“Robot” is a nonsense category
The Communications of the ACM recently ran an article titled, “How can we trust a robot?”
Thinking about the article led me to realize that the category "robot" is itself a piece of nonsense, drawn from science fiction, and having no basis in computer science.
We exist in a world in which computer programs control many real world outcomes. Often, those programs direct the operation of physical peripherals to achieve those outcomes. A payroll program that prints checks directs the operations of a printer. Is it therefore a "robot"? Should we ask the question, “How can we trust a payroll program?”
Well, of course we should, but not because it is some special entity called a "robot," but because this program will determine how much employees get paid, and if the program contains bugs, they will get paid the wrong amount. And whether we should trust it depends not on whether it conforms to "social norms," as the ACM article contends of "robots," but on whether it is written correctly.
And so it is with programs that control self-driving cars, or heart monitors, or "robotic" arms in an assembly line: we should trust these programs to the extent we believe they are written correctly to do what we want them to do, and distrust them to the extent that we suspect they contain bugs.
A program that controls a little cart driving around an office complex (as some student has recently been directing past my office) is not, in computer science terms, conceptually any different than a program controlling the number printed on an employee paycheck. Each should only be trusted to the degree we believe the program correctly implements the algorithm we want it to implement, and dis-trusted to the extent we suspect it fails to do so.
NOTE: No doubt, rob would like to point out that "robot" has an entry in Webster's Dictionary.
Thinking about the article led me to realize that the category "robot" is itself a piece of nonsense, drawn from science fiction, and having no basis in computer science.
We exist in a world in which computer programs control many real world outcomes. Often, those programs direct the operation of physical peripherals to achieve those outcomes. A payroll program that prints checks directs the operations of a printer. Is it therefore a "robot"? Should we ask the question, “How can we trust a payroll program?”
Well, of course we should, but not because it is some special entity called a "robot," but because this program will determine how much employees get paid, and if the program contains bugs, they will get paid the wrong amount. And whether we should trust it depends not on whether it conforms to "social norms," as the ACM article contends of "robots," but on whether it is written correctly.
And so it is with programs that control self-driving cars, or heart monitors, or "robotic" arms in an assembly line: we should trust these programs to the extent we believe they are written correctly to do what we want them to do, and distrust them to the extent that we suspect they contain bugs.
A program that controls a little cart driving around an office complex (as some student has recently been directing past my office) is not, in computer science terms, conceptually any different than a program controlling the number printed on an employee paycheck. Each should only be trusted to the degree we believe the program correctly implements the algorithm we want it to implement, and dis-trusted to the extent we suspect it fails to do so.
NOTE: No doubt, rob would like to point out that "robot" has an entry in Webster's Dictionary.
For the record: I agree with the main points you make in this post. I had not until now realized that you were in yesterday's post similarly redefining the term "beating the odds" away from the normal dictionary definition, or I wouldn't have brought it up.
ReplyDeleteRob,
DeleteYou're misunderstanding. Gene is saying the common usage is wrong. You're saying "but the common usage is..."
I agree that is what Gene is saying here. I had missed that in the earlier post Gene was similarly criticizing the normal usages of terms like "beating the odds" . In his view they are "incoherent". I disagree , but at least see his point that pointing out that this was the normal usage was irrelevant to his argument.
DeleteGotcha. :)
Delete