We Need an Open-Source Approach to Weed Out Bad Quality Patents

“The open-source approach is the only way to ensure the creation of world-class patent tools that are assistive to innovators, examiners and other patent professionals alike, leading to only truly promising patents being granted.”

artificial intelligenceMuch has been written about patent quality. But many authors approach this problem with a bias against the very idea of a patent system.

These critics would “solve” the patent quality problem by cutting down the total number of issued patents rather than focusing on problem patents. They suggest increasing examiners and examination time will weed out bad quality patents. And this might throw up additional roadblocks to inventors obtaining a patent by increasing the time and cost of securing an allowance. But this does not necessarily improve patent quality. Instead, it merely reduces the total number of patents issued.

Rather than “more examination,” solutions to the patent quality problem need to focus on “better examination.” In theory, “better examination” should stop invalid claims from ever getting issued while simultaneously streamlining allowance for valid claims.

One proposal for improving patent quality is to use artificial intelligence (AI) to find better and more relevant prior art references. Unfortunately, the proliferation of prior art sources has made high-quality patent searching more difficult. Moreover, many of the best sources of prior art, such as technical literature, are not stored and indexed in a single database the way patent documents are.

An AI search tool can search more broadly and provide more relevant prior art references to human examiners. This allows examiners to focus on understanding and applying the prior art to the claimed invention instead of spending all their time searching for prior art.

This necessarily raises the question: What is the best approach for leveraging the potential of AI to improve patent quality? There are several compelling reasons to believe that an open-source approach will produce the fastest, fairest, and most complete AI search engine.

The Problems of a Fragmented Approach

Everyone is affected by the patent quality problem. Those who disagree with the premise of the patent system must understand that refusing to participate in the solution will only allow the system to produce more low-quality patents. And those who believe in the patent system know that a distrusted system will call their patents into question, thereby increasing the costs of obtaining and defending their patents.

Diverse groups of people, including corporations in the business of patent tools, researchers and academicians at universities, tech enthusiasts, and patent offices across the globe, are exploring and developing patent search AI. While these groups of people might have different motivations, they share an interest in ensuring that only truly innovative technology gets patented.

But a piecemeal approach will have the following results.

Slower Development of a World-Class Patent Search AI

A cooperative development effort can direct greater resources and more diverse talents to develop a near-perfect patent search AI. A fragmented approach, by contrast, will result in smaller and less diverse teams that will take longer to solve the inevitable problems that will arise.

But the patent quality problem is not small. According to the U.S. Patent and Trademark Office’s (USPTO) patent quality metrics, roughly 8.2% of allowances were erroneously granted in 2021. Given that the USPTO grants over 350,000 patents per year, even a brief delay in developing a high-quality patent search AI could result in tens or even hundreds of thousands of invalid patents.

Possibility of Blind Spots Due to Biases in Training AI

After developing an AI, it must be trained. But biases in training will create blind spots in the AI’s capabilities.

These blind spots are not immediately apparent. Biases were found in AI products as diverse as facial recognition to loan processing software only after they produced flawed outcomes.

A fragmented approach could cement blind spots in an AI patent search engine’s learning that produces a hopelessly flawed product.

Proprietary Products that Give Unfair Advantages to Bigger Players

A high-quality patent search AI should be available to all users, whether they are patent offices, patent owners, or patent skeptics. Whether you use patent search engines to examine patents, obtain patents or invalidate them, you should have access to the highest-quality search results possible.

A fragmented approach will produce proprietary products that come with a cost to use. This approach could allow one group of developers to corner the market for patent searching. Worse yet, a developer might only use the product internally to give itself an advantage in obtaining its patents and invalidating its competitor’s patents.

The Game-Changing Open-Source Approach

Open-source development can solve these problems. Everyone has an interest in ensuring that only valid patents get issued. Patent owners cannot continue to spend millions of dollars defending their patents. And accused infringers cannot continue to face the decision to modify their products, spend millions in litigation or risk billions in patent damages.

Solving this problem in a fair, fast, and thorough way will require a cooperative effort. This immediately brings to mind an open-source approach. Open-source software (OSS) is a significant driver of freedom, trust and innovation in the digital age.

No one company will own the fruits of AI-powered patent searching. Instead, advanced patent tools will remain available to those needing them, including patent offices and solo inventors worldwide.

All developers will have an opportunity to contribute to a patent search AI and ensure it avoids biases, thereby producing a better product.

All businesses will have the opportunity to utilize a patent search AI internally to make better patenting decisions. They will also have the ability to commercialize patent search AI created through open source. Like Mozilla and Linux, an open-source patent search AI can create spin-off products that fill needs no one even knew existed.

Advanced Patent Tools: Open Source and Open Access

The open-source approach is the only way to ensure the creation of world-class patent tools that are assistive to innovators, examiners and other patent professionals alike, leading to only truly promising patents being granted.

Additionally, no one should have a monopoly on these advanced patent tools. Open access to these tools will level the playing field for innovators especially. Innovators, no matter how small, should be able to determine if they have invented something deserving of a patent. Currently, there is a huge gap where examiners have access to advanced patent tools; some under-resourced inventors cannot even afford a professional patent search.

Moreover, the USPTO is expanding initiatives for under-resourced inventors and first-time filers. Open access to advanced patent tools created through open source can significantly assist these initiatives. We hope that open access will drive more diversity and inclusion in the patent space.




Warning & Disclaimer: The pages, articles and comments on IPWatchdog.com do not constitute legal advice, nor do they create any attorney-client relationship. The articles published express the personal opinion and views of the author as of the time of publication and should not be attributed to the author’s employer, clients or the sponsors of IPWatchdog.com. Read more.

Join the Discussion

6 comments so far. Add my comment.

  • [Avatar for Anon]
    November 16, 2022 07:39 am

    To the point as presented at https://ipwatchdog.com/2022/11/09/moving-toward-design-patent-bar-progress-ip-community/id=152828/#comment-2854008 :

    I contrast your statement here of,
    Rewarding the applicant for strengthening their own applications by commissioning a clear and thorough patentability search before examination provides evidence establishing Average Skill in the Art significantly reducing the probability that invalidating evidence will be found after the patent issues.

    Input quality at the front gate is precisely what applicants who want a high quality patent need.

    I would also add that your own post here explicating the necessary skill of a searcher (which IS a separate skill than that of inventors) speaks against putting that onus on the applicant.

    Further, as the link indicates, YOU already believe that the US had the most stringent ‘front gate.”

    Lastly, I would remind you of the repeated posts ‘at that other blog’ (even as it has been awhile since PatentlyO covered this) in that US examiners LARGELY ignore art provided by applicants. Certainly, this choice of examiners is not universal – I have seen many a time in which non-diligent examiners (incorrectly) parrot provided art, although this tends to come from examiners who simply are not as ‘up’ to the subject technical merits.

  • [Avatar for PA Crier]
    PA Crier
    November 15, 2022 11:55 am

    Another commentator who has never handed the results of a “clear and thorough” patentability search or validity search to a patent attorney yet again claiming that AI produces “more relevant prior art” than humans. This is a fake made-up fact. Unless you are a skilled patent searcher you lack insight into that question.

    I can see that at some point in the future a human asking “computer provide me evidence related to this abstract or these claims” and getting it. I don’t see that happening anytime in the near future.

    Don’t confuse what patent examines do with private sector searchers. Planning a field of search and conducting the search is the only way presently to improve patent quality. I don’t know how the 8.2% erroneous allowances was derived but the number is likely closer to 80% which you allude to with the comment “tens or even hundreds of thousands of invalid patents.”

    The key to patent quality is what do applicants want. Large corporations do not obtain patents to assert therefore low quality patents as produced by the PTO are fine for them. A much much smaller segment of applicants require a high quality patent. Rewarding the applicant for strengthening their own applications by commissioning a clear and thorough patentability search before examination provides evidence establishing Average Skill in the Art significantly reducing the probability that invalidating evidence will be found after the patent issues.

    Input quality at the front gate is precisely what applicants who want a high quality patent need. The fact that the US has an undeclared two-tiered patent system is real.

    On the other hand I applaud the work PQIA is doing. I have used it myself. However, AI is the fourth tool in the professional searchers tool-box and still a long ways from replacing the first three.

    It does not help the patent quality debate to assert that AI is a near term solution.

  • [Avatar for Model 101]
    Model 101
    November 10, 2022 09:46 am

    Pro Say


  • [Avatar for Anon]
    November 10, 2022 07:48 am

    Ms. Burke,

    Thank you for your comment, and I could not agree more that a proper emphasis for “Quality Patents” must be aimed at the examination function and explicitly, improving (rather than moving backwards) the very things that you identify.

    Our patent system should NOT aim for a “Sport of Kings” model, either with charging more fees or for putting onus on the ‘input quality’ at the front gate, as both of these will tend to move inputs to those types of innovators that may well ALREADY be established in the market, and away from disruptive innovators that most often challenge those who are already established.

    For those that have studied and understand innovation, and who recognize that the established players WILL play ‘the patent game’ only so far as to move that game to non-innovation factors, keeping the conduits OPEN to the less-monied, less-sophisticated, but more likely to be disruptive smaller players is how the patent system should be built to accommodate.

    This does mean that the Office examination system should be built to be ROBUST to handle a full spectrum of inputs (and much more than what well-monied applicants would — or could – drive those inputs to be).

    The “answer” to patent quality should NOT be ‘higher fees’ or more strenuous efforts on the part of applicants. We ALL should be aware of (and beware of) siren calls sounding in the Efficient Infringer mantra of, “0h N0es, Tr011s.”

  • [Avatar for Pro Say]
    Pro Say
    November 9, 2022 05:13 pm

    “We Need an Approach to Weed Out Bad Quality CAFC Judges”

    There. Fixed.

  • [Avatar for Julie Burke]
    Julie Burke
    November 9, 2022 12:35 pm

    Hello Nitesh-

    Thank you for your article. I agree that more examiners and/or more examination time will not lead to a better quality search and examination process that results in higher quality patents. While I appreciate your optimism, I doubt that simply giving examiners access to AI-driven search tools will fix the USPTO’s patent quality problem either.

    Why? The patent examiner’s performance and appraisal plan (PAP) is the vehicle that embodies the USPTO’s strategic goals to drive examiner behavior.

    Planning a field of search, conducting the search and even applying the references obtained from the search under 35 USC 102 and 103 are all categorized by the USPTO as BASIC activities in the quality element of the examiner’s PAP.

    As defined by Office of Personnel Management, the performance of basic tasks DO NOT REQUIRE a “sound understanding of those aspects of procedural and substantive law generally and of the statutory and case law applied to patents specifically, which are applicable to the patent examination process.”

    Instead, basic level tasks are assigned to GS-7 trainee-level employees and generally undertaken with a quick cursory review. An example of task appropriately labelled as basic would be checking that the brief description of the drawings matches the set of drawing submitted with the application.

    The PAP indicates that even seasoned GS-14 and GS-15 primary examiners are reviewed for their search and application of prior art under 35 USC 102 and 103 at the BASIC level. The analysis of the disclosure and claims for compliance with 35 USC 112 (enablement, written description and indefiniteness) is another series of legal-level tasks relegated by the USPTO to trainees to perform at the basic level.

    As long as the USPTO downgrades advanced and legal examination functions to basic trainee-level tasks in their PAP, the public, Congress and the IP community should expect to receive office actions, allowances and patents of BASIC level quality.

    To quote my SPE at an art unit meeting circa 1998: your job here isn’t to make Cadillacs. Your job is to make Yugos.

    With that sort of managerial mindset, the USPTO has since gone on to repeatedly downgrade the quality element of the examiner’s performance and appraisal plan, driving the United States towards a Yugo-level patent system.

    For more on this topic: https://ipwatchdog.com/2021/07/01/usptos-roadmap-improved-patent-quality-lead-lake-wobegon/id=135152/

Add Comment

Your email address will not be published. Required fields are marked *