![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
ABSTRACT
Institutional and academic debates have intensified regarding the recent efforts to claim inventorship of AI-related patent applications, as has notably been seen in the known cases of Thaler v Comptroller (‘DABUS’) that have been examined in various jurisdictions. The pertinent question that has emerged is whether artificial intelligence systems can independently produce patentable subject matter. What has to be looked at, first, is the preliminary question of what the claim of producing inventions ‘autonomously’ can possibly mean under a technological perspective – an essential stage in the debate that is usually bypassed in legal commentary. Once such a technological explanation has been provided, a legal question can reasonably arise as to whether an AI process, such as software, may make a contribution that rewards a patent. AI inventions are legally approached and analysed as processes and as to their relationship with their direct products. Thus, where a process (AI) ‘creates’ or ‘makes’ a product, the focus is reasonably put on if and to what extent disclosing the product can provide a contribution separate to that which has already been provided by the process that created it. It is stressed that the current push for AI-generated products bypasses this key question which is essential in assessing the invention.
Acknowledgements
The author thanks Dr Sadaf Shariat of the University of South Wales for discussions on the topic.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
Notes
1 See e.g. X Yang and others, ‘Concepts of Artificial Intelligence for Computer-Assisted Drug Discovery’ (2019) 119 Chemical Reviews 10520, 10521–23.
2 ibid.
3 UKIPO, ‘Artificial Intelligence – A Worldwide Overview of AI Patents and Patenting by the UK AI Sector’ (2019) <www.gov.uk/government/publications/artificial-intelligence-a-worldwide-overview-of-ai-patents>.
4 Thaler v Comptroller [2020] EWHC 2412 (Pat); Thaler v Comptroller [2021] EWCA Civ 1374; recently permission to appeal was granted by the Supreme Court <www.supremecourt.uk/news/permission-to-appeal-july-august-2022.html>.
5 Thaler v. Hirshfeld, 558 F. Supp. 3d 238 (E.D. Va. 2021); Thaler v. Vidal, No. 21–2347 (Fed. Cir. 2022) (USA); Thaler v Commissioner of Patents [2021] FCA 879; Commissioner of Patents v Thaler [2022] FCAFC 62 (Australia).
6 e.g. see Thaler v Comptroller [2021] EWCA Civ 1374, [1].
7 BL O/741/19 (2019 IPO Hearing Decision) at 29.
9 SJ Russell and P Norvig, Artificial Intelligence: A Modern Approach (4th edn Pearson, 2022) 19–20.
10 ibid 102–05.
11 ibid 133.
12 ibid 132.
13 ibid 19–20.
14 see <en.wikipedia.org/wiki/Artificial_intelligence>: ‘optical character recognition is frequently excluded from things considered to be AI having become a routine technology’.
17 Russell and Norvig (n 9) 42–44.
18 ibid 35–38.
19 ibid 801.
20 ibid 801.
21 ibid 44.
22 M Negnevitsky, Artificial Intelligence: A Guide to Intelligent Systems (Pearson, 2011) 13–14.
23 Russell and Norvig (n 9) 133.
24 Negnevitsky (n 22) 13–14.
25 Russell and Norvig (n 9) 160.
26 ibid 132.
27 ibid 130.
28 ibid 160.
29 ibid 160.
30 ibid 160.
31 ibid 160.
33 see also: <en.wikipedia.org/wiki/Ant_colony_optimization_algorithms>.
34 see also: <en.wikipedia.org/wiki/Artificial_bee_colony_algorithm>.
35 Russell and Norvig (n 9) 160; see also: <en.wikipedia.org/wiki/List_of_metaphor-based_metaheuristics>.
36 WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence (AI) – Second Session, Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence, prepared by the WIPO Secretariat (13 December 2019) – WIPO/IP/AI/2/GE/20/1.
37 ibid 3.
38 ibid 4.
39 ibid 4.
40 SIEMENS AG response: re Draft Issues Paper on Intellectual Property Policy and Artificial Intelligence (14 February 2020).
41 ibid 2.
42 ibid 2.
43 e.g. see Concise Oxford Dictionary (Clarendon Press, 1990); <en.wiktionary.org/wiki/autonomy>.
44 Concise Oxford Dictionary (n 43); <en.wiktionary.org/wiki/automatic>.
45 Thaler v Comptroller [2020] EWHC 2412 (Pat); Thaler v Comptroller [2021] EWCA Civ 1374.
46 US Patent No. 5,659,666.
47 R Abbott, ‘I Think, Therefore I Invent: Creative Computers and the Future of Patent Law’ (2016) 57 Boston College Law Review 1079, 1083.
48 R Abbott, ‘Everything Is Obvious’ (2019) 66 UCLA Law Review 2, 29.
49 Abbott, ‘I Think, Therefore I Invent’ (n 47) 1083, 1087.
50 ibid 1088.
51 e.g. see ‘succeeded in independently generating patentable results’ ibid 1086.
52 ibid 1084.
53 Abbott, ‘Everything Is Obvious’ (n 48) 2.
54 D Kim, ‘AI-Generated Inventions’: Time to Get the Record Straight?’ (2020) 69 GRUR International 443, 446.
55 ibid 448.
56 M McLaughlin, ‘Computer-Generated Inventions’ (2019) 101 Journal of the Patent and Trademark Office Society 224, 233.
57 ibid 243–47.
58 G Chimuka, ‘Impact of Artificial Intelligence on Patent Law. Towards a New Analytical Framework – the Multi-Level Model’ (2019) 59 World Patent Information 101926.
59 E Fraser, ‘Computers as Inventors-Legal and Policy Implications of Artificial Intelligence on Patent Law’ (2016) 13 SCRIPTed 305.
60 D Watson, ‘The Rhetoric and Reality of Anthropomorphism in Artificial Intelligence’ (2019) 29 Minds and Machines 417.
61 Russell and Norvig (n 9) 60: ‘to the extent that an agent relies on the prior knowledge of its designer rather than on its own percepts and learning processes, we say that the agent lacks autonomy. A rational agent should be autonomous – it should learn what it can to compensate for partial or incorrect prior knowledge. For example, a vacuum-cleaning agent that learns to predict where and when additional dirt will appear will do better than one that does not’.
62 ibid. ‘To the extent that an agent relies on the prior knowledge of its designer rather than on its own percepts and learning processes, we say that the agent lacks autonomy. A rational agent should be autonomous – it should learn what it can to compensate for partial or incorrect prior knowledge. For example, a vacuum-cleaning agent that learns to predict where and when additional dirt will appear will do better than one that does not’.
63 T Fong and others, ‘Autonomous Systems Taxonomy’ (2018) Autonomous Systems CLT Meeting. No. ARC-E-DAA-TN56290 <https://ntrs.nasa.gov/citations/20180003082>.
64 ibid.
65 E Ackerman, ‘Roomba Inventor Joe Jones on His New Weed-Killing Robot, and What’s so Hard About Consumer Robotics’ (2017) 6 IEEE Spectrum.
66 Society of Automotive Engineers – SAE International, J3016, ‘Surface Vehicle Recommended Practice: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles’ (2021).
67 ibid 34.
68 ibid.
69 ibid.
70 ibid.
71 ibid.
72 S Kugele and others, ‘Towards a Taxonomy of Autonomous Systems’ (15th European Conference on Software Architecture, ECSA 2021).
73 Thaler v Comptroller [2021] EWCA Civ 1374, [1]: ‘At first sight, and given the way this appeal is presented by both parties, the case appears to be about artificial intelligence and whether AI-based machines can make patentable inventions. In fact this case primarily relates to the correct way to process patent applications through the Patent Office’ (Birss LJ).
74 ibid [92]: ‘Foreseeable technology: One of the submissions made was that the technology in this case was unforeseeable at the time of the 1977 Act. I do not believe it matters but if it does then I was not satisfied that the parties had examined that question sufficiently to make such a statement with any confidence. More work would have been required. What has changed since the 1970s is terminology. Edward Feigenbaum’s pioneering work on thinking machines, then called “expert systems”, was published well before the 1977 Act’ (Birss LJ).
75 ibid [103]: ‘Whether or not thinking machines were capable of devising inventions in 1977, it is clear to me that Parliament did not have them in mind when enacting this scheme. If patents are to be granted in respect of inventions by machines, the 1977 Act will have to be amended’ (Laing LJ).
76 Russell and Norvig (n 9) 35.
77 Known as the DENTRAL program; Russell and Norvig (n 9) 40.
78 Edward FeigenBaum Interview, ‘Annals of the History of Computing’ (2012) 74–81.
79 Political scientist who did interdisciplinary work in the fields of computer science, economics, and cognitive psychology – see <en.wikipedia.org/wiki/Herbert_A._Simon>.
80 IBM 701 was IBM’s first series production mainframe and IBM’s first commercial scientific computer, e.g. see <en.wikipedia.org/wiki/IBM_70>.
81 The test is that if a machine can engage in a conversation with a person without being detected as a machine, it can be assumed that it demonstrates human intelligence.
82 Russell and Norvig (n 9) 1057.
83 ibid 1035.
84 ibid 1035–37.
85 e.g. see P Barry and P Crowley, Modern Embedded Computing Designing Connected, Pervasive, Media-Rich Systems (Elsevier, 2012) 437: ‘[in] a virtualised platform, whenever an operating system thinks it is interacting with or observing or modifying [etc]’.
86 D Patterson and J Hennessy, Computer Organisation and Design, the Hardware/Software Interface (5th edn Morgan Kaufmann, 2014), 446: ‘[w]hen the operating system decides to change from running process P1 to running process P2 []’.
87 ibid 437: ‘if the operating system chooses to replace the page, []’.
88 CIPA Journal, ‘Artificial Intelligence: CIPA's Response to the IPO Consultation’ (2021) January–February, 10.
89 ibid.
90 Russell and Norvig (n 9) 1057: ‘Consciousness remains a mystery’.
91 ibid 1037.
92 Merrell Dow v HH Norton & Co Ltd [1996] RPC 76, 86: ‘an invention is a piece of information’.
93 s.125 PA 1977.
94 e.g. see PA 1977, s48B(1)(c), s55(1)(a), s60(1).
95 Hospira UK Ltd v Genentech Inc [2014] EWHC 3857 (Pat), [127] (Birss J).
96 Halliburton Energy Services v Smith International (North Sea) Ltd [2006] RPC 2, 94 – intangibles are typically excluded from patentability but may be allowed if there is an appropriate technical effect.
97 Biogen Inc v Medeva plc [1997] RPC 1.
98 Generics (UK) Ltd v H Lundbeck A/S [2009] UKHL 12, [99].
99 Aerotel Ltd v Telco Holdings Ltd [2006] EWCA Civ 1371, [43]; also Halliburton Energy Services v Comptroller [2011] EWHC 2508, [7] and [50].
100 Gale's Application [1991] RPC 305, 328.
101 Merrill Lynch’s Application [1989] RPC 561, 569.
102 e.g. under PA 1977, s.1(2)(c).
103 Halliburton Energy Services v Comptroller [2011] EWHC 2508.
104 Thaler v Comptroller [2021] EWCA Civ 1374, [2]–[3].
105 ibid [6].
106 ibid [7].
107 BL O/741/19 (2019), para 30.
108 Sixteen month period specified by the Patents Rules, Rule 10(3).
109 ‘To be clear, the Office accepts that DABUS created the inventions set out in the two applications’ BL O/741/19, para 15.
110 Thaler v Comptroller [2020] EWHC 2412 (Pat), [5].
111 Thaler v Comptroller [2021] EWCA Civ 1374, [113] (Arnold LJ).
112 BL O/741/19 (2019), para 29: ‘As the applicant says, inventions created by AI machines are likely to become more prevalent in future and there is a legitimate question as to how or whether the patent system should handle such inventions. I have found that the present system does not cater for such inventions and it was never anticipated that it should [] it is right that this is debated more widely and that any changes to the law be considered in the context of such a debate, and not shoehorned arbitrarily into existing legislation’.
113 Thaler v Comptroller [2020] EWHC 2412 (Pat), [5].
114 Catnic v Hill & Smith [1982] RPC 183 (HL), see also Kirin-Amgen Inc v Hoechst Marion Roussel Ltd [2004] UKHL 46; [2005] RPC 9, 18–22 and 45–48; also L Bently and others, Intellectual Property Law (Oxford University Press, 2018) 656–57.
115 Bently and others (n 114) 656.
116 Aerotel Ltd v Telco Holdings Ltd [2006] EWCA Civ 1371, [43]; also Halliburton Energy Services v Comptroller [2011] EWHC 2508, [7] and [50].
117 Kirin-Amgen Inc v Hoechst Marion Roussel Ltd [2005] RPC 9, [69].
118 Russell and Norvig (n 9) 801: “[] the networks trained by deep learning methods are often called neural networks, even though the resemblance to real neural cells and structures is superficial.”
119 GB2308476 A (WO96/12221), page 2, lines 11–22.
120 e.g. see <en.wikipedia.org/wiki/Artificial_neural_network>.
121 GB2308476 A, page 5, line 20 – page 6, line 7 and figs 5, 11, 17.
122 GB2308476 A, e.g. see ‘coffee mug design’.
123 ibid page 11, lines 14–20.
124 ibid page 11, lines 29–32.
125 ibid page 12, lines 10–13.
126 ibid page 13, lines 16–18.
127 ibid page 14, lines 28–31.
128 ibid page 3, lines 6–9.
129 ibid page 3, lines 19–22.
130 GB2308476 A, page 8 lines 6–18; also page 13, lines 1–12: ‘neural networks embedded within the same computer code’.
131 ibid. see claims 34–35: ‘algorithmic computer code’-these claims were excised from the granted specification.
132 ibid page 11, lines 5–6: ‘information may be put on a spread sheet to be used for training the two separate neural networks’, also ‘spread sheet’ page 13, line 6.
133 e.g. see SK Goudos, Emerging Evolutionary Algorithms for Antennas and Wireless Communications (IET 2021).
135 US5719794, published 17 February 1998, inventors EE Altshuler and S Linden, Assignee: United States of America, as represented by the Secretary of the Air Force.
136 G Hornby and others, ‘Computer-Automated Evolution of an X-Band Antenna for NASA's Space Technology 5 Mission’ (2011) 19 IEEE Transactions on Evolutionary Computation 1, 20 and 31.
137 ibid 31.
138 JR Koza, ‘Human-Competitive Results Produced by Genetic Programming’ (2010) 11 Genetic Programming and Evolvable Machines 251.
139 e.g. see JR Koza and others, ‘Evolving Inventions’ (2003) 288(2) Scientific American 52.
140 US5719794, col. 2, lines 35–39.
141 ibid col. 26, lines 30–33.
142 ibid col. 26, lines 35–37.
143 The first objects in space to ‘be designed by a computer and not a human engineer’, see G Hornby and others, ‘Computer-Automated Evolution’ (n 136) 31.
144 ibid 1.
145 GB2308476 A, page 1, line 3.
146 Thaler v Comptroller [2020] EWHC 2412 (Pat), [5].
147 see above: ‘Use of the word “autonomy” in the technical literature’.
148 GB2308476 A, page 13, lines 25–26.
149 ibid page 3, line 19–21.
150 US5719794, col 25, lines 55–60.
151 ibid col 2, lines 56–60.
152 GB2308476 A, page 5, line 29–page 6, line 1.
153 ibid page 3, lines 21–23.
154 e.g. EP1461045A2, claim 17: ‘A product created by the process of claim 13’; US5948040A, claim 37: ‘A product created by the process as described in claim 35’.
155 Kirin-Amgen Inc v Hoechst Marion Roussel Ltd [2004] UKHL 46; [2005] RPC 9.
156 Manual of Patent Practice (MoPP) 14.120.1.
157 PA 1977 s.60 (Meaning of infringement).
158 e.g. see Generics (UK) Ltd v H Lundbeck A/S [2009] UKHL 12, [22]: ‘[] patent are likely to seek to obtain protection, not for a single compound, but for a class of compounds and sometimes an almost unimaginably large class []’ (Lord Walker).
159 Bently and others (n 114) 648.
160 Halliburton Energy Services Inc v Smith International (North Sea) Ltd [2006] RPC 2, [91].
161 Kirin-Amgen Inc v Hoechst Marion Roussel Ltd [2005] RPC 9, [100].
162 Magnesium Elektron Ltd v Molycorp Chemicals & Oxides (Europe) Ltd and another [2015] EWHC 3596 (Pat).
163 Thaler v Comptroller [2021] EWCA Civ 1374, [113].
164 Thaler v Comptroller [2020] EWHC 2412 (Pat), [5].
165 GB2308476, as discussed above.
166 Actavis Group PTC EHF v v ICOS Corp [2019] UKSC 15, [83]: ‘the notional skilled person’s task is to implement the ex hypothesi valid patent’.
167 ibid 53.
168 Actavis Group PTC EHF v v ICOS Corp [2019] UKSC 15, [57]; H Lundbeck A/S v Generics (UK) Ltd [2008] EWCA Civ 311, [35].
169 A granted patent is ‘ex hypothesi’ valid i.e. sufficiently enabled until proven otherwise.
170 See above, ‘What is “DABUS”?’.
171 E.g. expert opinions are given after the priority date but concern the state of affairs before the priority date, see Generics (UK) Ltd (t/a Mylan) v Yeda Research & Development Co Ltd [2013] EWCA Civ 925, [52]–[65].
172 ibid.
173 s.1(1)(a).
174 s.2(1).
175 .s.2(2).
176 .Also known as anticipation.
177 E.g. Merrell Dow Pharmaceuticals Inc v HN Norton & Co Ltd [1996] RPC 76, [89].
178 E.g. see Actavis Group PTC EHF v ICOS Corp [2019] UKSC 15, [53].
179 Under present UK law, an ‘individualised description’ may overcome anticipation of a prior, generalised disclosure. For example disclosure of ‘fixing means’ is not a disclosure of a nail. On the contrary, in pre-EPC UK law the general rule was that ‘disclosure of the class prima facie deprives its members of novelty’ and ‘prima facie a general disclosure of a class is a disclosure of all members of the class, however obscure and whatever the consequences’, see Dr Reddy's Laboratories v Eli Lilly [2009] EWCA 1362, [35].
180 Manual of Patent Practice (MoPP), 2.06.
181 Dr Reddy's Laboratories v Eli Lilly [2009] EWCA 1362, [32].
182 SmithKline Beecham Plc’s Paroxetine Methanesulfonate Patent (aka Synthon BV v Smithkline Beecham plc) [2005] UKHL 59; [2006] RPC 10.
183 ibid UKHL 59, [24].
184 ibid UKHL 59, [24].
185 Merrell Dow Pharmaceuticals Inc v HN Norton & Co Ltd [1996] RPC 76.
186 ibid 91.
187 ibid 80.
188 ibid 90.
189 ibid 90.
190 SmithKline Beecham Plc’s Paroxetine Methanesulfonate Patent [2006] RPC 10; Beecham’s patent GB2336364.
191 Synthon's application, WO9856787.
192 SmithKline Beecham Plc’s Paroxetine Methanesulfonate Patent [2006] RPC 10, 41.
193 ibid 34.
194 ibid 34.
195 ibid 34.
196 ibid 24 and 35.
197 ibid 21.
198 T 12/81 [1982].
199 Dr Reddy's Laboratories v Eli Lilly [2008] EWHC 2345, [81].
200 A type of stereoisomer not related as a mirror image.
201 T 12/81 [1982].
202 Thaler v Comptroller [2020] EWHC 2412 (Pat), [5]; Thaler v Comptroller [2021] EWCA Civ 1374, [6]–[9] and [113]–[114].
203 US5719794, col. 26, lines 30–33.
204 Hornby and others, ‘Computer-Automated Evolution’ (n 136) 1, 31.
205 E.g. see Abbott, ‘Everything Is Obvious’ (n 48) 2, 9: ‘proposes a novel framework for conceptualizing the transition from human to machine inventors’; and at 29: ‘reward early adopters of inventive machines which are able to outperform human inventors [] In 2006, for instance, NASA recruited an autonomously inventive machine to design an antenna that flew on NASA’s Space Technology 5 (ST5) mission’.
206 TW Dornis, ‘Artificial Intelligence and Innovation: The End of Patent Law As We Know It’ (2020) 23 Yale Journal of Law & Technology 97, 108.
207 Thaler v Comptroller [2021] EWCA Civ 1374, [114].
208 <https://web.archive.org/web/20220910064958/https://artificialinventor.com/dabus/>, see ‘other inventive machines’ and links therein.
209 According to the 1977 PA, an inventor is the actual deviser of the invention (s.7). Such a definition might not be very useful because the words devising and inventing are synonyms.
210 Thaler v Comptroller [2021] EWCA Civ 1374, [19], [50] and [147].
211 University of Southampton's Applications [2006] RPC 21.
212 ibid [24]–[25].
213 University of Southampton's Applications [2006] RPC 21, [42].
214 see Thaler v Comptroller [2020] EWHC 2412 (Pat), [5] and Thaler v Comptroller [2021] EWCA Civ 1374, [147]: the applicant ‘was unable to identify the inventor’.
215 University of Southampton's Applications [2006] RPC 21, [43].
216 ibid [39].
217 Pozzoli Spa v BDMO SA & Anor [2007] EWCA Civ 588.
218 Bently and others (n 114) 585.
219 Manual of Patent Practice, 3.26–3.28.
220 Schlumberger Holdings Ltd v Electromagnetic Geoservices AS [2010] RPC 33.
221 <www.exscientia.ai/dsp-1181>.
222 <www.bbc.co.uk/news/technology-51315462>: ‘Artificial intelligence-created medicine to be used on humans for first time’ accessed May 2022.
223 T Burki, ‘A New Paradigm for Drug Development’ (2020) 2(5) The Lancet Digital Health e226.
224 P Schneider and others, ‘Rethinking Drug Design in the Artificial Intelligence Era’ (2020) 19 (5) Nature Reviews Drug Discovery 353.
225 EP3608314.
226 Schneider and others (n 224) 353–64.
227 Abbott, ‘Everything Is Obvious’ (n 48) 2, 29.
228 Hornby and others, ‘Automated antenna design with evolutionary algorithms’ (2006) Space 7242.
229 ibid 5.
230 The ST5-33.142.7 antenna.
232 Abbott, ‘I Think, Therefore I Invent’ (n 47) 1079.
233 Dornis (n 206) 97.
234 ibid 108–09.