29 September 2020

DABUS Denied – Machines Cannot be Inventors, and the English High Court Agrees

Robot touchIn Thaler v The Comptroller-General of Patents, Designs And Trade Marks [2020] EWHC 2412 (Pat), the England and Wales High Court has dismissed an appeal by Dr Stephen Thaler from a decision of the UK Intellectual Property Office (IPO), which determined that one or more natural persons must be named as inventor(s) on a patent application and, therefore, that two UK applications naming an ‘artificial intelligence’ called DABUS as inventor were deemed to be withdrawn for failing to satisfy this requirement.  Thaler had refused to identify a human inventor on the basis that no natural person had made an inventive contribution to the inventions disclosed in the applications.  Furthermore, he had identified himself as the applicant, claiming ownership of the inventions on the basis of ownership and control of DABUS.  The court also rejected this theory, finding no legal basis for a machine either to own the intellectual property, or for rights in machine-generated inventions to be transferred to a human owner.

Attempts to name DABUS as inventor on corresponding applications have also been rejected by the European Patent Office (EPO), and the US Patent & Trademark Office (USPTO) [PDF 5.27MB].

Readers who have been following developments in this case will be aware that while Thaler is named as the applicant and developer of DABUS, the real driving force behind efforts to recognise machine inventors is Professor of Law and Health Sciences at the University of Surrey, Ryan Abbott, through his Artificial Inventor Project.  In addition to the patent applications filed at the UK IPO, the EPO, and the USPTO, there is also a single international patent application filed under the Patent Cooperation Treaty (PCT) encompassing both inventions, published as WO2020079499.  Abbott not only heads up the Artificial Inventor Project, he is also named as agent for the PCT application, he represented Thaler in the English High Court appeal, and it has been reported that he is also representing Thaler in a parallel Federal District Court appeal in the US.

The PCT application recently entered the national phase in Australia, as application no. 2019363177.  A formalities report has been issued, objecting that the application does not comply with the Patents Regulations 1991, because ‘it is the Commissioner’s understanding that inventors for the purposes of the Patents Act 1990 must be natural persons.  As DABUS is not a natural person the Commissioner is of the view that no inventor has been supplied for the purposes of Regulation 3.2C(2)(aa).’  Thaler has until 21 November 2020 to either: provide the name or names of natural persons who are the actual inventors; or provide submissions to the Commissioner explaining why DABUS, despite being an artificial intelligence, can be legitimately recorded as an inventor.  I anticipate that Thaler will take the latter option, and that his submissions will be rejected on similar grounds to the UK case, given that the scheme for establishing inventorship and entitlement to the grant of a patent under the Australian Patents Act 1990 is similar to the corresponding UK provisions.

In my opinion, Abbott’s Artificial Inventor Project is misconceived.  His arguments as to why we should recognise non-human inventors are unsupported by any evidence, and amount to little more than a personal opinion.  His methods of pushing the issue through patent offices and into the courts may be successful in generating attention, but they are doomed to fail, and wasteful of administrative and judicial resources.  Furthermore, the very idea that machines might be granted the ‘right’ to be recognised as inventors lacks credibility, and is unlikely to secure widespread public support.

In this article, I shall explain my position in more detail, while also taking the opportunity to discuss the English High Court decision, which is likely to be influential in the Australian case.

A Framework for Discussing Non-Human Invention

For the most part, I use the word ‘invention’ in this article to refer to a process, as opposed to the product (‘an invention’) of that process.  Thus an ‘inventor’ can be identified as one who (successfully) executes the process of ‘invention’.  I make no assumptions, a priori, about the nature of any entity that may qualify as an inventor.  It is only by agreeing on what qualifies as ‘invention’ that we can determine who – or what – we are prepared to recognise as an ‘inventor’.  As I will discuss later, I believe that the broad consensus at the present time is that ‘invention’ is an activity that inherently involves human mental activity, insights, and creativity.  For anyone who subscribes to this view, it necessarily follows that an ‘inventor’ must be a human being.

In an effort to create a more formal structure for discussing the issues, I propose a framework in which there are three dimensions to the process of invention:

  1. philosophical dimension – what range of activities are we, as sentient human beings, prepared to recognise as having the necessary qualities of invention?
  2. legal dimension – what does the law recognise as constituting invention, for the purposes of determining whether patent rights may be conferred upon some legal entity (whether that be the inventor or some other party)?
  3. policy dimension – what activities do we, as a society, want to incentivise by recognition as constituting invention?

Abbott’s starting point appears to be the ‘policy’ dimension of this framework.  In his paper ‘I Think, Therefore I Invent: Creative Computers and the Future of Patent Law,’ Boston College Law Review, vol. 57 no. 4, 2016 (available on SSRN), he states:

This Article addresses whether a computer could and should be an inventor for the purposes of patent law as well as whether computational inventions could and should be patentable. It argues that computers can be inventors because although AI would not be motivated to invent by the prospect of a patent, computer inventorship would incentivize the development of creative machines. In turn, this would lead to new scientific advances.

His thesis is thus that there is a policy argument for recognising computational invention, and therefore that the law (i.e. the legal dimension) should be adapted, if necessary, to serve this policy objective.  As far as I can tell, however, Abbott is not much concerned with the philosophical dimension of the above framework.  If he were, then I would suggest that he would have a greater interest in the question of whether DABUS – and, indeed, other computational systems that he discusses in ‘I Think, Therefore I Invent…’ – is indeed capable of invention.  Instead, he is prepared to accept at face value Thaler’s claim that it is, in order to advance his project.

The stated object of the Artificial Inventor Project is to secure intellectual property rights for the autonomous output of artificial intelligence.  The means to this end, so far, has been the filing of patent applications in multiple jurisdictions naming Thaler’s DABUS machine as inventor.  As I have noted, this effort has so far been unsuccessful.  The recent decision of the English High Court is the first judicial consideration of issues relating to machine inventorship.

DABUS: Neither a Person nor an Inventor Under UK Law

Unsurprisingly, the English High Court (in common with the UK, European, and US patent office officials who have also considered the matter so far) has focussed primarily on the legal dimension of the framework. 

Regarding the court’s lack of interest in the philosophical dimension, the judge is clear, at [6], that he makes no findings with regard to DABUS’ capacity for invention, and proceeds on the basis that the asserted facts are correct:

First, the Patent Form 7 contains a number of factual assertions regarding DABUS, its capabilities and its role in the inventions specified in the Applications. None of these factual assertions has been tested in these proceedings. I proceed on the assumption – as has everyone in these proceedings – that these factual assertions are true. However, I should stress that I am making an assumption in Dr Thaler's favour, not a finding.

It is also clear that nobody is trying to assert that DABUS should be treated as a person.  This approach to the matter relates closely to the legal dimension, because there is no realistic prospect that a court – at least at this point in history – would consider a machine to have legal (or actual) personality, given the broad range of consequences that would follow from such a finding.  As the court notes, at [34]:

It is common ground that DABUS is not a person, whether natural or legal. DABUS is not a legal person because (unlike corporations) it has not had conferred upon it legal personality by operation of law. It is not a natural person because it lacks those attributes that an entity must have in order to be recognised as a person in the absence of specific (statutory) legal intervention.

In determining whether or not DABUS can be considered an ‘inventor’ within the meaning of the term in the UK Patents Act 1977, the court found at [40] that:

It is quite clear from the statutory scheme contained in the Patents Act 1977 that – whatever the meaning of the term "inventor" – a patent can only be granted to a person.

Since paragraph 7(2)(a) expressly provides that a patent may primarily be granted to the inventor (or joint inventors), the judge thus determined, at [46]:

As I have noted, I am proceeding on the basis that DABUS has "invented" the inventions the subject of the Applications. Nevertheless, I conclude that DABUS is not, and cannot be, an inventor within the meaning of the 1977 Act, simply because DABUS is not a person.

Further, the court rejected Thaler’s argument that he was entitled to apply for, and be granted, a patent on the basis of his ownership of DABUS, with the judge stating, at [49]:

Even if I accepted that DABUS was capable of being an "inventor" – which, for the reasons I have given, I do not – Dr Thaler's application would be hopeless, because DABUS would – by reason of its status as a thing and not a person – be incapable of conveying any property to Dr Thaler. In short, the ability to transfer, which DABUS lacks, is fatal to Dr Thaler's contentions. The same point can be put in a different way: because DABUS is a thing, it cannot even hold property, let alone transfer it.

As a result, Thaler’s appeal was dismissed.  The court has left it to the parties to frame an appropriate order, but presumably this will be that the UK IPO’s original decision should stand, such that Thaler’s applications will be deemed withdrawn due to his failure to identify a valid inventor, or a legitimate basis for his entitlement to apply for patents, within the 16 month time limit provided.

Problems with the Artificial Inventor Project

I disagree profoundly with Abbott’s entire project, and his efforts to gain recognition for machine invention.  I think his policy thesis is wrong.  I disapprove of the methods he is employing to raise the profile of the issue, and to push for changes in the law.  I have written, and spoken, previously about my scepticism of claims that DABUS – or any machine, for that matter – has been developed with the capacity for invention.  More generally, however, I believe that Abbott’s failure to engage meaningfully with the philosophical dimension of inventorship is a flaw in his approach that undermines his project.

Policy Perspective – Do We Really Need Machine Inventors?

The problem with Abbott’s policy position is easiest to deal with.  He posits, as I have noted, that the potential benefits of the further development of ‘creative machines’ are such that additional incentives are desirable, and in particular that this should be achieved by ensuring that the products of such machines should be patentable to the same extent that they would be if devised by a human inventor.

There is no evidence for this proposition.  Abbott certainly does not offer any in his published work.  Indeed, the value of the patent system as it presently exists – with its limitation of inventorship to ‘mental acts’ performed by humans – is already hotly contested within economic circles.  As the economists at the Australian Productivity Commission argued in Chapter 7 of their 2016 report on the nation's IP arrangements, there are (in their view) already too many ‘low-value’ patents, which ‘impede innovation by frustrating the efforts of follow–on innovators and researchers’, and which may be ‘deliberately used as a strategic tool for stalling or excluding market entry’.  As a result (they believe) that ‘[c]onsumers incur much of the costs from a multitude of low-value patents, either because new products and services are not developed, or because the higher costs of innovation are passed on.’

While I see little evidence that the patent system is as unbalanced as the Productivity Commission’s economists have asserted, I do believe that it is important to get (and keep) the balance right.  In my view, therefore, we should demand the production of persuasive evidence before opening the door to allowing patents on automated invention.

As an alternative to Abbott’s utopian world of productive and socially valuable machine invention, I can readily envisage a diametrically opposed outcome.  Arguably, a machine with virtually unlimited capacity to churn out inventions without the further exercise of any human ingenuity, judgment, or intellectual effort, is far more likely to find application as a ‘patent thicket generator’, if its products can be patented.  In this dystopian scenario, a small number of corporations with advanced AI programs and deep pockets could simply overwhelm the patent system in an ‘arms race’ to tie up rights in key fields of technology.  Of course, I too have no evidence.  But, then, I am not the one arguing for an expansion in the legal meaning of ‘inventor’.

Legal Perspective – Is it Right to Force the Issue in the Courts?

The approach being taken by Abbott and the Artificial Inventor Project is essentially to force the issue of machine invention via the provocation of having Thaler file patent applications in various jurisdictions, naming DABUS as inventor.  Abbott (presumably) knows perfectly well that the established law and practice in these jurisdictions requires a human inventor.  The adverse decisions received so far from the European Patent Office EPO), the UK IPO, and the USPTO, will have come as no surprise.  While the English High Court has already heard and decided an appeal from the UK IPO decision, an appeal was filed in the EPO case on 14 March 2020, and in the US case (in the Eastern District of Virginia) on 6 August 2020.  Further appeal to the higher courts may well be pursued in the UK.

Is this quixotic exercise an appropriate use of the limited resources of administrative and judicial bodies that exist primarily to resolve issues and disputes raised by genuine users of the patent and legal systems?  I do not think so.  These are not ‘test cases’ that we need to have – certainly not at this point in time, given the absence of evidence of any need or justification for a change in the law to recognise machine invention.

The law is what it is, and these cases will not change it.  Any change is ultimately a matter of policy that will need to come from the legislature.  I would respectfully suggest that Abbott’s time and energy as an academic researcher would be more appropriately expended on the policy dimension, i.e. by studying and modelling the potential social and economic impacts of granting patents based on machine invention to try to determine whether there is a sound policy argument for such a change.  If there is, then the case should be made to governments, not the courts.

Through these provocations, Abbott is doing little more than pushing his own opinion.  Regardless of what I think, however, his approach seems to be working.  Consultations on patents and artificial intelligence have been conducted, or are currently underway, at the UK IPO, the World Intellectual Property Organization (WIPO), and the USPTO, all of which encompass questions around machine invention that have been explicitly or implicitly prompted by Abbott’s project.  Ultimately, however, opinions alone will not result in changes to the law.  And, based on submissions to the various consultations, there appears to be little support for either the proposition that machines are capable of invention, or that there is any need for legislative action, from influential stakeholders such as IBM and the IP Owners Association (which counts Apple, Google, HP, IBM, Microsoft, Oracle, and Uber among its members).

Philosophical Perspective – Will Society Embrace Machine Inventors?

While the philosophical dimension encompasses fascinating questions such as whether, and in what circumstances, a machine can be said to possess intelligence, to ‘think’, or to be ‘conscious’, that is not the aspect that primarily concerns me here.  It is not so much Abbott’s failure to critically evaluate Thaler’s claims regarding DABUS’ creative capacity that I see as a flaw in his approach.  I regard this is merely a symptom of a more general failure to engage with the concept of ‘invention’, not only within the patent law, but within society at large.

As I have said, patent laws (like all laws) should be balanced and proportionate in view of their policy objectives.  Additionally, however, they also need to be perceived to be balanced and proportionate.  Laws that are out of step with community standards and expectations are liable to come under criticism, to lose support and authority, and thus to have their effectiveness undermined.  We have some relatively recent experience of this in relation to the Australian patent law, in the debate over so-called ‘gene patents’.

The debate over patenting of isolated gene sequences simmered uneasily in Australia for years.  Opponents – of whom there were a number, with substantial political clout and community support – were able to generate considerable adverse publicity, push legislative efforts, and initiate parliamentary enquiries in their efforts to outlaw gene patents.  As I wrote way back in the early days of this blog (i.e. 2011), the failure by supporters of gene patents (including patent attorneys) to engage with genuinely-felt ethical concerns within the community significantly undermined efforts to contribute constructively to the debate.  The fact that the law was seemingly settled did not deter those who believed it to be wrong.  Despite the Patent Office defending its practice of granting patents on isolated genes, and a total of six Federal Court judges (one at first instance, and five on appeal) upholding that practice, gene patent opponents were ultimately successful at the High Court in establishing that they had been right all along.

While I have not conducted any studies – although perhaps somebody should – I suspect that there may be little community support for allowing machines to be named as inventors on patent applications.  I expect that most people associate ‘invention’ with certain qualities – such as ‘creativity’, ‘initiative’, and ‘innovation’ – that they believe to be uniquely ‘human’, and which distinguish us from animals (and machines).  From this perspective, a law allowing a machine to be considered an ‘inventor’ is either elevating an unthinking automaton to ‘human’ status, or in some sense recognising artificial intelligence as somehow equivalent to human intelligence.  Many people would regard this as either insulting, or threatening.

While the progress (or lack thereof) of the DABUS patent applications has generated great interest within the IP professions, and some coverage by technology and business media, it has thus far received relatively little mainstream coverage.  To the extent that members of the general public are aware of efforts to have a machine named as an inventor, I imagine that most would currently regard this as a curiosity or a stunt.  However, genuine concerns might arise if a government started to discuss seriously the possibility of granting machines the ‘right’ to be identified as inventors.

In the debate over gene patents in Australia, despite deep-felt concerns within the community, submissions to formal consultations were significantly in favour of allowing patents on isolated genetic sequences.  This is because contributions to such consultation processes are always dominated by organisations and individuals with specific interests, and rarely reflect views from the wider community.  Contributions to the various consultations conducted by authorities such as WIPO, the UK IPO, and the USPTO will inevitably suffer from the same selection bias.

In my view, a change as profound as providing legal recognition of non-human inventorship should not occur unless it has broad community support.  The risk, otherwise, is that it could reduce respect for patent laws, and for the patent system generally.  That is why I believe that Abbott’s project is undermined by his failure to engage with the philosophical dimension of invention.

Conclusion – Automation’s Challenge to the Patent System

Rapid advances in artificial intelligence – or, more accurately and specifically, machine learning – undoubtedly raise important questions for the patent system.  But sophisticated software that can assist humans in developing novel products and solutions is not new.  ‘In silico’ design, analysis, experimentation, and optimisation is now used across almost all fields of technology.  There is no clear dividing line between the assistance provided by conventional ‘procedural’ programs (i.e. those that are explicitly coded to perform desired tasks), and that provided by machine learning systems (i.e. those that comprise structures ‘trained’ to perform desired tasks using available data).  Both approaches provide human users with tools that enable them to produce better solutions in less time and at lower cost than could be achieved without computer assistance.

Such tools potentially enable even the least imaginative technician to generate novel solutions that are ‘inventive’ in the sense conventionally required by the patent laws: that no non-inventive person of ordinary skill in the art would regard them as obvious over the available prior art.  On the other hand, applying common knowledge to execute a series of routine steps using widely-available tools (computational or otherwise), and without the exercise of any inventive ingenuity, would not generally be regarded as providing an inventive step.

There is, however, no requirement under existing patent laws to disclose how an invention was derived.  Indeed, the US law in particular expressly provides that ‘patentability shall not be negated by the manner in which the invention was made’ (35 USC § 103).  This provision was enacted to overrule the ‘flash of genius’ doctrine that had been established by the Supreme Court in 1941, recognising – to use the famous words of Thomas Edison – that ‘genius’ is often ‘one percent inspiration and 99 percent perspiration’.

On balance, then, no rational patent applicant would actually disclose any significant role played by computational tools in the development of an invention, if that could lead to a rejection on grounds of obviousness.  We therefore find ourselves in a potential Catch-22 situation: legally recognising machine inventors may lead to a proliferation of patent applications generated by automation, while not doing so may encourage applicants to conceal the role of automation, enabling patents to issue that should perhaps be regarded as lacking inventiveness.

In circumstances where the ‘99 percent perspiration’ can be delegated to a machine, the one percent inspiration that originates in the human mind has never been more central to the principles underpinning the patent system.  Long may it remain an essential requirement for patentability!


Post a Comment

Copyright © 2014
Creative Commons License
The Patentology Blog by Dr Mark A Summerfield is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Australia License.