Enhancing the Value of Patents as Corporate Assets in South Africa: How can Artificial Intelligence (AI) Assist?

Although the accounting definition of assets contemplates intangible, abstract assets such as those embodied in intellectual property (IP), South African company law largely views IP as a legal and not a business asset. This paper tentatively suggests an approach that uses artificial intelligence (AI) to mitigate weaknesses in the South African patent law relating to the absence of patent searches and examinations. It is hoped that using AI will enable the filing of quality patents that satisfy the prescribed patentability criteria. High-quality patents will allow companies to accumulate patents as corporate assets. The approach is based on the algorithmic use of AI technologies such as machine learning, natural language processing, deep learning alongside the Internet of Things, and IP analytics to strengthen South Africa's IP system and create asset value for corporations. The paper recommends using the proposed AI technologies by companies and the Patents Office to enable the filing of high-quality patents, which will lead to the accumulation of corporate assets in the form of patents. The methodology is doctrinal, and the paper relies on recent literature on IP and AI, South African law, case law and examples drawn from studies conducted in other countries.


Introduction and background
The importance and impact of intellectual property, known to the subject cognoscenti as IP, cannot be gainsaid.IP permeates almost every aspect of human life.It creates and enables access to high and low culture 1 but at the same time is responsible for the high price of prescription drugs and the huge profits design companies like Apple make. 2 It determines whether farmers can plant the seeds they bought in more than one farming season 3 and whether inventions are the same or different from each other. 4In classical literature, IP is indeed an example of an oxymoron. 5e creation and protection of IP may be viewed as part of what is loosely referred to as the IP system, a public policy tool aiming to promote aspects of the economy, society, and culture by stimulating creativity and innovation in technology. 6Therefore, IP refers to rights given to persons (natural and juristic) over the creations of their minds.Such rights are referred to as IP rights (IPRs), which are timebound exclusive rights under national law. 7uring the subsistence of the IPR the creator can prevent the use of the product by others in specific ways without prior authorisation. 8The creator of an IPR can reap economic benefits from his or her creation directly or indirectly by authorising others.The permission can take the form of licences, which may be voluntary or compulsory. 9It is important to emphasise that despite international harmonisation efforts, 10 IPRs are A necessary consequence of a company's legal personality is that it can own property and assets which belong to it. 12The different manifestations of IP can be owned by corporate entities as corporate assets forming part of the asset register.Most forms of IP will be created through human endeavour, while advances in technology spurred by the fourth and other industrial revolutions imply that it may be possible to generate IP using machines.This possibility implicates machine learning and artificial intelligence (AI), which may help ensure that the IP generated is original and of high quality.For example, in 2020 the European Patent Office (EPO) declined to grant patent protection to an AI inventor, the "DABUS machine", which was described as "a kind of connectionist artificial intelligence". 13The matter was dealt with by the United Kingdom Intellectual Property Office, 14 the EPO, 15 and the United States Patent and Trademark Office (USPTO), 16 with all the different patent offices coming to broadly similar conclusionsthat the invention could not be registered as a patent because it was created by the AI machine. 17though IP corporate assets may take the form of trade marks, industrial designs, copyright and patents, patents are the most suited for protecting most innovations; hence they are discussed here alongside AI.
This paper discusses the relationship between patents and AI by looking at patentability requirements generally and in South Africa specifically.It aptly illustrates how applying strict patentability criteria may be aided by AI to Protection of Literary and Artistic Works, an international agreement governing copyright, which was first accepted in Berne, Switzerland in 1886.11   See generally Zhao 2017 QMJIP 137-155.12   Cassim "Legal Concept of a Company" 36-38.13   The issue attracted a lot of academic debate and for a conspectus of views on it, see the following authorities: Flint 2020 Bus L Rev 151-152; Deshpande and Kamath  2020 JIPLP 879-889; and Kidd 2020 Australasian Biotechnology 40-42.14   The UKIPO decision was issued on 4 December 2019.It is available at UKIPO 2019 https://www.ipo.gov.uk/p-challenge-decision-results/p-challenge-decision-resultsbl?BL_Number=O/741/19.The USPTO decision of 2020.See Tronson 2020 https://www.stlgip.com/wpcontent/uploads/2020/04/USPTO_AI_as_Inventor-r.pdf.17 For further legal arguments on AI related patentability, see Hattenbach and Snyder  2018 Colombia Science and Technology Law Review 313-318.L NDLOVU PER / PELJ 2021 (24)  4 develop inventions capable of beefing up IP corporate assets in the South African legal context.
The paper is divided into two thematic parts.The first part defines and conceptualises patentability criteria and exposes the real possibility of their interfacing with AI.The second part explores possible practical uses of AI approaches, such as machine learning and deep learning.The AI approaches are linked to the accumulation of corporate assets and the enhancement of patentability criteria.Finally, the paper optimistically concludes on a futurist note that AI will solve most problems associated with the granting of "wrong" patents before sketching a brief research agenda for the future.
2 Patents, corporate assets, and AI: Definitions, issues, and relationships

A short primer on patents
Patent law is domestic, and each country has its version of patent laws, which may not necessarily obtain in another country. 18Most patent laws, however, protect inventions that are new, inventive, and useful in life.
Patents for products and processes will be granted to inventors or the first persons to file for a patent for 20 years. 19Inventions must be new, 20 must involve an inventive step, 21 and must be capable of industrial application. 22dequate disclosure of the invention is mandatory so that the invention may be carried out by a person skilled in the art. 23The requirement that inventions must be new (novelty) is traceable to 16 th century England in Darcy v Allen. 24In Darcy v Allen, the granting of patent monopolies was 18 Hestermeyer Human Rights and the WTO 19. 19 The commencement of the 20-year period is known in patent law as the priority date.confined to products that were previously unknown in England.This aimed to curb the danger posed by patent monopolies, which could induce a patentee to demand unreasonably high product prices. 25The Statute of Monopolies, 26 which lasted for 200 years, is still viewed as the first important patent legislation that still informs the patentability criteria to date. 27cording to South Africa's current Patents Act of 1978, patents will be granted for 20 years from the date of the first application, and prescribed renewal fees have to be paid for the patent to subsist. 28South Africa's repealed Patents Act 29 of 1952 provided a term of 16 years from the date of lodgement of the complete specification at the Patents Office. 30On the ground of inadequate remuneration and/or war loss during the normal term, the 16 years could be extended upon application to the Commissioner of Patents. 31 A patent application consists of a specification which describes the invention and a section explaining the technology applicable to the invention circumscribing the protected rights claimedthe claims. 32The claims define the patentee's rights, and the specification instructs the public about the rest of the details.Generally, as national institutions, 37 patent offices examine compliance with patentability criteria in terms of each country's national laws. 38If the requirements are satisfied, patents are granted and published 39 for information and possible opposition.
Rights in product patents prevent third parties from making, using, offering for sale, selling, or importing the patented product without the patentee's consent. 40Similarly, with process patents, third parties with no prior authorisation from the patent holder are prevented from using, offering for sale, selling, or importing the product or products of the patented process. 41he enjoyment of patent rights is on a non-discriminatory basis, discounting where the product or process was invented, the technological field, and whether it was locally produced or imported. 42The unauthorised use of the invention and the product of the process may attract a claim for damages or an application for an interdict. 43 is worthwhile for the patentee to invest in product rather than process patents because market exclusivity for the products is guaranteed for 20 years; other inventors producing the same product through a different process may pose stiff competition for the owner of a process patent. 44 is easier to prove the infringement of product rather than process patents because infringing products are easily identifiable, while the same cannot be said of process patents, because the process of obtaining an identical product may be different from that of its patented counterpart. 45 some countries the grant of a patent may be contingent on working it locally first. 46Because of the local working requirement, the inventor will be compelled to manufacture the product or use the process locally. 47The local 37 working requirement originated from French law. 48The law may also disallow the exploitation of certain inventions on the grounds of protecting public order or morality, human, animal, or plant life, health, and the environment. 49

Corporate assets and IP
A company's assets may be in the form of corporeals (tangibles) 50 such as land and offices, and incorporeals (intangibles) 51 such as goodwill and IP.
The intangible assets of a company are usually referred to as its intellectual capital. 52Almost everything written about intellectual capital may be attributed to Stewart, who wrote a series of articles and published a book on the subject. 53How a company manages its intellectual capital depends on the existence and effectiveness of its business strategy. 54A company's actions and decisions to achieve set business goals considering business competition form the hallmark of its strategy. 55Steps required to reach business goals help to guide the process of decision-making. 56 effective business strategy ensures that corporate assets, including incorporeal assets such as IP, are managed for the company's benefit.IP is primarily regarded as one of the most critical assets for any business organisation. 57 is now regarded as both a business and a legal asset, and its value has steadily increased since the 1970s. 58 reduced to material form and commercialised in the context of each business in line with the business strategy.For example, suppose a pharmaceutical company invents a new drug to treat arthritis and registers a patent.In that case, the patent is a potential asset that is not truly useful to the company unless commercialised and aligned to its business strategy of developing and marketing blockbuster drugs that treat diseases common in elderly patients.
The management of IP as a business asset entails a multidisciplinary approach, including a business strategy, strategic litigation, intellectual asset management, risk management, the licensing of intellectual property, royalty collection and maximisation, franchising, and IP valuation. 60Most of the above-mentioned IP management concepts are self-explanatory, but one may note in addition that strategic litigation, risk management, licensing franchising, and royalty maximisation will, in all likelihood, increase the company's asset value and please shareholders.
To view a patent properly as a corporate asset, it is crucial to focus on what the patent does for the business.Patents will increase the value of a company's assets directly through sales of the patented products and indirectly if there are licensing agreements in place.The company can identify the value brought by each patent in its portfolio and attribute the value to a specific business unit in terms of the business strategy.Consequently, some departments or divisions in the company may become more IP intensive than others.In the long run, specific divisions in the company may become responsible for the bulk of the IP outputs.Whether or not a patent constitutes a corporate asset will depend on whether the invention has been commercialised or not.
To gain from IP generally as a corporate asset and specifically from patents, three considerations are essential.Firstly, the company should define its expected gains from IP management and secondly, it should determine how IP will support its business. 61Thirdly, the company should choose and follow a defined IP strategy to accomplish the two preceding goals. 62 A good example of such an IP strategy would be to incentivise staff working in units that produce more commercialised patents and to channel financial and technological resources to such units to boost the company's asset base through the instrumentality of IP accumulation.
L NDLOVU PER / PELJ 2021 ( 24) 9 What companies expect from their IP assets may be motivated by several factors.The motivation may be defensive, 63 cost-minimising, profitcentered, integrated, collaborative, impromptu, or at an ideal level, visionary. 64The defensive mode implies that a company's goal is to maximise its IP assets through accumulation, irrespective of incurring IP administrative costs.The goal is to amass IP for its own sake and to be counted among IP holders and owners.In the cost-minimising context, companies remain defensive but begin to view IP as costly, thus shifting their attention to limiting the costs associated with creating and maintaining their IP assets.The profit-centered approach to IP as a corporate asset kicks in once a company commercialises its IP and licenses it for profit through royalties to support the business.When IP assets are no longer confined to specific departments or units, the approach becomes integrated, and the company may transform into a hub of innovation and invention.At the visionary level, the highest IP management level, companies may use the IP to create more strategic value, considering IP's role in the economy. 65ming to patents, one may ask the rhetorical question: what business roles can patents play as corporate assets?When commercialised, patents can generate profits for companies through the direct sales of patented products and licensing agreements.In this manner companies will derive value from their assets associated with a registered patent.The company may also use patents to strategically position itself as an inventor and innovator, enhancing its image before its peers.To sum up, a company can extract value from its IP generally and from patents, particularly by selling the IP, licensing it to outsiders, using the IP in joint ventures, using it as a market access tool, using the IP to protect its products and services, and creating companies through which the IP is commercialised. 66 For companies to take full advantage of patents as corporate assets, they should have the infrastructure to spur innovation.Such infrastructure will ensure that only the best inventions get registered as patents.
As highlighted earlier in this paper, for inventions to be patentable, specific patentability criteria must be met.To enforce the patentability criteria, companies should take advantage of the 4 th Industrial Revolution and its attendant technological advances, such as AI.In most jurisdictions patentability criteria are enforced by human beings in patent searches and examinations, leading to some erroneous results.Some of these inaccurate results militate against a company's objective of maximising IP as a corporate asset, because patents may be opposed, revoked or litigated on infringement grounds.Therefore, AI may be used to obtain more accurate results on patentability criteria and enable companies to accumulate durable IP assets whose validity is unlikely to be opposed, revoked, or litigated.
To starkly illustrate how IP may effectively be leveraged as a corporate asset, one needs to look no further than the example drawn from the pharmaceutical industry by Drahos and Braithwaite, 67 reproduced in the appropriate context by Khotha and Stern. 68In the example, typically, when a pharmaceutical company develops a new drug it locates such development within an existing IP strategy that implicates different IP forms.
All aspects of the compound, including the dosage methods and manufacture, are protected by patent law, while some vital knowledge is not disclosed and remains protected under trade secrets law.Brand name identity is protected under trademark law, and copyright protects the new drug's mass documentation.This simplistic example shows that strategic planning on a company's part may be interwoven effectively into IP knowledge and an existing pre-planned IP strategy relevant to each company's context.
Before exploring the potential utility of AI in enhancing patents as corporate assets, it is appropriate to give an expository contextual account of the patentability criteria.

Patentability criteria
Requirements for patentability relate to whether a product or process 69 is patentable.Simply put, patentability criteria refer to specific attributes/requirements that must be satisfied before a patent can be granted. 70Patentability criteria relate to patentable subject matter. 71To establish if an endeavor is patentable, one needs to look no further than the TRIPS Agreement 72 and the South African Patents Act. 73 industrially applicable or useful. 74It is also possible for a defendant in a patent infringement suit to admit liability but counterclaim for the patent's revocation based on the failure to meet the patentability criteria.In such cases, the onus of proving invalidity rests on the defendant. 75 invention is new if it does not form part of the state of the art immediately before its priority date. 76The state of the art consists of all matter, whether a product, process, information about either, or anything else which has been made available to the public in South Africa by written or oral description, by use or in any other way. 77An invention will not be regarded as new if it is anticipated; 78 for example, an invention that has taken over all the essential integers of a previous invention is anticipated. 79The test to apply in establishing whether an invention has been anticipated is found in Gentiruco AG v Firestone SA (Pty) Ltd, 80 wherein Trollip JA made it clear that something that has been patented or described in any printed publication is anticipated. 81In one case, It was held that an invention whose claims related to purifying water using chlorine, a process well known and documented in scientific literature and other patent specifications, was anticipated and formed part of the state of the art. 82ether an invention involves an inventive step requires a three-legged enquiry. 83Firstly a question must be asked about what the state of the art was immediately before the priority date of the invention in question; secondly, whether the invention claimed is a step forward on the state of the art and finally, whether considering the state of the art, the step was inventive (not obvious). 84 how the step goes beyond the state of the art, and whether taking the step is not obvious to a person skilled in the art. 85practical example could be invention A, a bitter square pill, black and not aesthetically appealing, to cure mild headaches.Invention B, a claimed improvement on A, is now yellow in colour, round and sweet, but having the same chemical composition and curing the same disease with the same efficacy.Having regard to the state of the art before invention B, the invention is obvious to someone skilled in the art and does not involve an inventive step.The only additions were the shape, the colour, and sweetness; hence these cannot be regarded as significant components of inventiveness.
The manner it deems fit for its national context. 90The autonomy alluded to above may then be used to inform strict patentability criteria that may be embedded in AI prior-art-search algorithms, which are discussed in detail in paragraph three and four below.
Before inventors can experiment with ideas or processes to develop new products, they must search the available patent information database. 91The patent information assists inventors in conceiving, screening, and developing ideas into patentable final products. 92arching the thicket of patent information is not an easy task, and human endeavours alone will not guarantee the quality of such information.Other technology-based approaches, such as AI, will help make the search process yield more accurate results.
The main challenge relating to patentable subject matter in South Africa is that despite the prescriptive provision in section 25, patents are not examined in South Africa to ascertain whether the patentability criteria are met but to ascertain compliance with formalities. 93To achieve legal certainty, recourse must be had to case law.Therefore, an effective AI strategy entails reading the case law closely, alongside the science applicable to the field of technology.This calls for a multidisciplinary approach in creating the relevant and effective prior-art-search algorithm.AI is part of the 4 th Industrial Revolution; hence, brief remarks are necessary for full contextualisation.

Brief remarks on the 4 th Industrial Revolution and AI
The 4 th Industrial Revolution is also called Industry 4.0, 94 and many associate it with technological innovation.However, it also entails "an assemblage of novel technologies and forms of application, with discrete degrees of technical maturity and systemic effects". 95There have been 90 This has a textual basis in Art 1.1 of the TRIPS Agreement, which provides that "Members may, but shall not be obliged to, implement in their law more extensive protection than is required by this Agreement".three other industrial revolutions, each one having its distinct features and hallmarks. 96e Industrial Revolution in England between 1760 and 1840 is now largely referred to as the First Industrial Revolution. 97During the First Industrial Revolution machines began to replace human labour and automation became fashionable. 98The First Industrial Revolution is also known as "the age of manufacture". 99e Second Industrial Revolution was born of the first, and its main hallmark is electromagnetism, a science responsible for the generation of up to 90% of our electricity even today. 100 The Third Industrial Revolution happened in the USA in 1960.It introduced semiconductor technology, which is essential in computers and smartphones and is famed for automating communication using ones and zeros. 101It is responsible mainly for the automation of industrial production through electronics and information technology. 102e term Fourth Industrial Revolution (41R) was coined by the founder and executive chairman of the World Economic Forum, Klaus Schwab, in his renowned book. 103The 4IR 104 is described as a confluence of all the technologies around cyber, physical, and biological technologies, more specifically in that: The 4th Industrial Revolution is an era where people are using smart, connected and converged Cyber, Physical and Biological systems and smart business models to define and reshape the social, economic and political spheres.The 4IR is associated with the emergence of many specific technologies, most of which will lay the foundation or superstructure for further specialised applications. 106 the heart of the 4IR are biotechnology, AI, cloud computing, blockchain technology, the Internet of things, quantum computing, 3D printing, additive manufacturing, robotics, and the production of new materials. 107iotechnology is important in the IP context when one looks at, for example, pharmaceutical patents and the industrial manufacture and production of medicines.An important industrial application of biotechnology is in the manufacture of biological medicines and related products using genetically modified animals, plants, cells, fungi, and yeast. 108The internet of things is responsible for the invisible embeddedment of the information and communication systems around us, thus enabling device-to-device and human-to-device interactions. 109Blockchain technology, among other things, allows the safe and secure transfer of encrypted data and information between devices and generally ensures safe communication. 110uantum computing facilitates the creation of super-fast computers, while cloud computing allows the provision of computing power as a service. 111hese technologies are useful practically and will make manufacturing, the production of goods and the provision of services fast and efficient.
Companies employing such technologies will, of course, manage the IP spinoffs thereof.
To summarise the opportunities likely to be brought by the 4IR in the context of government and business, one can reiterate the apt observation by Xu, David, and Kim that the 4IR will bridge the gap between inventors and markets, facilitate an active role for AI, integrate different techniques and domains, and improve lives through techniques such as robotics and improve interconnectedness. 112cause this paper deals with using AI to enhance patents' value as IP assets, it is essential to define AI and its associated terminology.

AI and important associated terms
Although a precise definition of AI is difficult to pin down, AI generally refers to the technology that allows humans to build intelligent machines. 113Such machines will then exhibit aspects of human intelligence. 114AI has been defined as "the science of making machines do things that would require intelligence if done by men". 115Another author, John McCarthy, defined AI as "the science and engineering of making intelligent machines". 116I found the definition below, extracted from a 2016 Stanford University report on AI in 2030, to be very informative: Artificial intelligence is that activity devoted to making machines intelligent, and intelligence is that quality that enables an entity to function appropriately and with foresight in its environment. 117 becomes clear from the definitions that AI has been around for quite a while, albeit in different guises.For example, the scientific calculator (a machine) performs activities that the human brain can perform but performs mathematical functions faster than humans.Therefore, by making intelligent machines, tasks that traditionally have been performed by humans can now be performed by machines, which do not have human fallibilities.Good examples of human fallibilities are getting tired, making mistakes, getting emotional, being absent-minded, and paying less attention to detail.
The reliance on intelligent machines is likely to guarantee more production and efficiency.In the context of this paper the production of more IP products will boost companies' asset registers.However, one must not ignore the fact that intelligent machines do age through wear and tear, and are only as good as the information fed into them by human beings.
Machine learning, which has been described as a statistical approach to making machines intelligent, is an important offshoot of AI in the context of this paper.Natural language processing, deep learning, computational intelligence and cloud computing are other AI technologies relevant to patent searches and patentability criteria.Most AI uses computer programmes (algorithms) 118  natural language processing (forms of machine learning) are based on data and statistics. 119Such data can, in the context of this paper, be patentrelated data.With the advent of big data, which uses AI technologies to analyse vast amounts of data, the management of bulky and complicated patent information will become easier.This can be made possible by designing an algorithm that can analyse complex data, make predictions and read with more accuracy.
As shown in the analysis immediately below, AI technologies have a lot of potential to ensure that companies boost their corporate assets and appropriately apply patentability criteria.Using AI, machine learning, and deep learning, patent data can be analysed accurately and adequately to isolate the state of the art so that companies can extract business value from IP. Self-learning algorithms can be used as tools for effective IP management with positive results for companies.
The relationship between AI and law, on the one hand, and AI and IP on the other is now the subject of a substantial body of literature.The seminal texts include the 2018 Research Handbook on the Law of Artificial Intelligence, edited by Barfield and Pagallo. 120 The handbook organises law and AI themes into five parts dealing with law and AI generally; the regulation of AI; AI, fundamental rights and constitutional law; AI and IP, including AI and its interface with patents and copyright; and AI applications.A detailed analysis of the scope of AI under all the listed themes is beyond the scope of this paper, which focusses on the possible use of AI to aid patent searches and examinations in South Africa for companies to have good quality patents as corporate assets.

Discussion and analysis: The interface between patents as corporate assets, AI, and patentability criteria
The central argument this paper has advanced thus far is that AI can be used in patent searches and examinations with accurate results, leading to higher quality patents.Such high-quality patents will form part of a corporate entity's assets.Because such patents are unlikely to infringe on other existing patents, they are assets worth investing in and keeping for the long term.Viewed in this way, AI can assist corporates in accumulating patents as corporate assets.The discussion following immediately below elaborates on this initial premise.

IP and corporate assets
The point that IP is a corporate asset cannot be overemphasised.IP becomes a corporate asset of value when it becomes commercialised and can generate income for the company owning it.Commercialisation "constitutes immediate, measurable market acceptance for outputs" 121 such as patents.Many scholars view commercialisation as a process through which innovation is introduced into the market. 122Razak, Murray, and Roberts, cited by de Koker and Du Plessis,123 express the opinion that "commercialisation success is about converting ideas and new innovations into new marketable products". 124The components of a commercialisation life cycle, namely "idea generation, research, concept development and testing, analysis, product or service development, market testing, and marketing and commercialization", 125 apply mutatis mutandis when viewing IP as a corporate asset.Before IP or a patent can become a corporate asset, it starts as an idea which a company embraces, which leads to the company investing in related research, further development, testing analysis, patenting, marketing, and commercialisation (at which stage the IP becomes an exact IP asset with value).
The value of an asset depends on the business worthiness attributed to it. 126hrough exclusivity, IP allows companies to build their competitive and business value. 127To maximise the business value of IP, companies must embrace a well-established IP culture, which through an enabling IP policy incentivises IP generation, values IP, evaluates risk management strategies, licenses the IP, markets it, and takes IP dispute resolution seriously. 128Different forms of IP have different business value depending on the type of industry. 129For example, a technology company will value patents more than it does copyright, which may be of great importance to a publishing company.
For companies to invest profitably in IP, innovation, and related products, it should not be easy to patent anything by copying, and the state of the art should be easily accessible to all prospective investors and patent offices.For those companies that can afford it, an investment in machine learning technologies capable of detecting state of the art information accurately is worthwhile.For others, dependence on the state remains the only viable option through the Companies and Intellectual Property Commission.It is also possible for state departments to invest in AI technology to weed out undeserving patents, and this is encouraged.Patents that are not part of the state of the art are likely to complete the statutory 20-year term and weather the patent examination, opposition, and revocation storms.It is almost an axiom that where strong patents exist, companies that own them will have their IP assets guaranteed and boosted.
An AI prior art search would render a patent a better corporate asset than is currently the case in South Africa for reasons lying outside current South African law.Although there is no evidence of AI enhancing the value of patents in South Africa, such evidence does exist in studies carried out elsewhere. 130

AI and patentability criteria
Currently in South Africa one can do a physical and Boolean online prior art search.For a physical patent search an application is made to the Registrar of Patents, and upon payment of a prescribed fee requested documents may be furnished to the applicant.This type of search is laborious, cumbersome and not very accurate as some prior art may be missed in the thicket of patent documents in the Patents Office.The online Boolean search may be viewed as the better alternative, but it also has its significant shortcomings in the South African context.For example, most patents searched online have only summaries attached to them, 131 and it will not be easy for applicants to compare the prior art without access to the specification and the claims.
A typical application for a patent in South Africa commences with the applicant submitting the application forms for either a provisional or a complete patent to the patents office upon paying the prescribed fee. 132The Patents Office examines the application on formalities only and does not 130 See, for instance, an illustration by Genin and Zolkin 2021 World Patent Information 1-4 relating to the enhancement of patent assets quality using AI methods, artificial neural networks and machine learning in similarity searches in Russia.24) 20 look at the substance. 133Suppose the formal requirements, such as filling in the correct documents using the correct forms and paying the prescribed fee, are satisfied.In that case the application is accepted, published in the Patent Journal, and a patent granted to the applicant 134 for 20 years. 135lthough a prior search is advisable, it is not mandatory, and patent examination is not contemplated; hence many underserving patents are granted in South Africa. 136The South African online patent search facility provides for a simple and advanced search.A simple search can be done based on the application number, patent title, or the inventor's name. 137An advanced search expands the list of possible search items in addition to those applicable in a simple search. 138However and very peculiarly, in both types of search no provision is made for tying in keywords associated with the state of the art.
In contradiction, in the United States of America, when an applicant files an application with the USPTO, the office assigns the application an examiner who is specialised in the field of technology to which the application belongs. 139To determine whether the application is new, involves an inventive step, and is useful to a person in the relevant field, the examiner searches the database for previously issued patents, publications, patent applications, and other related inventions. 140Members of the public, researchers, and inventors can search the USPTO electronic database and access both the full text and the image databases. 141It is possible to conduct three types of online patent searches: a quick search, an advanced 133 Although s 34 of the Patents Act 57 of 1978 provides that the "registrar shall examine in the prescribed manner every application for a patent and every complete specification accompanying such application", in practice this does not happen.
134 This is the so-called depository system, which results in the granting of weak and undeserving patents.According to the publication by Fix the Patent Laws Patent Barriers to Medicine Access 11, patents granted under this system do not meet the country's patentability criteria.search, and a patent number search.The advanced search is the most comprehensive.Through it one can access complete information about the patent, including the title, abstract, cross-references to existing patents, claims, description, summary, descriptions of drawings and detailed descriptions. 142 The US system, which employs patent examinations and mandates patent searches, has an easily navigable online search system.However, the system is not foolproof and does require further improvement and refinement, considering the AI opportunities presented by the 4IR.Although South Africa recently recruited and trained patent examiners, 143 they are yet to start examining patents and, in the meantime, assist in patent searches. 144The weaknesses in the patent systems of the two jurisdictions compared above may be remedied using AI.
How then can AI aid patent searches and examinations?

AI and patent searches and examinations
Many patent lawyers spend a lot of time analysing patent claims and placing them before patent authorities. 145The better part of the claim analysis is spent on scrutinising novelty, inventiveness, and the possible utility of products or processes encapsulated in a patent. 146The law could benefit from AI due to the time and resources saved by the use of automated processes of establishing the applicability of the patentability criteria.
The possibilities of using AI in patent searches and examinations are vast, and tapping into this technology for the benefit of patentability and the accumulation of IP assets by corporates is not a remote possibility.The actual and potential use of AI may also be gleaned from other legal fields, such as health and human rights, 147 and closer to home, copyright. 148For example, using algorithms machines can compare prior art more accurately than humans can 149 and ensure that only deserving patents are filed and registered.
The Internet of things allows for valuable data to be retrieved and collected for use in predicting future scenarios, such as the likelihood of patent growth in specific fields of technology.Because everything will be connected through the Internet of things, 150 it will be easier to retrieve the most important information about patent specifications, especially the claims that define the essential integers of an invention.
Machine learning, natural language processing, and deep learning have led to the automation of tasks traditionally the preserve of humans. 151Some deep learning models use knowledge graphs that, in addition to technical features, also contain those technical features' essential relationships.Additionally, it has been reported that deep and machine learning may be used to establish novelty, the existence of an inventive step, and the utility of an invention. 152Therefore, machine and deep learning AI approaches may help patent searching, going beyond the traditional South African online Boolean search.
Machine learning algorithms 153 can detect patterns in data, learn these over time through repetition, automate complex tasks and make predictions. 154his may be adapted to patent searches with such retrieved patent data enabling the prospective inventor to sift through relevant and irrelevant data (search results  159 The pre-processing stage involves collecting data, extracting information, and preparing it to check its quality, correctness, and completeness. 160The information is analysed using various methods to aggregate, identify, and cluster it for meaning at the processing stage. 161he post-processing stage is also known as covered knowledge and leads to evaluating the results from the processing stage to support decisionmaking strategically.improve "keyword stemming, concept-semantic, and relational word searching capabilities". 165e South African patent system has been caustically criticised on several grounds, including that it is depository, encourages evergreening, and does not accommodate patent searches, examinations, and pre-grant opposition. 166Most if not all of the problems identified emanate from the fact that patentability criteria are not strictly enforced in South Africa for several reasons, including financial or administrative ones.
Embracing AI and its associated technologies will eliminate most of the problems identified here.It may be time for the Presidential Commission on the Fourth Industrial Revolution to come to the party and conceive a project that will help the CIPC integrate AI into its systems, especially patent searches and examinations. 167 While it may not be apparent to a skeptical reader how the arguments and descriptions above demonstrate how AI would make patents better assets, the following must be emphasised in closing this analysis section.
Although AI-generated prior art searches may yield more accurate results, 168 such searches are not expressly mentioned in the current South African Patents Act.While the statutory position remains unclear, the current legal regime, over and above the other weaknesses discussed above, may be characterised as woefully inadequate to deal with the recent growth in the use of AI systems.The prevailing legal position is that "only human and technology-assisted intellectual output is protected by IP law"; 169 hence AI-assisted patent searches will be permitted in South Africa.
Because AI-assisted prior art searches are more accurate than other searches, the resultant patents will not form part of the state of the art, and companies owning such patents will have valuable corporate assets, in addition to other assets they own.Therefore, AI will play a significant role in enhancing the value of patents as corporate assets.

Valedictory remarks and possible areas for further research
In this paper, I have demonstrated the relationships and linkages between IP, patents, corporate assets and AI to show that companies and patent offices can enhance patents' quality by resorting to AI.The use of AI, specifically data analytics generally and IP analytics, will ensure that accurate prior art search results containing few or no errors are obtained through the instrumentality of machines before further scrutiny and analysis by human subjects.The paper has reviewed the legal and other literature on AI and IP and contextualised it in the South African context.There has been an increased interest in the relationship between patents and AI, considering increasing instances of patent evergreening and the use of previously granted patents to block new patents and stifle innovation. 170AI, specifically the use of machine learning and data analytics, is presented in this paper as the panacea.The paper has also partially confirmed the potential utility of AI in enhancing patent quality in the South African context.If patent quality improves, companies will expand their IP assets and extract value from IP generally and patents in particular, especially as corporate assets.
Further research is required to identify and recommend specific AI technologies in other IP fields such as plant variety protection, trade secrets, copyright, trade marks, and industrial designs.AI remains unchartered territory for IP scholars in many parts of the world.For example, legal practice requires high cognitive abilities, 171 and AI algorithms cannot replicate humans' most advanced intellectual skills, such as analogical reasoning.With time and further research, the situation may change and will surely change.
As a futurist technology AI does have its blemishes.Some authors have described these blemishes as challenges. 172From the perspective of a developed and developing country, there is a reasonable apprehension of creating technology monopolies, which may exacerbate social inequalities.This may further manifest itself in unequal countries such as South Africa, where the divide between rich and poor is vast and keeps widening.
There is a real need to initiate a multidisciplinary study that will explore novel ways of using AI to improve and enhance the IP landscape in South Africa. 62

63Grimaldi,
Grecob and Cricelli 2021 Journal of Business Research 156-164.64 For a full discussion of these strategies and their implications, see Alexey, Criscuolo and Salter 2009 MIT Sloan Management Review 71-77.65 Sullivan and Harrison 2008 WIPO Magazine 2. 66 Sullivan and Harrison 2008 WIPO Magazine 2.

119Panch,
Szolovits and Atun 2018 Journal of Global Health 2. 120 Barfield and Pargallo Research Handbook on the Law of Artificial Intelligence.
iponline.cipc.co.za/iponlineTemp/ patents/65293_20210201185914NI.pdf.132 In South African law a specification can either be provisional or complete as contemplated by s 32 of the Patents Act 57 of 1978.L NDLOVU PER / PELJ 2021 (

170See
Hill 2014 Expert Opinion on Therapeutic Patents 839-843, tackling evergreening from the perspective of pharmaceutical patents.171 Surden 2014 Wash L Rev 87.172 See for example Xu, David and Kim 2018 International Journal of Financial Research 92 See ss 31 and 33 of the South African Patents Act 57 of 1978 for specifics about priority dates.
20 Novelty is sometimes synonymous with "non-obvious", as explained in Roman Roller CC v Speedmark Holdings (Pty) Ltd 1995 BP 199 (A) 212-221.Courts generally determine the issue of non-obviousness (see Gentiruco AG v Firestone 23 Per s 25 of the Patents Act 57 of 1978.24 Darcy v Allen 72 Eng Rep 830 (1603).Playing cards were the products at stake.
TRIPS Agreement is not directly prescriptive of how each World Trade Organization (WTO) Member ought to define what constitutes patentable subject matter and what may be excluded from patentability.However, there are prohibitions against discrimination based on fields of technology.Patents Act 39 of 1970 as amended.For the legal historical developments and a clear articulation of India's negotiation stance on the issue at the WTO, see Watal "Patents" 295-320.Joint Resolution No 118/2012, 546/2012 and 107/2012 (Ministry of Industry, Ministry of Health and National Institute for Industrial Property) published in the Official Bulletin on 8 May 2012.
86Although the TRIPS Agreement does not allow Members to discriminate against a field of technology, differentiation between fields of technology does occur in practice.Each field of technology is unique, and differentiation may be necessitated by practical considerations such as standards for examining patents.87Forexample, in pharmaceutical patents India does not recognise patents for naturally occurring substances, new forms of known substances in which enhanced therapeutic efficacy is not shown, new uses of known substances, mere admixtures (combinations), and treatment methods.88Meanwhile, in Argentina since 2012 pharmaceutical patents on any new form of known substances will not be allowed, regardless of increased efficacy.89Novelty,inventiveness, and industrial applicability are not defined in the TRIPS Agreement; hence each member state may define the terms in any 85 Ensign Bickford v AECI 1999 1 SA 70 SCA.See also the earlier related case of Roman Roller CC v Speedmark Holdings (Pty) Ltd 1996 1 SA 405 (A) 413.86 TRIPS Agreement Art 27.1.89 See Argentina's

135
Section 46 of the Patents Act 57 of 1978 provides that the duration of a patent shall be 20 years from the date of application, subject to the payment of renewal fees by the patentee.
). Machines can detect data patterns, which may help quantitative legal prediction, 155 the main component in patent searches and examinations.The science behind the analysis of large amounts of patent information, comparing it to existing data to expose relationships and patterns, constitutes patent analytics.156Patentdata are the largest repository of technical information, 157 and the latest technology and the state of the art are found in this repository.Patent analytics becomes very relevant and valuable to AI when used in conjunction with machine learning and deep learning 158 to analyse patent data.Intellectual property analytics in a business context has three stages: pre-processing, processing, and post-processing.
162Therefore, the application of AI methods, namely machine learning, deep learning, and artificial neural networks to intellectual property analytics will enable examiners and prospective inventors to access accurate information relating to the state of the art and thus to comply with the patentability criteria.South Africa is urged to consider taking advantage of this technology.On a related note, in as early as 2016 the editorial of the World Patent Information published an article in which the Australian Patents and Trade Mark Office, IP Australia, and the USPTO was said to have shown signs of embracing AI technologies to assist the advancement of IP, albeit in a slightly different manner in two examples.163Thefirstexamplewas the reference to IP Australia's launch of Alex, a virtual assistant who assists online visitors with trade mark queries.In March 2015, during a Patent Quality Summit, the USPTO announced that it would be launching a more advanced automatic search tool to replace the Patent Linguistic Utility Service, which uses an algorithm that identifies and analyses frequently used words to retrieve US patent data.164Thenew US tool is expected to 162 Aristodemou and Tieztze 2018 World Patent Information 42. 163 Editorial 2016 World Patent Information A1-A3.164 Editorial 2016 World Patent Information A1-A2.
Speculum Juris 70-100.See further DTI 2018 https://www.gov.za/sites/default/files/gcis_document/201808/ippolicy2018phasei.pdf14-27.167Inthe Report of the Presidential Commission on the 4th Industrial Revolution, the phrase "intellectual property" is mentioned nine times, and in those nine instances where IP is referred to, there is no express reference to leveraging AI technologies to aid the current flawed patent system (Department of Communications and Digital Technologies 2020 https://www.gov.za/sites/default/files/gcis_document/202010/43834gen591.pdf).