adobe" is pretty darn generic _for_houses_. If I started a house building company called "Abobe Huts, Inc" and then sued every time a tour guide pointed someone to a quaint adobe hut, I would get lose my shirt. Adobe is very unique for computer software. (How many pieces of software had you heard of called "Adobe" before this compnay showed up?)
Um. Except that article was in the future tense ("someday, in the desert, she will lift large--over two ton--bricks"); this article is in the past ("in the desert, she has lifted large bricks.")
Saying that it is currently on an exp. curve doesn't mean that it will someday take over the universe. If the portion of their curve under examination matches an exponential curve, this is a fact. But I like the statistics!
No, not really. If you turn in "more then a few phrases" of your own work, you are being dishonest. A friend of mine once had a student turn in the same essay twice: "but it fit both requirements," the student said. Nevertheless, the question was to ask students to do research and to demonstrate a variety of knowledge. I'm not at all convinced that turning in the same essay twice counts.
Hm. As a student, it takes me about one hour per page to turn out a reasonable-quality essay.
As a TA, it takes me about one minute per page to grade the same essay. Which balances out at the sixty-person class. Most classes are smaller, but then again, most classes require a lot of preparation besides the grading itself: students dropping by to ask questions, emailing, composing reasonable answers.
Look: we're looking for good, coherently expressed ideas. Do you have a meaningful argument? Do you express your ideas clearly? It takes under a minute to read a page of a journal article; how long should it take us to read your essay?
As a TA at a Large Public Institution, we also used a cheating detector. Check the software at
http://www.cs.berkeley.edu/~aiken/moss.html
(other variants are at
http://www.findsame.org
http://www.plagiarism.org)
That system would return us a list of all programs that it--working conservatively--thought may include some degree of plagiarism.
I manually went through all code that it found was a good match, and removed those that seemed (to me, subjectively) to be coincidences.
There were very few of those. If the system had fingered you, chances are you really had matched it.
It was actually pretty clear which pairs of students had even worked together: they'd tend to solve problems in pretty much the same way, and their code would look different (different variable names, different choices of what precisely went inside or out of loops), even though their ordering of functions was the same and their variable declarations were similarly placed.
In contrast, students who had copied-and-pasted were obvious. Their code would jump out at us. Most coders had pretty specific styles, and when it varies between capitalization, function names, comment styles, indentation styles, and even (sometimes) the language of comments and variable names...
... well, there's no question that someone has added a snippet of code from someone else. Since the system would show us the someone else, we pretty much knew the culprit.
That said, we accused no one of plagiarism. It's too much of a pain--at most universities--to deal with the long chain of accusations and defenses. It just isn't worth the effort. So we knock their GPA down a fraction of a point and go on with our lives.
Which is disturbing, actually, because we found some students who had been known--among other TAs and professors--as cheaters. Who had been confronted semester after semester. Who had always gotten a few points off and gone on with their lives.