The question of whether a machine can be an inventor for the purposes of patent law has become very acute with the breakthrough of artificial intelligence technology in the 21st century.
The Copyright, Designs and Patents Act 1988 (CDPA) only recognised humans as authors of copyright-protected works (see this post for a discussion about copyright in monkey selfies), and even section 9(3) dealt with computer-generated works by identifying a human as the author. Similarly for patents under the Patents Act 1977: under section 7, only a "person" may make an application for a patent (possibly jointly with other persons), which may be granted to the inventor or joint inventors or certain other persons, and (the tailpiece to s.7(2) stresses) "to no other person". As in the case of the monkey selfie, the word "person" seems to limit the grant of a patent to human beings. But the 1977 Act does not credit a human as the effective inventor in the way that the copyright law substitutes a human author for a computer.
DABUS-the-inventor
The EPO has recently declined applicants in two related cases, EP 18 275 163 and EP 18 275 174, because both named the machine (DABUS) as the inventor. The decisions were widely reported by media and legal platforms alike (see Practical Law here (subscription required) and BBC news here).
In a nutshell, a machine has an owner, who is an individual. A machine, therefore, cannot transfer rights to a successor in title. The failure to name a natural person as an inventor, as required by the EPC, meant that the applications had to be refused. The applicants have a right to appeal the decisions, so we will continue monitoring the developments of these cases.
Analysis of these EPOs decisions shows that they are based on outdated law, but this is not uncommon: indeed, it is inevitable, because the law cannot anticipate technological change. Similar legal systems are in force all around the globe. However, since AI inventions are becoming increasingly more common, it seems that the world would have to adapt to the new reality.
Unlimited patents for minimum investment
As mentioned above, CDPA contains provisions for copyrights given to computer-generated works, which seems to be as close as English law has got to recognising AI copyright. Once the machine copyright is widely recognised, it seems logical that the actual human and/or the company that owns the machine would naturally get the reward from the machine's inventions. In fact, we would only have a longer chain, but the principle would stay the same - the humans would be the ultimate beneficiaries of AI copyrights.
Allowing AI patents could open the floodgates for various AI machines producing inventions almost like a conveyor belt. After all, a machine needs no food or sleep. It only needs a couple of engineers to maintain it, which poses a risk AI could inadvertently aid employee redundancies (though this is a wholly different topic!).
This ultimately means that companies could receive maximum profits from their patents with minimum investments. All a company has to do is to employ an inventor (or a small group of such) to produce one AI machine, which, in turn, would potentially produce an infinite number of inventions in any given area of science or medicine.
For such inventions to be successfully patented, however, the criteria of patentability would have to be slightly amended, so as to re-define an inventive step for machine inventions (see my arguments below).
Patents? Who needs patents!
Understandably, the development of AI can question the whole idea of patenting. To be granted a patent, an invention must be:
An inventive step, however, is a tricky category to assess. Inventiveness as a criteria is assessed in light of obviousness, which for years worked well for human inventions. However, how would one assess obviousness of a machine invention? Where a machine can learn from everything ever invented in the world, it may be that everything would seem obvious!
One may argue that a machine would be placed in an unfair advantage against its human competitors. As mentioned above, for an inventor it would seem a more attractive option (and would make more commercial sense) to create an AI machine to produce something that is guaranteed to satisfy all patent requirements rather than waste time and effort to produce something that would (most likely) never sustain any competition with an artificial intellect.
If this was the way the humankind wanted to go, it may well mean that the whole patenting system could crumble and patent laws could become redundant. Would the world be sad if that happened? Patent attorneys would probably shed a tear or two.
So what shall we do?
In light of the above, the humanity would have three obvious ways: 1) to abandon patents all together; 2) to restrict patenting only to human inventions; 3) to allow patenting of AI inventions but reform the patent requirements and award patents to the owners of AI technology.
If the second option is chosen, there have to be very strict criteria and close supervision over human inventors to make sure there were no AI used in order to aid a real “human invention”. I find this a very costly and not very practical solution.
The first and the third options seem to be the most likely ones in light of the current developments in technology. Both of these have to be closely scrutinised to make sure they can sustain the test of time, before any one of them is chosen.
To sum up, I firmly believe that the legal recognition of AI inventors is only a matter of time and AI will no doubt be recognised as a part of our world's technological progress. Just like an industrial revolution saw factories emerge, AI is only a natural step forward for humankind, whether we like it or not.
We will no doubt keep you posted on the developments in this curious area of law.
The Copyright, Designs and Patents Act 1988 (CDPA) only recognised humans as authors of copyright-protected works (see this post for a discussion about copyright in monkey selfies), and even section 9(3) dealt with computer-generated works by identifying a human as the author. Similarly for patents under the Patents Act 1977: under section 7, only a "person" may make an application for a patent (possibly jointly with other persons), which may be granted to the inventor or joint inventors or certain other persons, and (the tailpiece to s.7(2) stresses) "to no other person". As in the case of the monkey selfie, the word "person" seems to limit the grant of a patent to human beings. But the 1977 Act does not credit a human as the effective inventor in the way that the copyright law substitutes a human author for a computer.
DABUS-the-inventor
The EPO has recently declined applicants in two related cases, EP 18 275 163 and EP 18 275 174, because both named the machine (DABUS) as the inventor. The decisions were widely reported by media and legal platforms alike (see Practical Law here (subscription required) and BBC news here).
In a nutshell, a machine has an owner, who is an individual. A machine, therefore, cannot transfer rights to a successor in title. The failure to name a natural person as an inventor, as required by the EPC, meant that the applications had to be refused. The applicants have a right to appeal the decisions, so we will continue monitoring the developments of these cases.
Analysis of these EPOs decisions shows that they are based on outdated law, but this is not uncommon: indeed, it is inevitable, because the law cannot anticipate technological change. Similar legal systems are in force all around the globe. However, since AI inventions are becoming increasingly more common, it seems that the world would have to adapt to the new reality.
Unlimited patents for minimum investment
As mentioned above, CDPA contains provisions for copyrights given to computer-generated works, which seems to be as close as English law has got to recognising AI copyright. Once the machine copyright is widely recognised, it seems logical that the actual human and/or the company that owns the machine would naturally get the reward from the machine's inventions. In fact, we would only have a longer chain, but the principle would stay the same - the humans would be the ultimate beneficiaries of AI copyrights.
Allowing AI patents could open the floodgates for various AI machines producing inventions almost like a conveyor belt. After all, a machine needs no food or sleep. It only needs a couple of engineers to maintain it, which poses a risk AI could inadvertently aid employee redundancies (though this is a wholly different topic!).
This ultimately means that companies could receive maximum profits from their patents with minimum investments. All a company has to do is to employ an inventor (or a small group of such) to produce one AI machine, which, in turn, would potentially produce an infinite number of inventions in any given area of science or medicine.
For such inventions to be successfully patented, however, the criteria of patentability would have to be slightly amended, so as to re-define an inventive step for machine inventions (see my arguments below).
Patents? Who needs patents!
Understandably, the development of AI can question the whole idea of patenting. To be granted a patent, an invention must be:
- Something new;
- Something that can be put to practical use; and
- Something inventive.
An inventive step, however, is a tricky category to assess. Inventiveness as a criteria is assessed in light of obviousness, which for years worked well for human inventions. However, how would one assess obviousness of a machine invention? Where a machine can learn from everything ever invented in the world, it may be that everything would seem obvious!
One may argue that a machine would be placed in an unfair advantage against its human competitors. As mentioned above, for an inventor it would seem a more attractive option (and would make more commercial sense) to create an AI machine to produce something that is guaranteed to satisfy all patent requirements rather than waste time and effort to produce something that would (most likely) never sustain any competition with an artificial intellect.
If this was the way the humankind wanted to go, it may well mean that the whole patenting system could crumble and patent laws could become redundant. Would the world be sad if that happened? Patent attorneys would probably shed a tear or two.
So what shall we do?
In light of the above, the humanity would have three obvious ways: 1) to abandon patents all together; 2) to restrict patenting only to human inventions; 3) to allow patenting of AI inventions but reform the patent requirements and award patents to the owners of AI technology.
If the second option is chosen, there have to be very strict criteria and close supervision over human inventors to make sure there were no AI used in order to aid a real “human invention”. I find this a very costly and not very practical solution.
The first and the third options seem to be the most likely ones in light of the current developments in technology. Both of these have to be closely scrutinised to make sure they can sustain the test of time, before any one of them is chosen.
To sum up, I firmly believe that the legal recognition of AI inventors is only a matter of time and AI will no doubt be recognised as a part of our world's technological progress. Just like an industrial revolution saw factories emerge, AI is only a natural step forward for humankind, whether we like it or not.
We will no doubt keep you posted on the developments in this curious area of law.
Posts
No comments:
Post a Comment