[GuestPost] The Court of Appeal's latest AI decision begs the question - what inventions are deserving of patent protection?
 |
| The IPKat is ready for your answers in our AI survey |
Over to Peter:
The High Court decision in November 2023 provoked a great deal of debate and discussion. The invention in this case related to a technique for providing media file recommendations to a user. The claimed method would provide recommendations for media files that were similar to an input file, based on apparently similar tastes of other humans. This was achieved by training an artificial neural network (ANN) using a training set that included media files that could be described objectively, based on their physical characteristics, and subjectively, in written form. The method was able to identify similar media files by merging an assessment of these two criteria together, mimicking the way in which they might be assessed by a human. The output was a file recommendation for a user. This file recommendation was determined to be “better” in the sense that it is more likely to meet the needs of a human user by providing an appropriate recommendation.
The High Court decision had two notable aspects. First, the Court decided that an ANN did not involve the use of a computer program at all. Accordingly, the Court decided that an ANN could not be excluded from patent law on that basis. Second, the High Court decided that even if the exclusion had been engaged, the relevant invention would have been allowable because it involved a technical contribution.
The UKIPO was quick to update its examination guidance to be consistent with the High Court decision. Their updated guidance instructed examiners not to object to claims relating to ANNs on the basis that they are programs.
The Court of Appeal judgment has now entirely reversed the High Court decision. According to the Court of Appeal, an ANN does indeed involve a program for a computer, and, in the particular case, it amounted to no more than a computer program because there was no technical contribution.
It is a little difficult to provide a positive overall message about the Emotional Perception case. The flip-flopping of guidance from the courts has caused a great deal of uncertainty among applicants. Some applicants will have updated their filing strategies and approach to claim drafting in an attempt to take advantage of the changed UKIPO guidance. The Court of Appeal decision reverses all of that and takes us back to the situation, much as it was in November 2023.
One positive that can be taken is that the decision does, at least, maintain a certain consistency between the law in the UK and the EPO. There is also a pleasing symmetry in the treatment of simulations in G1/19 and ANNs by the Court of Appeal. Both were seen as sub-categories of computer programs, and neither was held to deserve any special treatment from a patentability perspective. The basic principles for examining computer implemented inventions should be applied in both cases, and the decisive question should be whether the relevant invention involves a technical contribution, or, if you prefer, whether it provides a technical solution to a technical problem.
A potential negative from the Court of Appeal decision relates to the way they chose to define a computer and a computer program. The fact that the Court even decided to provide a definition is something of a surprise (to this writer, at least). They could have chosen to decide that an ANN is a computer and involves a computer program without adopting a particular definition that could prove to be problematic when applied to other technologies. The definition provided for a computer is a "machine which processes information" and a program for a computer has been defined as “a set of instructions for a computer to do something”. These are broad definitions. It could be argued that a transistor is a basic form of computer, given that it can process digital information. A transistor could be controlled by a basic element like a timer, and the timing signals could be interpreted as a set of instructions for the transistor to do something. Similarly, a sextant could be considered a kind of mechanical computer for calculating angles. The instruction manual for the sextant could be interpreted as a program for a computer. This last example may stretch the definition a little, but they demonstrate the potential problems generated by the new definitions. It would be a concern if the UKIPO were to use this wording to fit the computer program exclusion to a broader category of devices.
More broadly, the decision causes us to re-examine the requirements for obtaining a patent in this area. The requirement for technical subject matter has become a familiar part of the landscape for patent attorneys working in high tech. As a group, we have become adept at explaining the tests as applied by the EPO and the UKIPO to clients and often confused practitioners from other countries. As a captive can develop affection for their captor, we are in thrall to the technical contribution and sometimes this familiarity means that we fail to question whether it really makes sense. It is in this context, I would suggest, that many have chosen to celebrate the Court of Appeal decision in Emotional Perception. However, at an intellectual level it is difficult to explain why the media file recommendation engine developed by Emotional Perception should be any less deserving of a patent than an invention that provides a technical contribution. It seems like such an arbitrary place to draw the line. Why should it matter that the file recommendation in Emotional Perception can only be judged as being better from a subjective, human perspective?
This raises a fundamental question about what kind of inventions are really deserving of patent protection, and whether we are ready to move on from the “technical contribution” tests."
Create your own user feedback survey
"The recent case involving Emotional Perception, both in the High Court and the Court of Appeal, has divided opinions within the patent profession. Some have argued that the High Court decision provided a welcome change: an opportunity for the UK to lead the way in AI patents. Others called out the High Court decision as rogue and soon-to-be-overturned. Whatever your view, the Emotional Perception case does shine a spotlight on the borderlands of patentability, and forces us to re-examine where the dividing lines should be drawn. More generally, it has re-opened the debate on whether patent law remains fit for purpose in the modern age, especially when it applies to computer programs. Should we continue trying to stretch the Patents Act to deal with new technologies, or is it time for a legislative overhaul and a fresh look at what kind of inventions should be considered patentable?
The High Court decision in November 2023 provoked a great deal of debate and discussion. The invention in this case related to a technique for providing media file recommendations to a user. The claimed method would provide recommendations for media files that were similar to an input file, based on apparently similar tastes of other humans. This was achieved by training an artificial neural network (ANN) using a training set that included media files that could be described objectively, based on their physical characteristics, and subjectively, in written form. The method was able to identify similar media files by merging an assessment of these two criteria together, mimicking the way in which they might be assessed by a human. The output was a file recommendation for a user. This file recommendation was determined to be “better” in the sense that it is more likely to meet the needs of a human user by providing an appropriate recommendation.
The High Court decision had two notable aspects. First, the Court decided that an ANN did not involve the use of a computer program at all. Accordingly, the Court decided that an ANN could not be excluded from patent law on that basis. Second, the High Court decided that even if the exclusion had been engaged, the relevant invention would have been allowable because it involved a technical contribution.
The UKIPO was quick to update its examination guidance to be consistent with the High Court decision. Their updated guidance instructed examiners not to object to claims relating to ANNs on the basis that they are programs.
The Court of Appeal judgment has now entirely reversed the High Court decision. According to the Court of Appeal, an ANN does indeed involve a program for a computer, and, in the particular case, it amounted to no more than a computer program because there was no technical contribution.
It is a little difficult to provide a positive overall message about the Emotional Perception case. The flip-flopping of guidance from the courts has caused a great deal of uncertainty among applicants. Some applicants will have updated their filing strategies and approach to claim drafting in an attempt to take advantage of the changed UKIPO guidance. The Court of Appeal decision reverses all of that and takes us back to the situation, much as it was in November 2023.
One positive that can be taken is that the decision does, at least, maintain a certain consistency between the law in the UK and the EPO. There is also a pleasing symmetry in the treatment of simulations in G1/19 and ANNs by the Court of Appeal. Both were seen as sub-categories of computer programs, and neither was held to deserve any special treatment from a patentability perspective. The basic principles for examining computer implemented inventions should be applied in both cases, and the decisive question should be whether the relevant invention involves a technical contribution, or, if you prefer, whether it provides a technical solution to a technical problem.
A potential negative from the Court of Appeal decision relates to the way they chose to define a computer and a computer program. The fact that the Court even decided to provide a definition is something of a surprise (to this writer, at least). They could have chosen to decide that an ANN is a computer and involves a computer program without adopting a particular definition that could prove to be problematic when applied to other technologies. The definition provided for a computer is a "machine which processes information" and a program for a computer has been defined as “a set of instructions for a computer to do something”. These are broad definitions. It could be argued that a transistor is a basic form of computer, given that it can process digital information. A transistor could be controlled by a basic element like a timer, and the timing signals could be interpreted as a set of instructions for the transistor to do something. Similarly, a sextant could be considered a kind of mechanical computer for calculating angles. The instruction manual for the sextant could be interpreted as a program for a computer. This last example may stretch the definition a little, but they demonstrate the potential problems generated by the new definitions. It would be a concern if the UKIPO were to use this wording to fit the computer program exclusion to a broader category of devices.
More broadly, the decision causes us to re-examine the requirements for obtaining a patent in this area. The requirement for technical subject matter has become a familiar part of the landscape for patent attorneys working in high tech. As a group, we have become adept at explaining the tests as applied by the EPO and the UKIPO to clients and often confused practitioners from other countries. As a captive can develop affection for their captor, we are in thrall to the technical contribution and sometimes this familiarity means that we fail to question whether it really makes sense. It is in this context, I would suggest, that many have chosen to celebrate the Court of Appeal decision in Emotional Perception. However, at an intellectual level it is difficult to explain why the media file recommendation engine developed by Emotional Perception should be any less deserving of a patent than an invention that provides a technical contribution. It seems like such an arbitrary place to draw the line. Why should it matter that the file recommendation in Emotional Perception can only be judged as being better from a subjective, human perspective?
This raises a fundamental question about what kind of inventions are really deserving of patent protection, and whether we are ready to move on from the “technical contribution” tests."
To that end, time for a IPKat survey - is it time to overhaul the exclusions to patentability? The Kat team looks forward to your views and opinions in the comment section.
[GuestPost] The Court of Appeal's latest AI decision begs the question - what inventions are deserving of patent protection?
![[GuestPost] The Court of Appeal's latest AI decision begs the question - what inventions are deserving of patent protection?](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhA7tPmh6FacHeANUNGfp6c1UUhYHOu0ocD2vMSa844NtpY_Y3UCbJp9LNBt_qKMhNG6pQEn80MrpvBJ7AqwlQycVF66DXOS7H0QHyxMb9cgr8MSQEphANnEuEuPhwYPlFqH1zuaezmULMtf5NFuoaRpBdK9JAO7iowYJZrYH8KN5Ii-mbW0WqsQQ/s72-c/Cat%20ANN%20AI.jpg)
Reviewed by Annsley Merelle Ward
on
Wednesday, August 07, 2024
Rating:
Reviewed by Annsley Merelle Ward
on
Wednesday, August 07, 2024
Rating:
Follow the IPKat on LinkedIn 
![[Guest post] Thou shalt not discriminate the Lego block over any other kind of product – GRUR speaks out on C-211/24](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgWJYQUJNvYydrZcjwSHPuxd14xX3_48aBusHKBogY-Wg1BPSJSABbFFjmGiUV3aqG2npP9xYaIl1vtWK5AsqGOQPgRD2TWFLhyyc4xGm1fHDD-tt9kHcrSSgAyju2P9wSKz0VW6h7D5MD4Hxd9vREyxKYzlMRNXxxtxc3DiNsXppfVaxgSxAMV/s72-c/21742619125_7c9e53cd39_o.png)
![[Guest post] Comparison of wine and gin: the EUIPO’s unquestionable Similarity Tool](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiNfTEBXiPVS0XjZi4mNE1OtXLxXiEp1PUgcUK7MMMXlfyTkO5SMXv-VCMXXNdS_F1XdMMxYca1VbBTGvvDeTVpo5x-5YwR7atQJmK3x32T8LACpf4308kht3nEtaxPwEPcGGrB0s5xjGktmQPOiVwPVLRO5JbW19qupc7xUtiVutvaWHow0OpBkA/s72-c/Picture%201.jpg)
![[Guest post] Muddying the waters of copyright – a rapid reaction to today’s IPEC decision which sees the sinking of the WaterRower’s claims for copyright protection](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh7chFANC7ppmZl4_nlR6_hZEcg6cOhVEJNVXqki4p82BoVlcLFotuOXzHWYD5lKLiiK_kRoWeXmJjjaV2DpoKsAWq8U3LFioO5W_841a9gcfDIFjvZWcGqfnr0jWJNtyXptIYR_8bxCssaDkJ9KMokuSqy1g65Il7p2hPwo4MfRCEGoUPQJesvbg/s72-w400-c-h189/Screenshot%202024-11-11%20at%2018.20.45.png)
Posts
I agree entirely with the point 'As a captive can develop affection for their captor, we are in thrall to the technical contribution and sometimes this familiarity means that we fail to question whether it really makes sense … It seems like such an arbitrary place to draw the line.' It really doesn't make sense. Its roots lie a conflation between the concepts of program and algorithm.
ReplyDeleteAccording to standard definitions, a neural network is not a computer and it is not a program for a computer; it is an algorithm. The neural network algorithm may be implemented in hardware (i.e. without a computer and without a program for the computer) or software (i.e. with a computer and a program for the computer). This straightforward explanation would be agreed without qualification by any computer scientist, yet it is problematic for patent professionals because it highlights that there is in fact a distinction between 'algorithm' and 'program': it follows from this distinction that the 'new' thing claimed by a computer-implemented invention claim is a new algorithm, not a new program for a computer, so the program exclusion should not bite. Indeed, it is difficult to imagine what a claim bitten by the program exclusion (interpreted in accordance with standard definitions) might look like, just as it is difficult to imagine what a claim to any other literary, dramatic, musical or artistic work might look like. Therefore, in order to exclude Emotional Perception AI's invention, the patent offices and courts need to apply non-standard (and nonsensical) definitions of 'computer', 'hardware', 'program', and 'technology'. It has to be asked, why!? Does the tech sector (TRIPS!?) benefit from the burden of additional complexity and uncertainty?
I wrote about the non-standard definitions of 'hardware', 'program', and 'technology' here: https://whereistheprogram.wordpress.com/2024/04/08/emotional-perception-ai-the-fallacy-of-equivocation-and-original-sin/
I wrote about the non-standard definition of computer here: https://whereistheprogram.wordpress.com/2023/12/01/emotional-perception-ai-and-reflections-on-gales-application-where-is-the-computer-where-is-the-program-and-where-is-the-submarine/
You argue that “it is difficult to imagine what a claim bitten by the program exclusion (interpreted in accordance with standard definitions) might look like, just as it is difficult to imagine what a claim to any other literary, dramatic, musical or artistic work might look like”, but this highlights exactly where you misunderstand how the exclusions are applied. A patent to a “literary work” is not engaged only by a claim 1 that simply recites a story. The exclusion would, and should, still bite if claim 1 is “A book comprising a plurality of pages having a story thereon, the story comprising a beginning, a middle and an end, characterized in that the story includes [insert novel and non-obvious literary device]”. The courts favour substance over form. (As an aside, this is something which patent attorneys often fail to appreciate, with a tendency to purely focus on minutiae and technicalities while ignoring the bigger picture.) Evidently, the substance of my claim to a book is a literary work that should not be patented according to the clear intent of the legislator. Adding some notorious paper pages should not render the book patentable and there needs to be some way of throwing out such claims. The EPO chooses to do so under inventive step while the UK courts choose to do so up front by stripping out the form and looking at the substance of the claim. The end result is not that different. But your argument in respect of computer programs requires such a literal reading of the legislation that it renders the legislative intent worthless and thus cannot be correct. There needs to be a more holistic approach than your argument, which seems to be analogous to “claim 1 is not a literary work as such because it includes some extra physical requirements like pages so my non-obvious claim to a book should be allowed”.
DeleteThanks for your response!
DeleteI agree that, as you say:
the exclusion would, and should, still bite if claim 1 is “A book comprising a plurality of pages having a story thereon, the story comprising a beginning, a middle and an end, characterized in that the story includes [insert novel and non-obvious literary device]”
Here, the exclusion would and should bite because the only new thing claimed is a new literary work, nothing more: a new literary work 'as such'. Such things are within the purview of copyright.
Similarly, the exclusion should and would still bite if claim 1 is "a computer readable media comprising a plurality of storage blocks having stored therein [insert new program code implementing known algorithm using a non-obvious, perhaps beautiful, coding flourish]". The exclusion would and should bite because the only new thing claimed is a new program, nothing more: a new computer program 'as such'. Again, such things are within the purview of copyright.
In contrast to the above, the program exclusion should not bite if claim 1 is "a computer-implemented method comprising [insert new algorithm]". Here, the new thing claimed is categorically not a new program, it is a new algorithm, and according to a title in chapter 1 of CLRS, one of the most cited scientific publications of all time (circa 70,000 citations), published by the mecca of technology, MIT: 'algorithms are technology'. What is it TRIPS says about technology?
Importantly, programs and algorithms are categorically different things having different definitions. One is a symbolic/literary work, one is technology. The concepts of program and algorithm might be mistakenly conflated by a layperson, but, as the recent decisions have shown, it is impossible to correctly describe computer technology without the words algorithm and program referring to distinct things.
I do not agree that the principle of substance over form means that algorithms should engage the program exclusion. The difference between a program and an algorithm is analogous to that between a product and its aesthetic design - for example, any claimed product may be implemented using an infinite number of different aesthetic designs, and any claimed algorithm may be implemented using an infinite number of different programs. The fact that an item infringing a claim to a product would necessarily also embody an aesthetic design does not that mean claims to products should be excluded. That is not the principle of substance over form (in Gale's Application, the principle meant that a program stored on a floppy disc should be treated the same as a program stored in ROM).
I do not agree that interpreting 'program' using its dictionary definition renders the legislative intent worthless. Of the two computer-implemented invention claims I discussed above, one should be excluded, the other should not be. One is within the purview of copyright, the other is not. The one that is not within the purview of copyright is widely recognised to be technology.
I wrote about the above points in more detail here: https://whereistheprogram.wordpress.com/2024/04/08/emotional-perception-ai-the-fallacy-of-equivocation-and-original-sin/
On the topic of legislative intent, whilst much is obscured, the Lords did state that the reason for the particular positioning of program in the list of exclusions was:
Deleteto ensure that ... the exclusion will bite only on so-called computer software and not on programs which may be embodied in the hardware (The Earl of HALSBURY)
Other Lords were concerned the positioning wouldn't prevent this mistake from being made, and proposed explicitly defining program to ensure that the courts did not later confuse:
programs written in high-level programming languages, coded instructions and so forth, all of which is symbolic, and engineering activity which is producing ... programmed hardware ... which is not symbolic. In the words of the European Patent Convention, the latter is "susceptible of industrial application" (Lord IRONSIDE)
There is a lot of foresight in the Lords debates - for example:
sooner or later some judicial pronouncement has to he made on what is a purely technological matter (The Earl of HALSBURY)
Therefore, as far as the legislative intent can be determined, it is that program means program (symbolic/literary work), not algorithm (technology). Indeed, the qualification of 'as such' is a strong indicator that program means program, not something else.
ReplyDeleteThis is going to be unpopular view but it is said in good faith.
The law is very clear and has been for some time. G1/19 and the CoA were just reminding us of that fact.
The high court decision departed from the status quo and actually disadvantaged those who decided not to file knowing their competitors were also bound by the same law.
One of the ironies in this case is that emotional perception presumably had and will have FTO due to the lack of patents in this area.
This case and the exclusions beg the question whether patents are needed at all. There has been no shortage of software innovation in the last 5 decades in spite of the patent carrot that is allegedly needed. We can only speculate what might have happened were there more patents in play.
Removing the exclusions primarily benefits the patenting process, but that alone isn’t a good reason to change the law.
The law isn't clear at all and the various Emotional Perception AI decisions highlight that fact. For example, according to the latest decision, a neural network 'however it is implemented … is clearly a computer – it is a machine for processing information' (68). So, a new neural network architecture is (clearly) a new computer, not a new program for a computer, and new computers (e.g. new neural network architectures) are not excluded from patentability, nor are new methods of manufacturing computers (e.g. neural network pruning techniques and neural architecture search techniques). This is not inline with the original AI guidance published by the UK-IPO, previous CoA decisions, or EPO practice. This can be taken further: a binary search algorithm is a computer because, like the neural network algorithm (or indeed any algorithm), it is a machine for processing information that can be represented using 'nodes' and 'links'; therefore, a new binary search algorithm is a new computer, not a new program for a computer, and does not engage the program exclusion.
DeleteThe problem is not with the letter of the law, it is with the office and courts interpretation of it. As mentioned in my earlier comment, in order to exclude certain things from patentability, the offices and courts use non-standard definitions for various computer science terms. If the law was interpreted using GCSE computer science definitions of program, computer, and algorithm, it wouldn't be such a mess.
Whether computer-implemented inventions are the sort of things one should be able to get a patent for is a separate point. It is a significant policy decision that should not be left to the patent office. I am not convinced that the current interpretation of the law by the office can be justified on the basis that tech companies are 'free to operate'. They are not. The exclusion doesn't stop patents being granted for AI inventions, it just massively complicates the process, introducing a lot of uncertainty and cost. I don't see how this benefits anyone.
The law is clear. Technical contribution in UK or solving a technical problem at EPO is all that’s needed. What is or is not technical is a matter of judgment but there is enough case law out there to predict the likely outcome. The edge cases are few and far between.
DeleteHad this case been about a better computer in line with Symbian or Apple v HTC, it would have been allowed. It was not and correctly refused in line with the clear case law.
Specific paragraphs in the judgment don’t really change anything in substance. The broader point is that music recommendations is not something worthy of a patent. It’s no different to me recommending music to a friend. Using an ANN rightfully doesn’t change this. Had the ANN used the underlying hardware in an improved way, then yes a patent is possible, but the recommendation aspect would have no bearing. Substance over form in action.
The law may very well be misinterpreted by the courts in your view, but that doesn’t mean it can be ignored.
The bar for an AI patent is high and will require technical limitations, which opens up designarounds. Low quality patents to computer programs with no such limitations benefits no one, hence the SCOTUS striking out such patents. Patent attorneys might very well complain but there was wider concerns in allowing low quality software patents bogging down innovation.
My FTO point is to do with the status quo. Changing the law changes the FTO position. Judges are rightly concerned with this even though you won’t find it in a written decision.
As anyone who listens to the radio knows, if someone states that something is clear, it is usually anything but.
DeleteHost: Why is it that the patent office interprets the word ‘program’ to mean ‘algorithm’ when these are categorically distinct things. Doesn’t the ‘as such’ wording of the exclusion explicitly prohibit interpreting the exclusions to cover things other than those listed? This seems to be causing a lot of confusion, uncertainty, and expense for technology companies.
A: TRIPS is very clear that patents will be granted for any inventions in all fields of technology.
Host: The technology sector, academia, and science more generally consider algorithms to be technology. Why does the patent office disagree?
B: The office has been very clear on this: algorithms are not technology because technology is something that has technical character.
Host: You didn’t answer the question ...
B: If you would let me finish: technical character results from the use of technical means.
Host: What?
B: I have been absolutely clear: technical means have a technical purpose.
Host: Are you broken?
B: Look, I have always been and will continue to be absolutely clear about this: technical purpose means a technical problem is being solved.
Host: …
B: …
Host: …
B: Technology is whatever the patent office decides to grant a patent for.
I disagree that ‘specific paragraphs in the judgement don’t really change things in substance’. The paragraphs change things in at least two significant ways.
DeleteFirstly, and perhaps most importantly, the purpose of the paragraphs is to justify the decision. If those paragraphs don’t make technical sense, the decision is not justified. I find it disappointing that so many seem so willing to ignore palpable mistakes in reasoning provided that the resulting decision is what they expected/desired. Rule of lawyers replacing rule of law.
Secondly, the paragraphs set precedent that must be followed - for example, UK Statutory Guidance has now been updated to state that ‘an ANN is a computer’. Therefore, the specific paragraph I quoted unquestionably changed things. I have already filed several responses along the lines of: ‘neural networks are computers, and the claimed neural network architecture is not known; therefore, the contribution is a new computer architecture; and new computers are not excluded from patentability.’ I suspect this is the sort of thing you would consider ‘not … worthy of a patent’, but the law is the law and it is clear, right?
We seem to be arguing different aspects.
DeleteWhen I say clear, I am not passing judgment on whether it is correct or not. The law is clear on what is and what isn’t allowed and to that end predictable.
Clear in your case seems to equate to the law being wrong. Unfortunately, that’s for judges to decide. Have you considered that an algorithm is not limited to computer program? It is as much a mathematical method as an abstract thought process as it is the steps carried out once a program is executed.
To address your specific point, your new argument is the software implementation fallacy. Yes an EPO argument but not something that is easily ignored in the UK in view of the clear law.
Any argument that renders the statutory exclusion worthless is destined to fail.
I posted this on the wrong thread!
DeleteI don’t believe the case law is clear on what is and what isn’t an invention. For example, is a new neural network architecture an invention according to UK law? The law clearly states that a neural network is a computer. The list of exclusions clearly doesn’t include computers. It follows that a new neural network architecture is clearly an invention. Would you agree?
The law clearly states that a neural network is a computer, not a mathematical method, so the mathematical method exclusion doesn’t apply.
Say I invented the binary search algorithm, which, like a neural network, can be represented using 'nodes' having weights and 'links'. The law clearly states that this sort of thing is a computer. The list of exclusions clearly doesn’t include computers. It follows that the binary search algorithm is clearly an invention. Would you agree?
I’m afraid I don’t know what the software implementation fallacy is or how it applies to the above reasoning. I suspect it wouldn’t apply because UK law clearly states a neural network is a computer, not software. I also don’t see how this is an EPO argument because the EPO don’t consider neural networks to be computers, only UK law does.
I don't believe I have made an argument that renders the statutory exclusion worthless, though I suppose that depends on how you measure the exclusions worth. It is my view that the letter of the Act excludes programs (symbolic works), not computer-implemented algorithms (technology), and thus programs are all it should exclude. The exclusion of programs makes absolute sense in the context of the other exclusions, particularly the literary work exclusion. The exclusion interpreted in line with the letter of the Act would be as ‘worthy’ as the literary works exclusion. Why do you think interpreting it in this way would be worthless?
This is a very nice article, but if you wish to ask the broader question of what types of invention are worthy of patent protection, then perhaps the (policy) purpose of the exclusion for computer programs also needs to be analysed. The software/computer field has its own R&D structure and economics which will govern how investment in the field will happen and what sort of activities remain free of patent protection. The legislators clearly wanted a patent-free area for computer programs and software to be developed, and Birss might have been mindful of this when deciding on the matter at the Court of Appeal. To undermine (reduce) an exclusion is a serious matter, and legislator's intent is important to keep in mind when doing so.
ReplyDeleteI agree that legislative intent is important. However, I do not believe the current interpretation of the law is in line with legislative intent. For example, the qualification of 'as such' strongly indicates that the legislative intent was that 'program for a computer' should be interpreted to mean 'program for a computer', not something else. This intent is also apparent in the Lords debates. For example, the reason for the particular positioning of program in the list of exclusions was:
Deleteto ensure that ... the exclusion will bite only on so-called computer software and not on programs which may be embodied in the hardware (The Earl of HALSBURY)
Some members were concerned the positioning wouldn't prevent this categorical mistake from being made, and proposed explicitly defining program to prevent confusion between:
programs written in high-level programming languages, coded instructions and so forth, all of which is symbolic, and engineering activity which is producing ... programmed hardware ... which is not symbolic. In the words of the European Patent Convention, the latter is "susceptible of industrial application" (Lord IRONSIDE)
There is a lot of foresight in the Lords debates - for example:
sooner or later some judicial pronouncement has to he made on what is a purely technological matter (The Earl of HALSBURY)
Therefore, as far as the legislative intent can be determined, it is that program means program (symbolic/literary work), not algorithm (technology). This makes perfect sense - a new program implementing a known algorithm (a new program per se) is not technology because it provides no new effect, and new programs are protected by copyright (protection anticipated by the lords).
I don’t believe the case law is clear on what is and what isn’t an invention. For example, is a new neural network architecture an invention according to UK law? The law clearly states that a neural network is a computer. The list of exclusions clearly doesn’t include computers. It follows that a new neural network architecture is clearly an invention. Would you agree?
DeleteThe law clearly states that a neural network is a computer, not a mathematical method, so the mathematical method exclusion doesn’t apply.
Say I invented the binary search algorithm, which, like a neural network, can be represented using 'nodes' having weights and 'links'. The law clearly states that this sort of thing is a computer. The list of exclusions clearly doesn’t include computers. It follows that the binary search algorithm is clearly an invention. Would you agree?
I’m afraid I don’t know what the software implementation fallacy is or how it applies to the above reasoning. I suspect it wouldn’t apply because UK law clearly states a neural network is a computer, not software. I also don’t see how this is an EPO argument because the EPO don’t consider neural networks to be computers, only UK law does.
I don't believe I have made an argument that renders the statutory exclusion worthless, though I suppose that depends on how you measure the exclusions worth. It is my view that the letter of the Act excludes programs (symbolic works), not computer-implemented algorithms (technology), and thus programs are all it should exclude. The exclusion of programs makes absolute sense in the context of the other exclusions, particularly the literary work exclusion. The exclusion interpreted in line with the letter of the Act would be as ‘worthy’ as the literary works exclusion. Why do you think interpreting it in this way would be worthless?
Santa, it needs to be remembered that the computer program exclusion was drafted in the late 1960s/early 1970s, so essentially before the computer industry we know today had developed.
DeleteI wonder how the courts today would view a claim to a systolic array processor, as patented by the MoD in the early eighties.
Agreed Kant, but my point was just the simple one that we need to consider legislative intent in the process of deciding which inventions are patentable. Yes, legislative intent might be out of date, but that needs to be examined as part of the discussion
Deletein line with what a commentator said on a previous post, who is to judge what inventions are deserving of patent protection? is there a new patent police self appointed?
ReplyDeleteThe decision is correct in its result, but the reasons for it are not at all convincing. Discussing whether a neural network is a program appears rather beside the point and shows a total lack of understanding of the technique involved. On the other hand, it is more surprising that the UKIPO came up with this kind of argument.
ReplyDeleteOne thing is clear from the outset: there is no technical contribution from the invention in question. No effect other than the choice of similar music. Even with Aerotel this application should have been refused for lack of technical effect.
Both in the High Court’s and the Court of Appeal’s decisions, T 702/20 was cited. It appears that this decision was misquoted, or better, not understood. T 702/20 makes clear that “A neural network defines a class of mathematical functions which, as such, is excluded matter. As for other "non-technical" matter, it can therefore only be considered for the assessment of inventive step when used to solve a technical problem, e.g. when trained with specific data for a specific technical task.”
There is no technical contribution whatsoever, as the choice of music is by no means technical, and the invention was a not merely creating a program. This is confusing cause and effect. The output does indeed not turn the claimed invention into a system which produces a technical effect, but the training of the neural network is not merely excluded subject matter.
Furthermore, it does not appear that the trained artificial neural network has been described and, more importantly, that the training data for the artificial neural network has been disclosed.
Apart from the non-technical nature, there is also a clear problem of insufficient disclosure. T 161/18 made clear that in the absence of training data for a neural network, the description is insufficient.
Much is made of inventions using AI, but in order to obtain a patent, the learning data must be disclosed. Since data is a precious commodity in the modern world, I doubt that many applicants will be prepared to disclose it.
The High Court overturned the UKIPO’s rejection of the application. The High Court was right to say that in the case of a trained artificial neural network, it is not a program as such. The High Court has however overlooked the absence of technical effect.
In an article published in Chemistry World on August 1th by Victoria Atkinson, “Disappointing AI judgment could have a ‘catastrophic effect’ on UK science”, it is said that ‘Fundamentally, [the judges] didn’t understand the technology’. It is difficult not to agree with this statement.
Indeed, in view of the Common Law system, this decision might have dear, negative, consequences. One feels set back at the time of the “salami tactics”, the computer is known, and the program is not patentable, so nothing is patentable. Since VICOM, T 208/84 the situation has quite evolved.
Last but not least, the reasons for the exclusion of computer programs as such is a very practical one. There are many programming languages with their own peculiarities/syntax rules, and more over different versions.
At the time, and still nowadays, it is impossible to carry out a significant search only in programs. In the early days, the EPO accepted short excerpts of programs, but the chore of the invention had to be disclosed in plain wording.
Thanks @PeterArrowsmith for your interesting comments. I'm not surprised that the Court defined computer program. To do that, it's also not surprising that the Court defined 'computer'. If Step 3 in Aerotel is important (and not simply folded into Step 4), this is necessary.
ReplyDelete@PeterArrowsmith: I agree with you that there is considerably lack of clarity and uncertainty in the interpretation of Art 52(2). But I think that it stems from the definition of 'technical' and its application to particular facts. Other aspects of the provision are relatively clear in EPO and UK jurisprudence.
ReplyDelete@PeterArrowsmith: I disagree to some extent with your point about there being 'a pleasing symmetry' with EPO jurisprudence. The UK Court of Appeal overlooked some key points in EPO jurisprudence. For instance, EPO jurisprudence holds that 'technical character' includes any “teaching that methodically utilizes controllable natural forces to achieve a repeatable, causal, and perceivable result” including a “technical means of implementation”. The subsequent question is whether the technical character involves an inventive step (or putting it another way, is not obvious).
ReplyDeleteIn this Emotional Perception case, the Court of Appeal ignored the technical means of implementation (namely the ANN) and did not ask whether the ANN was an obvious or non-obvious technical means of implementation.
Accordingly the UK Court of Appeal has excluded a type of invention, which the EPO might have held it valid IF the ANN was disclosed with enough specific details that it disclosed a non-obvious inventive step. This is not pleasing symmetry in my view.
I also note para 48 in G0001/19 which allowed the simulation invention to be modelled on human factors such as "dissatisfaction function”, “inconvenience function” and “frustration function”, as well as physical, measurable parameters. Thus I do not agree that there was a pleasing degree of symmetry between this Court of Appeal judgement and EPO case law. The Court of Appeal was very dismissive of an invention that engaged with human factors such as opinions about the similarity and difference in two pieces of music.
A further comment from me: the invention reduced search costs and time, and resulted in a 'better ANN', which seems to indicate that it did indeed fall within the line of cases about 'better computers'; contrary to what the Court of Appeal said.
ReplyDeleteAs for legislative intent: It is simply not conceivable that the legislative intent was to allow computer programs to be the subject matter of a patent if they improved a drill bit (Halliburton) or the responsiveness of multi-touch gestures (HTC v Apple), but to disallow them if they gave consumers better recommendations for music and other files and reduced their search costs (Emotional Perception). Both the EPO and UK tests for interpreting Art 52(2) lost touch with any sensible legislative intent long ago.
ReplyDelete@PeterArrowsmith: but just to reiterate, genuine thanks for your post. Your comments were very interesting, clear, helpful and thought-provoking. Really appreciate you taking the time to post a blog on the case.
ReplyDelete