5 ms·

Good thing algorithms aren't patented and instead it's the application of the algorithm that gets the grant. Most are written as "a method for performing X by d

by deelowe 14y ago

Good thing algorithms aren't patented and instead it's the application of the algorithm that gets the grant. Most are written as "a method for performing X by doing Y." IMO, the issue with software patents isn't that businesses patent algorithms or code(which they don't do). Instead, it's that technology moves too fast for the monopoly time that's granted. When it comes to software, what's innovative 5 years ago is common place and often a commodity today.

Tycho 14y ago
Maybe there should be some burden of proof that many people have tried to solve the problem at hand but not come up with the solution in the patent application. For instance, if you invent a new sorting algorithm today, it would be reasonable to patent it; whereas if you invent a new voice activation scheme for a smartphone, it would be unreasonable to patent it because the whole field is relatively new.
- quahada 14y ago
  This touches upon the biggest problem I have with the patent system: The bar is very low on the "novel and non-obvious" criteria. If that were used properly, there would be fewer patents, and they would be higher quality.
- ScottBurson 14y ago
  Yes, yes, yes!! I think this is the smallest change that could be made to the current system to make it sane. The burden of proof needs to be on the patent holder to show, by a preponderance of the evidence, that they invented something nonobvious. That's a pretty low bar, but would require them to present some evidence, which they currently don't have to do.
  - boas 14y ago
    In my limited experience with both successful and unsuccessful patent applications, the first response from the patent office is often a "nonfinal rejection" due to obviousness (section 103 rejection). In response, you should present evidence that the invention is not obvious. Here are examples of types of evidence that the USPTO finds compelling: - commercial success (not due to better marketing). - failed attempts by other people to solve a longstanding problem. - awards or publications in the scientific literature. - the invention involves a combination of components that produce an unexpected result.
    - ScottBurson 14y ago
      Considering some of the patents I have seen, I would like to know what evidence of nonobviousness could have been presented for them. Perhaps one useful change would be to require that evidence to be part of the public record. In any case, as I understand, once the patent has been issued, the courts generally defer to the PTO on the obviousness question. I think plaintiffs in infringement cases should have to demonstrate nonobviousness again in order to proceed. While it would be better for the PTO to be more of a stickler on this point, the fact is that we don't give them the kind of resources they would need to do as thorough a job as they should do on every patent. For example, the PTO could maintain a team of engineers specifically for obviousness testing. When an application comes in, they would pose the problem the patent claims to solve to a selected panel of maybe 4 or 5 of them, who would have a couple of days to sketch some possible solutions. If they come up with the same idea as in the patent, that should pretty much kill it. But as I say, we don't spend enough money on the PTO for them to do that. Since we don't, having the courts re-evaluate each patent on this point seems the only solution.
- gcr 14y ago
  That strikes me as exactly backwards. The current system exists specifically to give inventors incentive to explore new fields. Having a patent over a completely new system means that as the one to birth that system, you get to have some control over it for a short while. That was the original intent, anyway.
baddox 14y ago
What's the difference between "a method for sorting an array of integers by doing <steps of Quicksort>" and simply patenting Quicksort? Or do you mean something more specific to a particular product, like "a method for sorting a list of restaurant recommendations by doing <steps of some machine learning algorithm>"? Either way, all I'm seeing is algorithms.
- JoshTriplett 14y ago
  Because you can't directly patent an algorithm, software patents use all sorts of fascinating workarounds to pretend like they've patented something else instead. A typical software patent will include the same patent claims written several different ways, based on legal precedents used with software patents in the past. A patent on an algorithm for fooing bars via steps A, B, and C, with the additional claim of doing D while fooing, might have one set of claims: 1. A method for fooing bars, comprising A, B, and C. 2. A method in accordance with claim 1, wherein said fooing further comprises doing D. 3. A system for fooing bars, comprising A, B, and C. 4. A system in accordance with claim 3, wherein said fooing further comprises doing D. 5. A computer program product for fooing bars, comprising A, B, and C. 6. A computer program product in accordance with claim 5, wherein said fooing further comprises doing D. This covers several different legal precedents; even if "method" gets invalidated as an approach, the patent may still work under the "system" or "computer program product" theories. This gets taken to a ridiculous extreme; many such patents describe the computer program product as a physical object consisting of bits on a medium such as a CD or disk, and include as a figure an illustration of such a medium.
  - diminoten 14y ago
    It's not pretending, they actually are patenting the method rather than the code itself. It's a critical distinction, as code is actually copywritten. There's a huge difference between design and implementation.
- hackinthebochs 14y ago
  (I'm no expert at software patents so take what follows with a healthy dose of NaCl) It seems to me the difference is a question of "real world implications". For sorting integers using quicksort, yes that's just a mathematical algorithm: an algorithm that performs operations on mathematical entities. The moment you apply that algorithm to something that affects the real world, that's when it becomes patentable. E.g. sorting restaurant recommendations by applying quicksort to the individual rating. This is no longer simply a mathematical algorithm; the entities being manipulated have a 1:1 correspondence with entities in the real world. It is recognizing the usefulness of this particular correspondence between mathematical entities and real world results that is patentable.
darkarmani 14y ago
> it's the application of the algorithm that gets the grant And that should simply be copyright for the expression of the algorithm.
- chii 14y ago
  no, the application would be the actual app. i would comply with copyright law if i simply wrote another app, whereas i will still violate the patent. the problem i have with algorithmic parents is that computer algorithms trend tend to be quite composable. other patents in other field aren't, e.g. pharma parents. when one algorithm could be such fundamental to a large amount of different apps, a patent on it gives the patent holder way to much than they deserve.
  - darkarmani 14y ago
    I know the law, but I'm saying that we already have protection for expression on an algorithm -- i don't think we need to patent an app that is simply an implementation of an algorithm.

by deelowe 14y ago

Tycho 14y ago

Maybe there should be some burden of proof that many people have tried to solve the problem at hand but not come up with the solution in the patent application. For instance, if you invent a new sorting algorithm today, it would be reasonable to patent it; whereas if you invent a new voice activation scheme for a smartphone, it would be unreasonable to patent it because the whole field is relatively new.

quahada 14y ago
This touches upon the biggest problem I have with the patent system: The bar is very low on the "novel and non-obvious" criteria. If that were used properly, there would be fewer patents, and they would be higher quality.
ScottBurson 14y ago
Yes, yes, yes!! I think this is the smallest change that could be made to the current system to make it sane. The burden of proof needs to be on the patent holder to show, by a preponderance of the evidence, that they invented something nonobvious. That's a pretty low bar, but would require them to present some evidence, which they currently don't have to do.
- boas 14y ago
  In my limited experience with both successful and unsuccessful patent applications, the first response from the patent office is often a "nonfinal rejection" due to obviousness (section 103 rejection). In response, you should present evidence that the invention is not obvious. Here are examples of types of evidence that the USPTO finds compelling: - commercial success (not due to better marketing). - failed attempts by other people to solve a longstanding problem. - awards or publications in the scientific literature. - the invention involves a combination of components that produce an unexpected result.
  - ScottBurson 14y ago
    Considering some of the patents I have seen, I would like to know what evidence of nonobviousness could have been presented for them. Perhaps one useful change would be to require that evidence to be part of the public record. In any case, as I understand, once the patent has been issued, the courts generally defer to the PTO on the obviousness question. I think plaintiffs in infringement cases should have to demonstrate nonobviousness again in order to proceed. While it would be better for the PTO to be more of a stickler on this point, the fact is that we don't give them the kind of resources they would need to do as thorough a job as they should do on every patent. For example, the PTO could maintain a team of engineers specifically for obviousness testing. When an application comes in, they would pose the problem the patent claims to solve to a selected panel of maybe 4 or 5 of them, who would have a couple of days to sketch some possible solutions. If they come up with the same idea as in the patent, that should pretty much kill it. But as I say, we don't spend enough money on the PTO for them to do that. Since we don't, having the courts re-evaluate each patent on this point seems the only solution.
gcr 14y ago
That strikes me as exactly backwards. The current system exists specifically to give inventors incentive to explore new fields. Having a patent over a completely new system means that as the one to birth that system, you get to have some control over it for a short while. That was the original intent, anyway.

baddox 14y ago

What's the difference between "a method for sorting an array of integers by doing <steps of Quicksort>" and simply patenting Quicksort? Or do you mean something more specific to a particular product, like "a method for sorting a list of restaurant recommendations by doing <steps of some machine learning algorithm>"? Either way, all I'm seeing is algorithms.

JoshTriplett 14y ago
Because you can't directly patent an algorithm, software patents use all sorts of fascinating workarounds to pretend like they've patented something else instead. A typical software patent will include the same patent claims written several different ways, based on legal precedents used with software patents in the past. A patent on an algorithm for fooing bars via steps A, B, and C, with the additional claim of doing D while fooing, might have one set of claims: 1. A method for fooing bars, comprising A, B, and C. 2. A method in accordance with claim 1, wherein said fooing further comprises doing D. 3. A system for fooing bars, comprising A, B, and C. 4. A system in accordance with claim 3, wherein said fooing further comprises doing D. 5. A computer program product for fooing bars, comprising A, B, and C. 6. A computer program product in accordance with claim 5, wherein said fooing further comprises doing D. This covers several different legal precedents; even if "method" gets invalidated as an approach, the patent may still work under the "system" or "computer program product" theories. This gets taken to a ridiculous extreme; many such patents describe the computer program product as a physical object consisting of bits on a medium such as a CD or disk, and include as a figure an illustration of such a medium.
- diminoten 14y ago
  It's not pretending, they actually are patenting the method rather than the code itself. It's a critical distinction, as code is actually copywritten. There's a huge difference between design and implementation.
hackinthebochs 14y ago
(I'm no expert at software patents so take what follows with a healthy dose of NaCl) It seems to me the difference is a question of "real world implications". For sorting integers using quicksort, yes that's just a mathematical algorithm: an algorithm that performs operations on mathematical entities. The moment you apply that algorithm to something that affects the real world, that's when it becomes patentable. E.g. sorting restaurant recommendations by applying quicksort to the individual rating. This is no longer simply a mathematical algorithm; the entities being manipulated have a 1:1 correspondence with entities in the real world. It is recognizing the usefulness of this particular correspondence between mathematical entities and real world results that is patentable.

darkarmani 14y ago

> it's the application of the algorithm that gets the grant And that should simply be copyright for the expression of the algorithm.

chii 14y ago
no, the application would be the actual app. i would comply with copyright law if i simply wrote another app, whereas i will still violate the patent. the problem i have with algorithmic parents is that computer algorithms trend tend to be quite composable. other patents in other field aren't, e.g. pharma parents. when one algorithm could be such fundamental to a large amount of different apps, a patent on it gives the patent holder way to much than they deserve.
- darkarmani 14y ago
  I know the law, but I'm saying that we already have protection for expression on an algorithm -- i don't think we need to patent an app that is simply an implementation of an algorithm.