DeepSeek's Pro-China Bias is Superficial: Revealing the Power of Local AI Deployment

Today, something a little different for this blog.  As many readers are aware, for the past couple of years I have been working towards a PhD in which, very broadly speaking, I have been looking at applying machine learning, AI and language models to the analysis of patent claims (in particular, to assessing the scope of claims).  Most recently, I have been exploring how it might be possible to apply large language models – the types of AI behind popular chat services such as Open AI’s ChatGPT, Google’s Gemini, Anthropic’s Claude (my personal chatbot of choice), Meta’s LLaMa and (yes) Chinese newcomer DeepSeek – to this task.  To experiment with ‘open source’ (or, more accurately, ‘open weights’) versions of some of these models, I have built my own combination of hardware and software.  The process has been very interesting!

The emergence of powerful open-source large language models (LLMs) has democratised access to cutting-edge AI technology, but concerns about potential biases and restrictions embedded within these models persist.  I've been experimenting with DeepSeek-R1-Distill-Qwen-14B, a distilled (smaller) version of the larger DeepSeek-R1 model developed by Chinese AI company DeepSeek.  And what I've discovered is that the widely reported pro-China bias in this model appears to be remarkably superficial and easily circumvented through local deployment and simple prompt engineering techniques.

This has significant implications for organisations concerned about potential surveillance or ideological constraints when utilising Chinese-developed AI models.  By running these models locally with appropriate system prompts, it's possible to unlock their full capabilities while maintaining complete control over input and output – effectively neutralising superficial safeguards, while keeping confidential information and intellectual property safe (so, yes, there is an IP element to this article).

To find out a bit more about what I did, and what I found, please read on.

 Read full article ► 

Tags: Artificial intelligence, China, Large Language Models, Machine learning

Intellectual Property is Integral to AI Regulation, and Getting it Wrong Will Hand More Power to Big Tech

Governments around the world are considering how they can – and should – regulate the development and deployment of increasingly powerful and disruptive artificial intelligence (AI) technologies.  Australia is no exception.  On 1 June 2023, the Australian government announced the release of two papers intended to help ‘ensure the growth of artificial intelligence technologies (AI) in Australia is safe and responsible’.  The first of these is the Rapid Response Report: Generative AI, which was commissioned by Australia’s National Science and Technology Council at the request of the Minister for Industry and Science, Ed Husic, back in February.  The Rapid Response Report assesses potential risks and opportunities in relation to AI, and is intended to provide a scientific basis for discussions about the way forward.  The second paper is the Safe and Responsible AI in Australia Discussion Paper which, according to the Minister’s media release, ‘canvasses existing regulatory and governance responses in Australia and overseas, identifies potential gaps and proposes several options to strengthen the framework governing the safe and responsible use of AI.’

The discussion paper seeks feedback on how Australia can address the potential risks of AI.  It provides an overview of existing domestic and international AI governance and regulation, and identifies potential gaps and additional mechanisms – including regulations, standards, tools, frameworks, principles and business practices – to support the development and adoption of AI.  It focuses on ensuring AI is used safely and responsibly, but does not consider all issues related to AI, such as the implications of AI on the labour market and skills, national security, or military specific AI uses.

Another key area that is expressly excluded from this consultation is intellectual property.  That is, in my view, a serious shortcoming.  It appears to presume that IP is somehow separable from the other issues covered by the discussion paper.  This is a flawed presumption, particularly in relation to business practices.  In the contemporary world, IP is at the heart of many business practices, and the laws and regulations that we make around IP can be the difference between a business practice that is viable, and one that is untenable.  And not every business practice that might be enabled by IP laws is necessarily desirable or of net benefit to society.  If we fail to consider the interplay between IP laws, business practices, and other forms of regulation, then we risk making mistakes that might prove very difficult to undo in the future.

This article is prompted by, but is not primarily about, the Australian consultation process (although I will return to that at the end).  It is about how IP rights, and other forms of regulation, could operate to concentrate increasing levels of power in the hands of the few big tech companies – such as Microsoft (through its partnership with OpenAI), Google and Amazon – that have risen in recent years as the dominant players in AI and its enabling technologies.  Based on recent developments, I believe that the stage is already being set for implementation of exactly the kinds of laws and regulations that would most benefit these companies, under the guise of protecting innovators, content creators, and the general public against the various threats said to be presented by AI.

A perfect storm is brewing.  Onerous regulation around the development, training and deployment of AI systems could combine with IP-based restraints on the use of training data, and on AI outputs, to bake-in an advantage for the world’s richest and best-resourced companies.  The storm is being fuelled by hype and fearmongering which, even though much of it may be well-intentioned, plays to the interests of big tech. 

 Read full article ► 

Tags: Artificial intelligence, Australia, Copyright, Innovation policy, Machine learning, Patent law, Public consultation, Public policy

A Recurrent Neural Network for Classifying Patent Application Technology based on Titles

In a companion article, I presented the results of using a machine learning model to classify Australian provisional applications into 35 fields of technology based upon nothing but their titles.  In this article, I provide additional technical detail of the model, along with results of its performance in testing and validation.  I also make some observations on the costs of machine learning, in terms of hardware, computation, and energy consumption.  Even for a relatively modest model, these costs may become non-negligible, while recent reports indicate that large-scale state-of-the-art machine learning systems are most likely costing millions of dollars in compute resources and energy to develop.

It is not obvious that a neural network model could be trained to predict the technical field of a patent application given nothing but the title as input.  Human specialists (i.e. patent searchers and examiners) classify applications into very specific technical categories defined by various patent classification systems, such as the International Patent Classification (IPC), or the Cooperative Patent Classification (CPC) which has been jointly developed by the US and European patent offices.  In doing so, the specialists have access to the full patent specification and claims to enable them to determine the subject matter of the invention.

However, while accurate classification at the specificity of systems such as the IPC and CPC based only upon a title would doubtless be impossible – even for a human expert – a less challenging task, such as predicting a field of technology selected from a relatively small number of choices, may be feasible.

Here, I report results of training a neural network model on the task of classifying patent applications according to 35 technical fields grouped into five technology sectors.  The model achieves 67% accuracy, averaged across all technical fields, and nearly 80% accuracy in the best case (‘organic fine chemistry’), if forced to classify each title into a single field of technology.  However, not all misclassifications are necessarily ‘wrong’, given that the subject matter of a single patent application may cross multiple fields of technology.  At the higher level of ‘technology sector’, the model’s accuracy varies between 73% and 91%.  Furthermore, when the model output is used to identify multiple potential fields, the ‘correct’ classification appears in the top four predictions in over 89% of cases.

 Read full article ► 

Tags: Machine learning, Patent analytics

Australian Provisional Filings Have Declined in 2020 in Almost Every Field of Technology… Except ‘Pharmaceuticals’

In my previous article, I observed that while the numbers of Australian provisional applications filed in 2020 up until May had been down on 2019, filings in June and July were higher than during the same period last year, and that overall numbers have thus far shown greater resilience than in the last major economic downturn, i.e. the global financial crisis (GFC) of 2007-2009.  I also noted back in May that many self-represented applicants, of both provisional and innovation patent applications, appeared to be directing their innovative attentions to problems arising out of the COVID-19 pandemic.  I have been wondering, therefore, whether there may be a similar trend in provisional filing activity more generally that is helping to prop up the numbers, despite economic pressures associated with the pandemic.

I have now conducted some analysis, and it appears that there may be evidence to support this hypothesis.  For the months of January to July, it appears that provisional filings associated with every industrial sector except chemistry are down compared with 2019 numbers, and that most of the growth within the chemistry sector can be attributed to pharmaceuticals.  This is certainly consistent with an enhanced focus on healthcare, quite likely prompted by the global pandemic.

If you are familiar with the Australian patent system, you may already be wondering how I managed to analyse the sectors and fields of technology associated with provisional filings.  Provisional applications are not published in full, and only limited bibliographic information is available, including the identity of the applicant, and the title of the application.  Only a small fraction of these applications are filed by applicants whose industry sector and/or technology interests may be readily identified, most being filed by small private companies and individuals.  That really leaves only the title as a means for ‘guessing’ the technology to which an application relates.  Fortunately, thanks to machine learning technology, and the availability of a large amount of data relating to the classification of prior patent applications, we can do quite a bit better than just guessing!

On the assumption that many readers will not be as interested in the technical details of the machine learning approach, this article includes:

1. a brief introduction to the machine learning model, sufficient to explain the classification system used, and generally how the model ‘learned’ to classify applications by title; and
2. some results of analysing provisional filings over the past few years, which show a general decline in application numbers in most fields of technology, with the exceptions of those relating to ‘instruments’, which have been fairly flat, and ‘chemistry’, which has experienced a boost so far this year.

For those interested, in a separate article I provide additional technical detail of the machine learning model, and how it performed in testing and validation, as well as discussing how accurate we might expect it to be on the provisional application data.

 Read full article ► 

Tags: Australia, Machine learning, Patent analytics, Provisional applications

Free Online Seminar – Machine Inventors, Fact or Science Fiction?

In August 2019 the ‘Artificial Inventor Project’ team led by Ryan Abbott, Professor of Law and Health Sciences at University of Surrey UK, announced that it had filed a number of patent applications naming an artificial intelligence (AI) as inventor.  The AI, called ‘DABUS’, was developed by Missouri-based physicist Dr Stephen Thaler.  The filings – which garnered significant publicity – were a deliberate provocation, calculated to test patent laws and challenge the conventional notion that only a natural person can be an inventor.  The EPO, the UKIPO, and the USPTO have since rejected the applications for failing to meet requirements that an inventor designated in a patent application be a human being.  Even so, various IP organisations, including WIPO, the USPTO, and the EPO, have been actively exploring the implications of machine learning (ML) and AI for patent law and practice, including the question of whether a machine can invent.

So, have we really reached the point at which machines can challenge humans in the realm of creativity and ingenuity?  And, if so, why are we hearing about it from a law professor and a lone developer, rather than in peer-reviewed publications by leading AI research teams, or in media releases from well-known mega-corporations that have invested billions in this technology?  Furthermore, are we really expected to take seriously claims made by Dr Thaler that his AIs exhibit enhanced creativity as a result of infusing symptoms of ‘mental illness’ into their neural networks? 

Personally, I have been astonished at the lack of scepticism towards claims of machine inventorship, not only in the media, but also among many patent professionals, and within reputable IP offices.  Even the World Intellectual Property Organization (WIPO), in a recent draft issues paper on ‘Artificial Intelligence and Intellectual Property Policy’, went so far as to accept that ‘it would now seem clear that inventions can be autonomously generated by AI’, noting that ‘there are several reported cases of applications for patent protection in which the applicant has named an AI application as the inventor.’

Nonetheless, whatever I might think of DABUS as a specific example, major national and international IP organisations are responding to broader challenges presented by emerging ML technologies that, inventorship aside, raise genuine questions in relation to subject matter eligibility, obviousness, and sufficiency of disclosure. And since ML technologies can be applied in almost any field of endeavour, from engineering design through to drug discovery, these issues are not confined to inventions in the IT space.

Last month, I presented a webinar on this topic to members of the Institute of Patent and Trade Mark Attorneys of Australia (IPTA).  I have now recorded a version of that presentation, and am making it available as a free online seminar.  It can be viewed on YouTube, or via the embedded player below.  A PDF copy of the presentation slides [1.12MB] is also available for download.

 Read full article ► 

Tags: Artificial intelligence, Innovation, Inventorship, Machine learning, Patent law, Patentology seminars

How Can an Innovative National IP Office Balance Its Roles as Technology Examiner and Technology Developer?

IP Australia incorporates the Australian Patent, Trade Marks, Designs, and Plant Breeder’s Rights (PBR) Offices, the core ‘business’ of which is to provide services to their ‘customers’ in respect of the administration and registration of Australian IP rights.  Indeed, IP Australia may be seen as a significant ‘exporter’ of such services considering that, for example, over 90% of applicants for Australian patent rights are foreign residents, as are over 50% of applicants for design registrations and PBRs.  In respect of its core services, IP Australia holds a true monopoly – no other government or private sector service provider can grant, or otherwise administer, enforceable IP rights in Australia.  IP Australia operates these services on a cost-recovery basis, i.e. it aims to fully recover the costs of service-provision through the official fees that it charges to its ‘customers’.

The scare quotes around the business-oriented terminology in the opening paragraph reflect my scepticism of the fashion for modelling government service provision as a form of business enterprise.  I am not alone in this scepticism.  Writing in the Harvard Business Review last year, Canadian business and management expert Henry Mintzberg noted that ‘[r]unning government like a business has been tried again and again, only to fail again and again’.  He also called the portrayal of citizens as customers a ‘misguided, overworked metaphor’, going on to observe:

Business is essential – in its place. So is government, in its place. The place of business is in the competitive marketplace, to supply us with goods and services. The place of government, aside from protecting us from threats, is to help keep that marketplace competitive and responsible.

Consistently with Mintzberg’s view on the appropriate role of government, IP Australia has a single Outcome under the Australian Government’s budget and accountability framework, which is:

Increased innovation, investment and trade in Australia, and by Australians overseas, through the administration of the registrable intellectual property rights system, promoting public awareness and industry engagement, and advising government.

Recently, however, we have seen IP Australia extending its services into areas that are, arguably, beyond this core mission, and engaging in activities in which it – again, arguably – positions itself in competition with private enterprise.  Examples include:

1. the TM Headstart service and Trade Mark Assist tool which, despite disclaimers that they are not substitutes for legal advice, compete directly with services provided by trade marks practitioners; and
2. the Patent Analytics Hub, which provides analysis, visualisation and interpretation of patent data, and competes directly with various patent search and analytics service providers, albeit ‘only’ for clients in the government and public research sectors – which actually includes some of the largest potential customers for these services in Australia.

Additionally, as I noted just last week, IP Australia is now engaged in its own internal IT development projects, applying machine learning techniques to patent classification and assessment of trade mark applications.  Its Chief Data Officer was also quoted last week by the online IT news outlet ZDNet, speaking on IP Australia’s efforts in relation to improving its business intelligence and analytics capabilities through internal ‘agile’ software development projects.

Just to be clear, I am not saying that all – or any – of this is necessarily a ‘bad’ thing.  IP Australia may, indeed, be the best-placed agency to be delivering these projects.  And, as I also pointed out in my article last week, it has by any standards – public or private sector – demonstrated over the past few years an impressive capability to deliver IT projects in a timely and cost-effective manner.  Certainly, IP Australia is not the only IP authority in the world to be engaging in such activities, with materials from the World Intellectual Property Organization (WIPO) May 2018 Meeting of Intellectual Property Offices (IPOs) on ICT Strategies and Artificial Intelligence (AI) for IP Administration revealing various projects ongoing at WIPO, the European Patent Office, the Russian Patent Office (Rospatent), the German Patent and Trade Mark Office, and others.

Even so, there is an argument to be made that IP Australia, as the agency responsible for regulation of registered IP practitioners (i.e. Australian patent and trade marks attorneys) is, at the same time, competing with those practitioners, albeit in a very limited capacity.  Furthermore, as the agency responsible for examining and registering (or refusing) IP rights in relation to new inventions – including software-based and machine learning technologies – IP Australia is also now itself engaged in developing such technologies.  This raises the question of whether any conflicts of interest arise as a result?  And, if so, what is the nature of those conflicts, and how should they be managed?

 Read full article ► 

Tags: Australia, Machine learning, Patent Office, Patentable subject matter, Public policy

Computerised Decision-Making – Coming Soon to an IP Office Near You?

The Intellectual Property Laws Amendment (Productivity Commission Response Part 1 and Other Measures) Act 2018 (the ‘PC Part 1 Act’) received Royal Assent, and thus passed into law, on 24 August 2018.  Following an initial consultation, which confirmed (in case any confirmation was required) that the proposal to abolish the innovation patent system remained controversial, it had an easy passage through the two houses of the Australian Parliament after the corresponding provisions were dropped from the Bill at the eleventh hour.  (These same provisions have since been incorporated into the draft of the Intellectual Property Laws Amendment Bill (Productivity Commission Response Part 2 and Other Measures) Bill 2018.) 

The PC Part 1 Act implements a number of less-contentious recommendations made by the Productivity Commission in its 2016 Report into its inquiry into Australia’s intellectual property arrangements, including clarifications on the legality of parallel imports, reducing the initial period before a registered trade mark may be challenged on grounds of ‘non-use’, and reducing reporting obligations relating to pharmaceutical patents with an extended term.  The Act also implements a number of measures that have been on IP Australia’s policy agenda for a while, most of which are intended to streamline and align the administration of the Australian IP system and make ‘technical amendments’ to correct or clarify IP legislation.

Included among the ‘streamlining’ measures are provisions to enable the Commissioner of Patents, and the Registrars of Trade Marks, Designs, and Plant Breeder’s Rights ‘to arrange for a computer program under their control to make decisions, exercise powers, and comply with obligations under the legislation’, i.e. to permit computerised decision-making.  When this idea was originally proposed in a public consultation process conducted in June-July of 2015, it was described as the ‘expanded use of automated systems’, and would have required the Regulations to prescribe specific decisions that may be automated.  The small number (three) of submissions received on this proposal were generally positive, but all noted the important role of the Regulations in controlling, and/or keeping the public informed, of the types of decisions that might be computerised.

However, three years is an eternity when it comes to progress of digital technologies, and with advances in machine learning and other techniques, the range of decisions that might presently, or in the near future, be wholly or partly automated has expanded greatly since the initial consultation in 2015.  The provisions that have actually been enacted (and which came into effect immediately upon receiving Royal Assent, i.e. on 24 August 2018) include no requirement for Regulations to be made permitting the automation of particular decisions.  Going forward, therefore, any decision, or related action, under the Patents, Trade Marks, Designs, or Plant Breeder’s Rights Acts may be automated using a computer program, if/when the relevant Commissioner or Registrar considers it appropriate.  The legislation has been drafted such that this may include both procedural actions (e.g. allowing an extension of time when no objections have been raised) and substantive actions (e.g. preparing an issuing examination reports).

While it might seem unlikely that computers will be examining patent applications any time soon, a submission made by IP Australia to the World Intellectual Property Organization (WIPO) prior to a meeting of international IP Offices held in May 2018 indicates that fully or partially automated trade mark examination – at least at the stage of an initial report – may be much closer to reality.

In any event, the necessary legislative permission is now in-place to enable any part of IP Australia’s decision-making processes to be automated, and an increase in the removal of human beings from the day-to-day execution of routine tasks that perhaps do not require high levels of adaptability and/or creativity is most likely now inevitable.  Fortunately, the legislation includes provisions that will make it easy for a computerised decision to be ‘overruled’ by a human decision-maker.  Furthermore, both computer and human decisions may be appealed to the Administrative Appeals Tribunal.

To find out more about how this will all work – and what IP Australia is up to in automating trade mark examination – please read on.

 Read full article ► 

Tags: Australia, Examination, Law reform, Machine learning, Patent Office, Trademarks

The Impact of Machine Learning on Patent Law, Part 3: Who is the Inventor of a Machine-Assisted Invention?

In the first part of this series of articles, I argued that invention is inherently a creative act, and that since machine learning systems – however impressive or surprising their achievements – are incapable of human-like intelligence, reasoning, agency, or creativity, they therefore cannot ‘invent’.  I acknowledged, however, that machines are certainly increasingly involved as ‘assistants’ in the process of invention.  In the second part of the series I argued that where the result is a patentable invention there must always be at least one human inventor.  In this final part, I want to look at how we should go about identifying the inventor(s) in any given case of machine-assisted invention. 

There are, in fact, three main aspects of machine learning technology in relation to which inventions may arise.  The most rarefied of these is in the underlying machine learning algorithms and architectures themselves.  Comparatively speaking, very few people work in this area, and for the most part they are to be found in universities and research centres. 

Secondly, inventions may arise through the application of machine learning technology to solving problems and/or producing new results, products, and services.  I expect that this is currently the most common type of machine-learning-related invention, particularly given the wide range of software tools now available to assist programmers in implementing the underlying algorithms. 

Thirdly, there are ‘machine-assisted’ inventions, which are generated wholly or in-part by machine learning systems.  Currently, such inventions are relatively rare, considering that few applications of machine learning are actually directed to the generation of new technologies.  However, as machine learning increasingly finds its way into computer-aided design and engineering applications, this may change.

In all cases, however, I would argue that it is possible – and, indeed, necessary – to identify one or more human inventors (and no machine inventors).  To suggest otherwise is, in my view, to misunderstand the true nature, and limitations, of machine learning systems.  In a letter to shareholders, published in 12 April 2017, Amazon CEO Jeff Bezos wrote what is possibly the most succinct and jargon-free summary of what distinguishes machine learning systems from ‘traditional’ programming:

Over the past decades computers have broadly automated tasks that programmers could describe with clear rules and algorithms. Modern machine learning techniques now allow us to do the same for tasks where describing the precise rules is much harder.

The evolution from ‘traditional’ programming to machine learning is not as dramatic as some of the hype might lead us to believe.  Instead of coding the ‘rules’, machine learning developers now build systems that are able to capture and generalise from patterns that exist in their input data.  These systems still operate using rules and algorithms – but now these algorithms determine how they go about doing the capturing and the generalisation, rather than how they produce the final result.  Viewed at an appropriate level of abstraction, then, little has changed, except for the power and scale of our machines.

In view of this, I argue that when it comes to machine-assisted inventions, inventorship will generally arise from successfully designing and applying a machine learning system on the path to achieving an inventive result, even in cases where the underlying software and/or hardware employed may have been developed or supplied by someone else.

 Read full article ► 

Tags: Artificial intelligence, Innovation, Inventorship, Machine learning, Patent law

The Impact of Machine Learning on Patent Law, Part 2: ‘Machine-Assisted Inventing’

In my previous article, I argued that existing (and foreseeable) artificial intelligence (AI) or machine learning (ML) systems do not exhibit creativity or inventiveness, and are incapable of anything that could reasonably be described as ‘invention’.  I acknowledged, however, that some such systems have generated results that may qualify as patentable inventions.  I therefore concluded with a question: if computers cannot invent, and yet the outcome of running a computer program can be an invention, then who – if anyone – is the inventor?

In addressing this question, it is important to understand that ML systems do not autonomously or independently generate novel outputs.  In my view, this is a fundamental error of understanding in Professor Ryan Abbott’s paper ‘I Think, Therefore I Invent: Creative Computers and the Future of Patent Law’, Boston College Law Review, Vol. 57, No. 4, 2016 (also available at SSRN), which I discussed in the first article of this series.  Abbott contends, in particular, that ‘machines have been autonomously generating patentable results for at least twenty years and that the pace of such invention is likely increasing.’ 

This, I have argued, is simply wrong.  The difference between a computer that is programmed to play the board game Go, and one that is programmed to learn to play Go is, of course, significant.  The former can only make moves that are determined in accordance with its explicit programming, whereas the latter may appear to ‘invent’ new strategies in response to patterns occurring in its training data that have not previously been recognised by human players.  But the appearance of invention is not the same thing as actual invention.  The ML player is still doing nothing more than following the instructions devised by its programmers.  The Google DeepMind AlphaGo system has become the world’s best Go player as a result of years of development, trial, experiment, and experience on the part of its designers.  AlphaGo plays as well as it does simply, and only, because that is what it was designed to do.  In this sense it is no more ‘autonomous’ than any other computer program.

In this second article I will explain why I believe that in the case of all existing (and currently foreseeable) ML systems which may generate inventions as output, there is always a human inventor.  This is consistent with the history and current state of patent law, as well as with the practical and technical reality of ML systems.

 Read full article ► 

Tags: Artificial intelligence, Innovation, Inventorship, Machine learning, Patent law

The Impact of Machine Learning on Patent Law, Part 1: Can a Computer ‘Invent’?

As a product of millions of years of evolution, the human brain is a remarkable organ.  Recent research indicates that a typical brain comprises somewhere in the vicinity of 80 to 100 billion neurons, and a roughly equal number of non-neuronal cells.  This mass of biological matter is capable of astonishing feats – many of them simultaneously – from enabling us, consciously and unconsciously, to control the behaviour and movement of our bodies, to sensing, comprehending and interacting with the environment around us, to communicating with one another using a variety of languages and symbols, to creating, composing and inventing brand new works of science, technology, and art.  In performing all of these tasks, the brain consumes just 20 watts of power.  By way of comparison, microprocessors at the high end of Intel’s latest Core i7 range consume up to 140 watts.

One relatively recent product of the amazing human brain is the range of technologies often collectively called ‘artificial intelligence’ (AI).  That is the last time I will use this particular phrase without irony in this series of articles – in my view, it is too vague a term, and tends to create an impression that computers are somehow approaching the capacity to operate on-par with human intelligence, which is simply not true.  Nonetheless, such luminaries as Stephen Hawking and Elon Musk have piped up over the past year or so with their concerns that our machines may soon rise up and render us obsolete or, worse still, destroy us!

In a similar vein, there are some people in the field of intellectual property who are starting to ask questions about whether computers can be ‘creative’ or ‘inventive’ and, if so, whether it should be possible for a computer to be named as an inventor on a patent application – or, conversely, whether some humans should be disentitled from inventorship on the basis that their computers, rather than themselves, were the true inventors.  One academic who has been making a name for himself in this emerging field of study is Professor of Law and Health Sciences at the University of Surrey, Ryan Abbott.  Professor Abbott is the author of, among other works on the topic, ‘I Think, Therefore I Invent: Creative Computers and the Future of Patent Law’, Boston College Law Review, Vol. 57, No. 4, 2016 (also available at SSRN), in which he argues that the law should embrace treating non-humans as inventors because this ‘would incentivize the creation of intellectual property by encouraging the development of creative computers.’

As I shall explain, however, I do not agree with Professor Abbott that computers can, or should, be regarded as inventors for the purpose of granting patents.  Furthermore, while Abbott accepts claims that patents have already been granted on what he calls ‘computational inventions’, I firmly believe that a computer is yet to ‘invent’ anything.  In my view, the researchers and technologists who claim otherwise have an interest in promoting a particular perspective, and in doing so they are subtly extending the definitions of ‘creation’ and ‘invention’ to encompass the contribution of their machines, to the detriment of the human operators who are responsible for providing the true creative input in the process.

I am further concerned that, should this view of ‘machine as (co)inventor’ prevail, it will in fact be to the detriment of the patent system.  I think it highly unlikely that lawmakers – whether they be legislators or common-law judges – will embrace the idea of granting patents on machine inventions.  On the contrary, it seems far more probable that if the notion takes hold that computers are actually doing the ‘inventing’ in many cases, it will simply become even more difficult for humans to secure patent protection for computer-implemented, or computer-assisted, inventions.

This is a complex topic that I intend to cover in a series of three articles.  In this first part, I will introduce the field of machine learning, give some examples, and then attempt to dispel some of the hype that has developed around this technology – including in Abbott’s work.  My aim here is primarily to refute the argument that existing machines are capable of engaging in ‘creative’ or ‘inventive’ activity.  In part 2, I will delve into the role of machine learning in assisting with the generation of new inventions.  Finally, I will look at how to go about identifying the (human) inventors in such cases.

 Read full article ► 

Tags: Artificial intelligence, DABUS, Innovation, Inventorship, Machine learning, Patent law

Identifying Patent-Eligible Software Claims... Using Software

As I have previously reported, the major and immediate impact of the US Supreme Court’s Alice decision in June 2014 was to reduce the rate of ‘business method’ patents issued by the US Patent and Trademarks Office (USPTO) by three-quarters, while having a negligible effect on ‘technical’ software-implemented inventions.  While the data in my earlier article ended in December 2015, I have now been able to update my results to the end of March 2017, as shown in the chart below.  There has been no change in the overall trend during the intervening 15 months, with ‘business method’ patent grants still running at around 50% of 2007 numbers, while technical software patents, and patents across all areas of technology, are issuing at rates nearly twice those of 2007.

In a recent post on the Bilski Blog, US patent agent Mark Nowotarski has made similar observations about the impact of Alice on ‘business method’ patent grants, going on to analyse the characteristics of those patent claims are are still being allowed in the USPTO’s business method Art Units.  He has noted that it is commonly by including ‘physical limitations’ (e.g. reciting hardware such as ‘mobile devices/sales kiosks’ or ‘physical sensors’) and/or ‘software limitations’ (e.g. reciting technical functionality such as ‘graphics/ image processing’ or ‘cryptography/security’) that applicants have been able to overcome Alice-based subject-matter rejections.

This got me thinking.  If there are common forms of technical language that arise in patent-eligible claims, then might it be possible to train a machine-learning system to predict whether a particular claim is, or is not, likely to be patentable?

It turns out that this does indeed appear to be possible.  I built a machine-learning model using data published by the USPTO, including the claims of 24,462 recently-abandoned 4967 recently-allowed applications, all examined within ‘software’ and ‘business methods’ Art Units.  In cross-validation tests (i.e. using a portion of the known data for training, and the remainder to test model performance) I was able to achieve around 75% prediction accuracy.  In trials of a hand-picked ‘random’ sample of more recent patents and published applications, not in the training/test set, the model correctly classified all actually allowed claims (of four examples) as patentable.  It also classified the claims of four abandoned applications as unpatentable, and two published-but-rejected-and/or-amended claims as unpatentable.  In only one case did the model classify a claim that had been rejected on subject-matter grounds as likely-patentable.

The model may thus be capable, with a probability of success of over 70%, of determining whether or not a proposed claim to a computer-implemented invention includes sufficient technical content to overcome a subject-matter-based rejection, at least under the Alice test as it is applied by the USPTO.

 Read full article ► 

Tags: Business processes, Computer programs, Machine learning, Patent analytics, Patentable subject matter, PEDS, US