ARTICLE

By: Reinaldo Franqui Machin*

Introduction

The patent system has been instrumental to the United States’ technological development.1 A new trend in patent applications has surfaced in recent years: innovations embedded with artificial intelligence (AI).2 Fueled by the exponential growth of computer power and its practical uses, the number of patent applications filed for AI-powered inventions doubled, from 30,000 in 2002 to 60,000 in 2018.3 This trend may be just beginning, as computers become more powerful, compact, and connect every device in our lives.4

A predictable patent system is essential to support this AI growth, yet current patent laws and court decisions are creating an unworkable AI patent system that does not foster innovation.5 For example, courts have established that if the invention consists of mental steps, it would be patent ineligible.6 By design, many useful AIs mimic human thought processes to generate an output.7 This framework discourages many inventors from disclosing their algorithms because they risk not receiving a patent or having their patent invalidated. There are also unanswered questions fundamental for AI patent protection, such as what constitutes prior art and who is the person of ordinary skill in the art necessary to determine obviousness, enablement, and disclosure.8 Without a clear answer to these questions, inventors will choose other mechanisms to protect their AI inventions, such as trade secrets, which fetters innovation and increases technology costs.9

The current patent system cannot effectively address AI protection because AI is radically different from other innovations. Thus, attempting to pigeonhole AI inventions using laws that did not foresee this technology will frustrate innovation. This article argues that the current patent laws are not suitable to protect AI innovations. Instead, a new type of patent exclusive for inventions that incorporate AI should be created. Some characteristics of this new patent category include the liberalization of the Alice-Mayo framework, building a digital depository for unexplainable but patentable AI, raising the person of ordinary skill standard, and allowing creative AI to be an inventor in a patent. Although creating a new kind of patent would require a significant effort from the legislature, it will be necessary to support this technological revolution and avoid trailing behind offshore competitors with better AI patenting systems.

The rest of this article is organized as follows: Section I provides a brief introduction to AI; Section II explains fundamental requirements to secure a patent and how AI inventions are incongruent with these requirements; Section III outlines the characteristics of the proposed AI patent category as well as potential problems of this solution; Section IV provides concluding remarks.

I. Brief Introduction to Artificial Intelligence

Artificial Intelligence are algorithms and computer programs that can solve problems that traditionally required human-level expertise.10 AI can differ from traditional rules-based programs because it does not always need a programmer to code all the rules; it can figure out the rules to produce the desired result.11 This process outsources the heavy lifting to the algorithm, saving time but, more importantly, allowing these programs to solve complex problems that would be impractical or impossible to code using the traditional rule-based system.12

Developing a new use for many types of AI requires training, which demands enormous amounts of data.13 For example, training an AI to detect human faces in pictures requires the algorithm to scan hundreds and thousands of images with faces in different settings to perform the task effectively. Without sufficient data, the algorithm will have problems detecting natural variations in faces; it may not notice a face if it is wearing sunglasses or have difficulty seeing faces of different races or genders.

Although AI’s growth is most noticeable now, the first AI programs were invented in the 1950s.14 The current expansion of AI technologies is possible by three factors outlined succinctly in Brad Smith’s book, Tools and Weapons.15 First, an increase in computing power has allowed the creation of complex AI.16 Moore’s law has predicted that computing power doubles every eighteen months, but in the case of AI, it has been doubling every 3.4 months, making more sophisticated uses of AI possible.17 Second, cloud computing and storage improvements have made it possible for more organizations and even governments to use AI and large data sets without a high price tag.18 Lastly, the explosion of digital data and and the current trends connecting all of our devises to the internet are allowing AI to permeate every industry.19 The overwhelming majority of all the world’s digital data were created in the past couple of years.20 This massive growth provides the necessary data to train AI and allow novel applications to better our lives.

II. Patenting AI: What Is the Problem?

An invention must overcome several hurdles to secure a patent. The biggest hurdles to patentability are codified in 35 U.S.C. §§ 101-03 and § 112.21 However, the invention also must survive legal frameworks devised by the courts to endure another party’s validity challenge.22 If the patent application meets all these challenges, the inventor gets a patent and control over the invention for a maximum of twenty years.23 AI inventions, however, are defying most of these statutory requirements and judicial frameworks. This section outlines some cornerstones of patentability and how AI is conflicting with these frameworks. Lastly, this section explores the likely effects of the incongruencies between AI and patent laws.

A. Patenting AI: Requirements and Conflicts

i. AI Subject Matter Eligibility

An invention must be subject matter eligible to secure a patent. As described in 35 U.S.C. § 101, subject matter eligibility includes new and useful inventions such as a “process, machine, manufacture, or composition of matter, or any new and useful improvement thereof . . ..”24 This statutory bar is low. The Supreme Court has stated that anything under the sun made by man could be subject matter eligible.25 However, courts have stepped in and created judicial exceptions that cannot be patented, such as “laws of nature, natural phenomena, and abstract ideas”, like those that merely involve mathematical concepts.26 These exceptions are reasonable. If Albert Einstein could have patented his famous equation, E=mc2, it would have monopolized an essential scientific research tool, preempting basic research for twenty years. However, the mathematical concepts exception to patentable subject matter is very troublesome for AI inventions becauseAI are mathematical expressions and algorithms.27 Furthermore, many of these AI programs either mimic mental steps or automatize human mental steps.28 But mental steps are also considered an unpatentable abstract idea.29 As a result, many useful AI inventions and claims have already been stricken down by the courts, creating uncertainty for inventors interested in protecting their inventions through patents.30

In an attempt to discern between patentable and not patentable subject matter, the courts created the Alice-Mayo framework.31 This two-step framework first asks if the claims are directed to a patent-ineligible concept, such as an abstract idea (i.e., a machine-learning algorithm).32 Even if the AI invention is directed to that ineligible concept, step two asks if the claim’s elements, individually and as a combination, transform the claims nature, making it patent-eligible.33 Drawing a line between a purely abstract idea and one with an “inventive step has been difficult.34 To generate predictability, the courts devised narrow exceptions and wide bars to the patentability of AI, which has resulted in a patchy system that does not support big innovative leaps.35

An example of the patchy system is the importance the courts give to an AI invention with a tangible, even if narrow, improvement in computer speed, storage, or other measurable metrics.36 If the AI is described to provide such an improvement, courts will likely deem its subject matter eligible, effectively circumventing the Alice-Mayo second step’s pitfalls.37 On the other hand, novel, industry-creating, AI inventions could be suspect for the courts because, under Alice-Mayo, particular focus is given to the existence of alternative mechanisms available to achieve the same result.38 The rationale is that if there are no alternative means to achieve the same product, the invention preempts the field and is directed to an abstract idea.39 Thus, when considered together, the current patent system rewards AI that offers minor improvements to a computer’s technology while restricting protection for a wide-ranging AI technologies that can be used in various industries. This framework is very dangerous now that the field is nascent and booming.40 The most groundbreaking AI inventions have most likely not yet been conceived, so inventors facing uncertainty in the basic concept of subject matter eligibility will be uneasy about choosing the patent system to protect their inventions.

 ii. AI Novelty and Obviousness

An invention must be novel to receive a patent according to Title thirty five of the U.S. Code.41 This requirement is met so long as every essential element of the invention does not already exist in some form, such as a physical embodiment or described in written form.42 In fact, this requirement is easy to meet, and it is reflected in the relatively few instances the issue is raised in court.43 However, some scholars have argued that the AI revolution has not yet occurred, positing instead that the AI boom consists of new applications of old algorithms allowed by advancements in computer technology.44 It remains to be seen if these old algorithms will challenge the novelty requirement of these new AI patents.

Another requirement to secure a patent involves 35 U.S.C. § 103, obviousness.45 An invention cannot receive a patent if it is merely an obvious variation of something that already exists.46 This obviousness determination is analyzed through the eyes of the person of ordinary skill in the art, a legal fiction.47 This hypothetical person knows all the available information in the field and can examine the invention to determine if it is not obvious.48 For example, an invention comprising a type of internal combustion engine would be deemed obvious (or not obvious) by a person of ordinary skill in the field of mechanical engineering and, to some degree, other fields incidental to the invention.

Nevertheless, in the case of AI, it is unclear if the person of ordinary skill in the art would be a programmer of AI or a technician in the field of the invention.49 Various AI systems take large teams to implement, requiring separate groups to program the original code and train and perfect it before it is usable.50 These interdisciplinary teams are all essential to make the invention, yet it is creating more uncertainty as to the appropriate person of ordinary skill in the art. Another open question is if the person of ordinary skill in the art for some inventions should include such person using an AI tool—a heightened person of ordinary skill standard.51 However, the Federal Circuit clarified that the person of ordinary skill in the art is not an automaton.52  So, an AI of ordinary skill in the art will not be entertained by the courts or USPTO. But, as an AI’s capacity to store and sort information continues to improve exponentially, the person of ordinary skill in the art could very well become an AI itself.53

Answering who a person of ordinary skill is, is not a mere academic exercise; the answer determines the pool of information that can be used to satisfy or frustrate the requirement of obviousness to obtain a patent.54 This brings us to a scenario where an unsuspecting inventor could draft and secure a patent, assuming they are a particular person of ordinary skill in the art, but a challenger could suggest a different standard that fits the current ambiguities. This other suggested person of ordinary skill in the art would bring in new, prior art to be considered for an obviousness analysis. Ultimately, this causes the AI patent to be an easy target for invalidation, since AI can be implemented in all sorts of fields and these issues will continue to compound as more complex and interdisciplinary AI-based inventions emerge.55 The requirement of obviousness is already one of the most commonly litigated issues in patent law, and AI patents would make matters worse.56 If these questions remain unanswered, the litigation costs will steer inventors away from patenting AI, and increase technology costs for those that do.

iii. AI Enablement and Disclosure

Enablement and proper disclosure are thorny issues for patenting AI inventions.57  Codified in 35 U.S.C. § 11258, this provision requires two things: that the patent (1) discloses the invention with enough detail so that it (2) enables another to recreate the invention without undue experimentation.59 This is part of the tradeoff to secure a patent’s benefits; the inventor ensures a temporary monopoly in exchange for full disclosure of the invention so others can make and use the invention when the patent expires. Additionally, proper enablement and disclosure help determine the contours of the invention, which is valuable for infringement and ensures the inventor does not claim things he or she did not invent.60

A fundamental problem of this requirement is that some AI technology variants, like neural networks, cannot be fully disclosed because few people understand how they work.61 This scenario is described as the AI “black box conundrum,” and it requires careful examination.62 An inventor who discloses an AI in a patent application, with no understanding of how it uncovers patterns or conclusions, may not sufficiently disclose it for another to recreate the invention without undue experimentation. If a person of ordinary skill in the art would have to undergo undue experimentation, then the inventor is deemed not to have actually disclosed the invention, and the patent could be denied or invalidated.63 Also, granting a patent to such an invention creates a one-sided bargain: the inventor obtains a monopoly in exchange for an inadequate disclosure that does not enable another to make and use the invention. This problem is made more complicated if the datasets used to train the AI are not public. A potential solution for this problem would be to require disclosure of the datasets used to train the AI to secure the patent. This would present no problems for many AI inventions trained with publicly available datasets. Still, technology giants like Facebook, Microsoft, Google, and many others would most likely not be interested in disclosing their private datasets for a patent application.

Another problem with AI patent disclosure is that an AI may have facilitated an invention, but the patent applicant may choose not to disclose its use to the patent office. Take as an example that there is currently no mechanism for the USPTO to determine if a highly effective anti-cancer therapeutic drug was the result of a predictive algorithm coupled with human testing or just a human inventor using traditional methods of experimentation.64 This loophole may seem trivial, but as these AI tools improve and accelerate human innovation, their nondisclosure will disadvantage innovators with no access or no knowledge that the devices are being used. Complicating matters, the USPTO recently took an official stance against allowing AI to be named inventors in patent applications.65 This stance will further disincentivize disclosure of cutting-edge AI tools that blur the lines between AI that merely assists an inventor and AI that is actually inventing. In other words, disclosing an advanced AI could lead the patent office to conclude that the AI was the inventor. Since AI cannot be a named inventor, the human inventor will be incentivized not to disclose the AI tool to secure a patent.

B. The Consequences of Poor AI Patent Protection

A viable alternative to patents is trade secrets, and AI inventions are particularly susceptible for protection under these laws.66 Trade secrets, unlike patents, do not require disclosure.67 To obtain trade secret protection, the owner only has to employ reasonable measures to maintain the invention a secret and reap a competitive advantage by keeping it a secret.68 Inventions that are easily reversed-engineered, like a mechanical clock, are not well-suited for protection under trade secrets because an individual can quickly figure out how it works. Reverse-engineering the clock would dissolve any trade secret protection.69 AI, on the other hand, is arguably easier not to disclose than to disclose, as was expounded upon in Section III(A)(iii). As a result, many would-be patentees are incentivized to protect their AI inventions through trade secrets. Moreover, trade secrets can retain protection for an unlimited period of time, contrary to patents, which have a maximum of twenty years of protection.70

Patents and trade secrets should not be considered antagonists. As trade secrets are considered an increasingly recommended alternative for many AI innovators71, protecting AI innovations as trade secrets could lead societal detriment because it will increase technology costs.72 As a whole, this system will also consolidate the market because only larger corporations have the resources to engage in duplicative research to develop tools or to secure a wide variety of narrow patents with little value.73 Thus, independent inventors and startups will suffer, and these have been shown to undertake the most significant innovative leaps when compared to more established risk-averse corporations.74

III. A Proposed Solution: A New Type of Patent for AI Inventions

Because there are many fundamental incongruities between AI and patent laws, alternate approaches to protect these inventions have been suggested. Some recommend protecting AI through sui generis, a type of protection that is unique or outside standard legal protections.75 In the United States, sui generis protection is given to mask works, which are two to three-dimensional layouts of an integrated circuit.76 Other countries also provide sui generis protection for some types of inventions.77 Congress could enact sui generis legislation tailored to the complexities of AI. These laws could forbid unauthorized extraction and use of an invention to protect the labor and economic investments of its creators. At first glance, this problem would seem to circumvent many of the difficulties patenting or protecting AI—assuming the laws are drafted correctly. However, the research conducted to make this article did not reveal significant efforts for this approach in the United States or internationally. On the contrary, most perspectives uncovered appear to advise against protecting AI through sui generis.78 Moreover, countries worldwide are responding by restructuring their respective patent laws and patent examination guidelines to accommodate AI-powered inventions.79 Thus, protecting AI through sui generis in the United States would most likely be disadvantageous in our globalized society if every major country is strengthening its patent system instead.

Because patents already provide a well-known set of rights and frameworks, the solution proposed in this article focuses on developing a new type of patent. The AI patent category would cater to the particularities of AI-powered inventions. This article follows the recent writing by Yanisky-Ravid and Jin recommending the creation of a new patent category designed explicitly for AI technologies.80 The general characteristics of this new patent category and some potential problems of this proposed solution are outlined below.

A. Characteristics of the New AI Patent Category

i. Liberalize or Eliminate Alice-Mayo

The Alice-Mayo framework gives way to heavy criticism. AI is already an easy target for invalidation under this framework, and it is forcing the creation of more narrow and less valuable patents.81 To solve this, Alice-Mayo should be eliminated or at least liberalized in the proposed AI patent category. A stronger focus on the utility of the invention, disregarding the abstract idea limitations, would be ideal. The Intellectual Property Law Association of Chicago, the Intellectual Property Owners Association, and the American Intellectual Property Law Association have given recommendations to change the subject matter eligibility requirement.82 The recommendations reflect a rejection of Alice-Mayo, merely requiring that the invention not be in the public domain or performed exclusively in the human mind.83  This strategy will reduce the unpredictability of patent validity challenges, which will encourage patenting of AI.

ii. A Depository for Unexplainable AI

As discussed above, in Section I (A)(iii) of this article, some forms of AI cannot be appropriately disclosed yet.84 But this is not the first time the patent office has encountered this problem. Patenting microorganisms is also not fully disclosable, as many tests for different features are continuously developed.85 The Budapest Treaty was enacted to solve this disclosure issue, allowing inventors to supply a sample of their invented microorganism that can be later tested for additional features if needed.86 A proper solution for the AI “black box problem” should be to require patentees to upload their code to a digital depository. As new tests reveal how these AI tools work, interested inventors could download a patented algorithm and test it, allowing for more disclosure as it becomes possible.

iii. Raise the Person of Ordinary Skill in the Art Standard to Include Ordinary AI Tools

Currently, the patent office only considers the person of ordinary skill in the art as a knowledgeable yet uninventive fictitious person.87 However, many fields readily use sophisticated AI tools, and the patent office should raise the person of ordinary skill in the art standard to include AI tools in those fields.88 Thus, the invention of a computer scientist could be rendered obvious if the person of ordinary skill in the art with ordinary AI tools in the field would have found it obvious. This heightened standard should not include the most novel tools available but rather those that are commonplace in the field. This would disallow patent protection to inventions that have no real inventive step since the ordinary AI tool rendered the invention in the first place.

iv. Grant Patent Protection to Creative AI, and its Inventions

The USPTO’s decision to disallow an AI to be a named inventor stems from the patent laws’ statutory language, which provides “[w]hoever invents…may obtain a patent” (emphasis added).89 Not allowing a creative AI to be a identified as an inventor will disincentivize innovators from creating valuable AIs capable of solving multidisciplinary problems independently. Additionally, it will foster nondisclosure of the use of cutting-edge AI tools in a patent application. Thus, the new patent category should allow an AI to be named “inventor”. Any rights of these patents would be bestowed on the patent owner, not the inventive machine. Promoting the patenting and full disclosure of these tools will catalyze innovation and allow anyone to access the most useful AIs.

v. Limit AI Trade Secret Protection

AI is too easily protectable through trade secrets, as previously stated in Section III.B. of this article. Even though variations of this technology are expected to permeate every industry in the coming years, AI protection will unfairly advantage established players with more resources to create powerful AI tools secretly.90 To prevent this unfair advantage, AI-powered tools should be given a more restricted trade secret protection. Instead of potentially infinite protection, only allowing five years of trade secret protection for AI innovations should balance the system. If misappropriation occurs after five years, no valid legal claim should materialize. Congress could make this change by amending the current federal trade secret laws. This will give a party sufficient opportunity and a strong incentive to share its technologies or seek patent protection. The ultimate goal of this approach is to deem AI innovations public goods, rather than private properties, thus allowing fast dissemination and access to useful AI inventions.

B. Potential Shortcomings of the AI Patent Category

Creating a new type of patent for AI inventions can cause various issues. One problem is the likely advantage to prominent players with more resources to develop and patent creative AI. This advantage could prevent the market entry of newcomers with fewer resources. For example, an independent inventor or a new startup would have difficulty competing with an established company with multiple creative AI tools in their patent portfolio creating new inventions. This would foster market consolidation—a problem also caused if trade secrets were to predominate AI invention protection. Creating different patent terms based on the inventor or patent owner’s status could reduce these adverse effects. An established prominent corporation patent owner would reap financial benefits from a patented AI system much faster than a startup or an independent inventor. Thus, granting a shorter patent term of ten years for these larger established corporations could help level the playing field. Smaller corporations or entrepreneurship projects would have a stronger incentive to innovate as they could secure the entire twenty years of protection, thus giving them sufficient time to reap financial benefits. However, a larger company would still be incentivized to develop these AI tools since these could ensure numerous AI patents and reap profits earlier, not requiring more extended patent protection.

The heightened person of ordinary skill in the art standard proposed here can also lead to some uncertainties. For starters, deciding if an AI tool is commonplace in a particular field could be difficult. Additionally, it will raise the standard of inventiveness and many innovators without these AI tools may be outdone by their AI-powered competitors. However, the increased disclosure and commercialization of those tools due to the proposed AI patent category would lead to faster access to these AI tools by all innovators, potentially at a lower cost.

However, raising the person of ordinary skill in the art standard to include AI tools could have a strange long-term effect. As more cutting-edge AI tools become commonplace, the person of ordinary skill in the art standard will keep increasing accordingly. If stretched to its logical conclusion, this could lead to a scenario in which everything would be obvious and unpatentable because powerful commonplace AI tools would foresee every new invention. At such juncture, patent protection principles would break down as AI would become more inventive than humans are. This could again lead to an increase in AI trade secrets, requiring even more restrictions on trade secrets than those proposed in Section III (A)(v). However, if done correctly, this may not be an undesirable result since other incentives—such as profitability—would continue to drive the development and dissemination of these tools. Additionally, the alternative would be to disregard the use of AI tools by the person of ordinary skill, which would result in patent protection for inventions that had no real inventive step. Thus, it is possible that patent singularity—the moment when patent protection is not possible due to advances in creative AI—will have a positive impact on society and help meet the goals intended for patents. Until then, raising the person of ordinary skill in the art standard would be the ideal and necessary solution to sustain the expansion of AI. 

Conclusion

AI is infiltrating virtually every industry, and this trend will continue as improvements in computer technology are expected to connect all of our surroundings.91 The patent system is an essential component of this expansion, but the courts are struggling to validate valuable AI patents as these do not fit current patent laws. Creating a new patent category specifically for AI-powered inventions will circumvent these problems and encourage disclosure of these new AI tools. This AI patent category could include eliminating Alice-Mayo, creating an AI depository, heightening the person of ordinary skill in the art standard, and allowing AI-made inventions to obtain patent protection. These features will mitigate current problems of subject matter eligibility, obviousness, disclosure, and enablement. Although it is likely that this solution will create new challenges, disregarding AI’s uniqueness in regards to intellectual property protection will do more harm than good. A revamping of patent laws is needed to cater to this increasingly important technology.


* J.D. Candidate at Howard University School of Law, Class of 2022; Patent Agent; Executive Publications Editor of Volume 65 for the Howard Law Journal; Molecular Biology Ph.D. from the University of Iowa; Microbiology B.S. from the University of Puerto Rico, Humacao. The author would like to acknowledge Professor Steven Jamar. Your guidance in the development of this article and your courses on intellectual property and artificial intelligence have been instrumental in my legal training. Thank you.

  1. See David Kline, Do Patents Really Promote Innovation?, THE MICHELSON INSTITUTE FOR INTELLECTUAL PROPERTY (April 24, 2017), https://michelsonip.com/patents-really-promote-innovation/ (summarizing the positive influence patents have had on innovation in the United States).
  2. See Rebecca Tapscott, USPTO Releases Benchmark Study on the Artificial Intelligence Patent Landscape, IPAWATCHDOG (October 28, 2020), https://www.ipwatchdog.com/2020/10/28/uspto-benchmark-study-artificial-intelligence-patent-landscape/id=126847/ (highlighting the critical findings of the report made by the USPTO about the current trends in AI patenting).
  3. Id.
  4. Id.
  5. See generally Ben Hattencach & Gavin Snyder, Rethinking the Mental Steps Doctrine and Other Barriers to Patentability of Artificial Intelligence, 19 COLUM. SCI. & TECH. L. REV.  313 (2018).
  6. Id. at 317.
  7. Neil Savage, How AI and neuroscience drive each other forwards, NATURE (July 24, 2019), https://www.nature.com/articles/d41586-019-02212-4.
  8. Uhthoff, Gomez Vega, & Uhthoff SC, Artificial Intelligence and its challenges for Intellectual Property, LEXOLOGY (September 30, 2020), https://www.lexology.com/library/detail.aspx?g=7af89c76-4645-411f-991a-0b19ade0dd19.
  9. Jessica M. Meyers, Artificial Intelligence and Trade Secrets, AMERICAN BAR ASSOCIATION, https://www.americanbar.org/groups/intellectual_property_law/publications/landslide/2018-19/january-february/artificial-intelligence-trade-secrets-webinar/ (last visited September 30, 2021).
  10. Darrell M. West, What is artificial intelligence? BROOKINGS (October 4, 2018), https://www.brookings.edu/research/what-is-artificial-intelligence/.
  11. Janelle Shane, You Look Like A Thing And I Love You20 (2019) (explaining the differences between rule-based and AI-based programing).
  12. Id. at 32 (explaining the impracticalities of using rule-based programing to teach a computer software what is a cat).
  13. Id. at 42-45.
  14. Rebecca Reynoso, A Complete History of Artificial Intelligence, LEARNING HUB (May 25, 2021), https://learn.g2.com/history-of-artificial-intelligence.
  15. Brad Smith & Carol Ann Browne, Tools and Weapons 195 (2019).
  16. Id.
  17. Cliff Saran, Stanford University finds that AI is outpacing Moore’s Law, COMPUTERWEEKLY.COM (December. 12, 2019), https://www.computerweekly.com/news/252475371/Stanford-University-finds-that-AI-is-outpacing-Moores-Law.
  18. Smith & Browne, supra note 15.
  19. Id.
  20. Bernard Marr, How Much Data Do We Create Every Day? The Mind-Blowing Stats Everyone Should Read, FORBES (May 21, 2018), https://www.forbes.com/sites/bernardmarr/2018/05/21/how-much-data-do-we-create-every-day-the-mind-blowing-stats-everyone-should-read/?sh=3607418760ba.
  21. 35 U.S.C. §§ 101-03, 112 (1952).
  22. 14 Famous Patent Infringement Cases that Changed US Patent Law, GREYB, https://www.greyb.com/famous-patent-infringement-cases/ (last visited September 30, 2021) (detailing the most prominent court cases that modified patent law in the United States).
  23. 35 U.S.C.A. §154.
  24. 35 U.S.C. § 101 (1952).
  25. See generally Diamond v. Chakrabarty, 447 U.S. 303, 309 (1980).
  26. Bilski v. Kappos, 561 U.S. 593, 601 (2010) (citing Diamond v. Diehr, 450 U.S. 175, 185 (1981)).
  27. Karen Hao, What is machine Learning?, Mit Technology RevieW (Nov. 17, 2018) https://www.technologyreview.com/2018/11/17/103781/what-is-machine-learning-we-drew-you-another-flowchart/.
  28. Id.
  29. Teaching artificial intelligence toadapt, SciENCE Daily (Dec. 16, 2020), https://www.sciencedaily.com/releases/2020/12/201216155201.htm; Google LLC v. Sonos, Inc., 498 F. Supp. 1138, 1142 (3d. Cir. 2020).
  30. See Purepredictive, Inc. v. H20.AI, Inc., No. 17-cv-03049-WHO, 2017 WL 3721480 (N.D. Cal. August 29, 2017).
  31. Alice Corp. Pty Ltd. v. CLS Bank Intern., 573 U.S. 208, 221-22 (2014).
  32. Id. at 221;  See also Digitech Image Technologies, LLC v. Elecs. for Imaging, Inc., 758 F.3d 1344 (Fed. Cir. 2014) (“Without additional limitations, a process that employs mathematical algorithms to manipulate existing information to generate additional information is not patent eligible.”).
  33. Id. at 225-26.
  34. See Susan Y. Tull & Paula E. Miller, Patenting Artificial Intelligence: Issues of Obviousness, Inventorship, and Patent Eligibility, 1 THE JOURNAL OF ROBOTICS, ARTIFICIAL INTELLIGENCE AND LAW 313, 315-17 (2018); Clark D. Asay, Artificial Stupidity, 61 WM. & MARY L. REV. 1187, 1203 (2020) (commenting that the Alice test can preclude most, if not all, existing software and AI patents); Ben Hattenbach & Gavin Snyder, Rethinking the Mental Steps Doctrine and Other Barriers to Patentability of Artificial Intelligence, 19 COLUM. SCI. & TECH. REV. 313, 318-27 (2018); Jordan Nimitz, Shattering the Looking Glass: How a Section 101 Revision Could Save FinTech From Alice, 30 FED. CIR. B. J. 55, 67-73 (2020-2021) (criticizing the application and effects of Alice in the digital technological revolution in the financial field).
  35. Hattenbach & Snyder, supra note 34, at 322-327; Asay, supra note 34, at 1205-06 (talking about how patent prosecutors are focusing too much on narrow computer improvements of the invention to avoid the Alice-Mayo minefield).
  36. Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1335-36 (Fed. Cir. 2016).
  37. Id.
  38. McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1315 (Fed. Cir. 2016).
  39. Id. at 1314-15.
  40. Nick Statt, The AI boom is happening all over the world, and it’s accelerating quickly, THE VERGE (Dec. 12, 2018), https://www.theverge.com/2018/12/12/18136929/artificial-intelligence-ai-index-report-2018-machine-learning-global-progress-research.
  41. 35 U.S.C. § 102 (1952).
  42. Gene Quinn, Patentability: The Novelty Requirements of 35 U.S.C. 102, IPWATCHDOG (June 10, 2017), https://www.ipwatchdog.com/2017/06/10/patentability-novelty-requirement-102/id=84321/.
  43. A search on LEXIS for the year 2020 showed that “novelty” and “patent” brought only 212 cases, while subject matter eligibility resulted in over 1,507 cases.
  44. Michael I. Jordan, Artificial Intelligence—The Revolution Hasn’t Happened Yet, HARV. DATA SCI. REV., Summer 2019, at 4-5 (arguing that the AI boom being observed in society are merely new applications of equations and algorithms developed during the 1950s).
  45. 35 U.S.C. § 103 (1952).
  46. Id.
  47. KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 417 (2007).
  48. Standard Oil Co. v. American Cyanamid Co., 774 F.2d 448, 454 (Fed. Cir. 1985).
  49. Tull & Miller, supra note 34, at 319-20.
  50. Anirudh V K, AI Job Roles: How to Become a Data Scientist, AI Developer, or Machine Learning Engineer, TOOLBOX (July 31, 2019), https://www.toolbox.com/tech/artificial-intelligence/tech-101/ai-job-roles-how-to-become-a-data-scientist-ai-developer-or-machine-learning-engineer/.
  51. Ryan Abbott, Everything is Obvious, 66 UCLA L. REV. 2, 37-–41 (2019) (arguing for raising the bar for obviousness by including machines as the person of ordinary skill).
  52. KSR Int’l Co., 550 U.S., at 420.
  53. Tull & Miller, supra note 34, at 320.
  54. Abbott, supra note 51, at 15-19.
  55. Tom Irving, Obviousness: Overcoming Obviousness Rejections by Attacking the Prima Facie Case, LEXIS NEXIS (June 2020), https://www.finnegan.com/images/content/2/5/v2/255327/PUBLISHED-Lexis-Practice-Advisor-Obviousness-Overcoming-Obvi.pdf.
  56. Isaac Kim & Richard Kurtz, The Role of Objective Indicia of Nonobviousness When Evaluating Whether a Patent Claim Is Obvious, ABA (Mar. 23, 2021), https://www.americanbar.org/groups/litigation/committees/intellectual-property/articles/2021/objective-indicia-nonobviousness-patent-claim/ (“For patent claims, the obviousness question under 35 U.S.C. § 103 has been the most common invalidity issue litigated in both district court and post-grant proceedings before the PTO.” (Citing Apple Inc. v. Samsung Elecs. Co., 839 F.3d 1034, 1074 (Fed. Cir. 2016) (Dyk, J., dissenting)).
  57. Will Knight, The dark secret at the heart of AI, Mit Tech Rev. (Apr. 11, 2017), https://www.technologyreview.com/2017/04/11/5113/the-dark-secret-at-the-heart-of-ai/; see also Tabrez Y. Ebrahim, Artificial Intelligence Inventions & Patent Disclosure, 125 Penn State L. Rev. 147 (2020); Shlomit Yanisky-Ravid & Regina Jin, Summoning a New Artificial Intelligence Patent Model: In the Age of the Pandemic 1, 27-29 (June 30, 2020) (unpublished) (on file with Social Science Research Network).
  58. 35 U.S.C. § 112.
  59. Gene Quinn, Patent Drafting: Understanding the Enablement Requirement, IPWATCHDOG (Oct. 28, 2017), https://www.ipwatchdog.com/2017/10/28/patentability-drafting-enablement-requirement/id=89721/.
  60. Quinn, supra note 59.
  61. Knight, supra note 57; Yanisky-Ravid & Jin, supra note 56, at 3-5.
  62. Yanisky-Ravid & Jin, supra note 56, at 27.
  63. Quinn, supra note 59.
  64. Ebrahim, supra note 56, at 190-91.
  65. FlashPoint IP, Ltd., 2020 Dec. Comm’r Pat. 1, 4-8.
  66. Meyers, supra note 9.
  67. Id.
  68. Id.
  69. Id.
  70. Meyers, supra note 9.
  71. Id.
  72. Holly M. Amjad, Patent vs. trade secret: Look at costs, industry, returns, Bus. J. (Feb. 3, 2002) https://www.bizjournals.com/kansascity/stories/2002/02/04/smallb3.html.
  73. Asay, supra note 34, at 1224-25.
  74. Maxwell Wessel, Why Big Companies Can’t Innovate, HARV. BUS. REV. (Sept. 27, 2012), https://hbr.org/2012/09/why-big-companies-cant-innovate.
  75. Sui Generis Right for Trained AI Models, Intell. Prop. Owners Ass’n 2 (Nov. 2, 2020), https://ipo.org/wp-content/uploads/2020/11/SG-model-rights-committee-paper-pub.pdf.
  76. Id. at 11-12.
  77. Id. at 12-13.
  78. Matt Hervey, EU report on AI-assisted creativity and invention, LEXOLOGY (Dec. 14, 2020), https://www.lexology.com/library/detail.aspx?g=6904f6ac-8973-4afd-8494-01e148111329; Christopher King, Inventorship, Patenting and AI: The Public Comments on Patenting Artificial Intelligence Inventions, JDSUPRA (Aug. 6, 2020), https://www.jdsupra.com/legalnews/inventorship-patenting-and-ai-the-15657/ (presenting the disagreements regarding of using sui generis protection for AI after the USPTO’s request for comments on patenting AI—most commentators agreed that patents are a good system for protecting AI even if the exact mechanism to do so was debated).
  79. Aaron Wininger, CNIPA Announces Amended Patent Guidelines for Patent Applications Covering AI and Blockchain, NAT’L L. REV. (Jan. 13, 2021), https://www.natlawreview.com/article/cnipa-announces-amended-patent-guidelines-patent-applications-covering-ai-and; Michael T. Renaud & Marguerite McConihe, European Patent Office Issues New Guidelines on Artificial Intelligence and Machine Learning, NAT’L L. REV. (Nov. 1, 2018), https://www.natlawreview.com/article/european-patent-office-issues-new-guidelines-artificial-intelligence-and-machine.
  80. Yanisky-Ravid & Jin, supra note 56, at 32-43.
  81. Ali Talip Pinarbasi, Effects of the Alice Preemption Test on Machine Learning Algorithms, IPWATCHDOG (June 19, 2020), https://www.ipwatchdog.com/2020/06/19/effects-of-the-alice-preemption-test-on-machine-learning-algorithms/id=122617/.
  82. IPLAC 35 U.S.C. § 101 Language Reform Resolution, Intell. Prop. L. Ass’n Chi., https://iplac.memberclicks.net/assets/docs/Briefs/Amicus_Briefs/IPLAC%20101%20Resolution%20Comparisons%20and%20Annotation%202018-04.pdf (last visited Apr. 3, 2021).
  83. Id.
  84. See Knight, supra note 56; Yanisky-Ravid & Jin, supra note 56, at 27-29.
  85. Introduction to the Budapest Treaty, World Intell. Prop. Org. 1, https://www.wipo.int/export/sites/www/treaties/en/registration/budapest/guide/pdf/introduction.pdf (last visited Apr. 3, 2021).
  86. Id. at 1-4.
  87. Standard Oil Co., 774 F.2d at 454-55.
  88. Roger Brown, Where is Artificial Intelligence Used Today?, BECOMING HUM. (Dec. 4, 2019), https://becominghuman.ai/where-is-artificial-intelligence-used-today-3fd076d15b68.
  89. 35 U.S.C. § 101.
  90. Joseph LaFave, Artificial Intelligence Is Expected To Permeate Every Industry, Setting These Companies Up To Win Big Contracts, SOFREP (Apr. 20, 2019), https://sofrep.com/news/artificial-intelligence-is-expected-to-permeate-every-industry-setting-these-companies-up-to-win-big-contracts/.
  91. Iman Ghosh, AIoT: When Artificial Intelligence Meets the Internet of Things, VISUAL CAPITALIST (Aug. 12, 2020), https://www.visualcapitalist.com/aiot-when-ai-meets-iot-technology/#:~:text=The%20IoT%20is%20empowered%20by,for%20real%20time%20data%20processing.
Share.