In other words, start with an idea that's not patentable because it's abstract. Merely specifying its implementation on a computer doesn't make the original idea more concrete or more patentable. Therefore the claim on the idea is invalid, therefore other claims based on that one are too. So what's left?
Computer implementations that are not "generic". A specific computer implementation of an abstract idea might be patentable. But a generic one (i.e., one that amounts to <abstract idea> + "on a computer") is not.
If the application to a computer is itself both novel enough and concrete enough, then that might still be considered patentable. Is that possible?
Exactly, though, really, the nature of general purpose digital computers is such that, if its not calling for novel hardware, that is probably indistinguishable from a generic computer implementation of a more concrete version of the abstract idea, which concrete version would probably itself be patentable independent of the computer implementation.
Take, for example, IBM's patents on alias analysis. You can't patent the abstract concept of alias analysis, even if you describe how to do it in very generic terms on a computer. However, you can probably patent Steensgard's specific, near-linear-time algorithm for doing so: http://www.cs.utexas.edu/users/pingali/CS380C/2007fa/papers/....
Why wouldn't Steensgaard's algorithm be an abstract idea?
There is a difference, since unlike hedging (Bilski) and clearing house (Alice Corp) it is a novel abstract idea. But it is still an abstract idea, and logically, should not be patentable.
Stålmarck's algorithm is a similar case. Can't argue it is not novel, but 100% abstract. Parabix is another.
I tend to think of these things as an engineer, rather than a mathematician. A mathematician will say that all algorithms are abstract ideas. I tend to think that "alias analysis" is the abstract idea, while Steensgaard's algorithm is a specific, concrete way of implementing the idea with desirable properties and trade-offs.
Then, are mergesort and quicksort both patentable as specific, concrete ways of implementing the idea of sorting with desirable properties and trade-offs?
Comments
Computer implementations that are not "generic". A specific computer implementation of an abstract idea might be patentable. But a generic one (i.e., one that amounts to <abstract idea> + "on a computer") is not.
Exactly, though, really, the nature of general purpose digital computers is such that, if its not calling for novel hardware, that is probably indistinguishable from a generic computer implementation of a more concrete version of the abstract idea, which concrete version would probably itself be patentable independent of the computer implementation.
Take, for example, IBM's patents on alias analysis. You can't patent the abstract concept of alias analysis, even if you describe how to do it in very generic terms on a computer. However, you can probably patent Steensgard's specific, near-linear-time algorithm for doing so: http://www.cs.utexas.edu/users/pingali/CS380C/2007fa/papers/....
Why wouldn't Steensgaard's algorithm be an abstract idea?
There is a difference, since unlike hedging (Bilski) and clearing house (Alice Corp) it is a novel abstract idea. But it is still an abstract idea, and logically, should not be patentable.
Stålmarck's algorithm is a similar case. Can't argue it is not novel, but 100% abstract. Parabix is another.
I tend to think of these things as an engineer, rather than a mathematician. A mathematician will say that all algorithms are abstract ideas. I tend to think that "alias analysis" is the abstract idea, while Steensgaard's algorithm is a specific, concrete way of implementing the idea with desirable properties and trade-offs.
Then, are mergesort and quicksort both patentable as specific, concrete ways of implementing the idea of sorting with desirable properties and trade-offs?
This is a bad example, because the actual steensgaard patents are highly generic.
But your general point holds :)