What happens to innovation if....
What happens to innovation if....
Generative AI Kills the Ability to Stop Others
How might new generative AI solutions such as GPT3 or ChatGPT impact the patent system and innovation more broadly?
According to Big Think, Generative AI is:
“a branch of artificial intelligence that enables computers quickly and convincingly to create original content ranging from images and artwork to poetry, music, text, video, dialog, and even computer code…These systems are master imitators of human creativity. They have been trained on millions upon millions of human artifacts such as documents, articles, drawings, paintings, movies, or whatever else can be stored in databases at scale. These systems have no conceptual understanding of the information they process — to a computer, it’s all just patterns of data — and yet, these Generative AI tools can create new pieces of content that are original and awe-inspiring.”
The release of ChatGPT a couple of weeks ago has catapulted Generative AI models into the mainstream consciousness. There have been dozens of mainstream media articles, blogs, LinkedIn and Twitter posts dedicated to the subject including samples of output.
This onslaught of media and general user attention, (it was reported that there were more than 1 million signups in less than a week),resulted in many attorneys (for the first time) taking an interest in the idea that technology might have the potential to at least partially replace them. This has been a long-time coming. Recently released Generative AI tools from OpenAI are garnering a new level of attention from the legal community.
I have worked in legal tech for the past 20 years, and my father was a pioneer in the industry, building tools to make the legal profession more efficient since 1993. Incremental advances in legal tech such as e-billing, e-discovery, and docket management have been embraced by many law firms over the years. However, software that disrupts the core value-adds that attorneys provide such as contract writing/simplification tools, patent claim drafting/examiner response tools, and litigation decision analysis tools have barely taken off.
After experimenting with ChatGPT and reading about the experiences many others are having with these tools, it is easy to see how generative AI tools apply to many legal tasks at the core of the work that attorneys perform for clients. I tested the tool and reviewed colleagues posting examples of patent application drafting, patent claim writing, agreement drafting, prior-art searches, and even proposed legal arguments.
I am not suggesting that these tools will replace attorneys entirely but I can certainly see how these tools will reduce the time attorneys spend on these time consuming tasks. I can also imagine how certain clients that are simply interested in being 80% (or even 70%) certain with the results of a legal task would deploy an AI solution rather than spending the additional money and time it takes to engage with an attorney.
But what if new generative AI models were used to disrupt the patent system in a different way? I had a thought the other day about the potential for Generative AI tools to disrupt the innovation system while simultaneously (potentially) blowing up the entire patent system.
This idea is as much Sci-fi as it is a possible reality. There are a number of technical challenges that would need to be overcome in order for this idea to become a reality. However, the thought pattern fascinated me, so I decided to write about it.
First a bit of background:
The pace of technology development has increased rapidly over the past couple decades. Whether looking at advanced battery technology, mobile communication devices, electric vehicle technology, robotic surgical devices, or really any other high-tech area, technology has infiltrated nearly everything we do. Software and semiconductors have become more and more prevalent in our society and I believe this is a trend that won’t change anytime soon.
In one specific example it is clear that for many years , the pace of innovation is occurring faster than government patent offices can review patents. As of 2022 there were more than 700,000 yet to be examined patent applications in the United States.
The rate of patent applications are increasing. According to a recent World Intellectual Property Office report, “Innovators around the world filed 3.4 million patent applications in 2021”. That is over 9000 patent applications filed every single day!
With approximately 8000 patent examiners in the US, and half a million or so new patents filed each year, the backlog is unlikely to be solved by humans alone.
In light of this rapid development of new ideas and inability of the patent system to keep pace, I question whether or not the patent system will survive. Their are a variety of reasons why I believe this to be true, but my main thesis is simply: due to the the speed at which new technologies are developed, the patent system will be unable to keep up with the pace of innovation.
It is possible that technology, specifically artificial intelligence models will be up for the challenge - to help people and government systems keep pace with the pace of innovation.
However it is also possible that these same artificial intelligence technologies become the the reason that the patent system is unable to keep up.
Incentivizing Innovation:
Some argue that the patent system “serves two primary functions: it provides an incentive for research and development and promotes the diffusion of ideas and information.”
To understand how that argument is justified, we must briefly review the patent application process.
An patent applicant files a patent in a specific country, let’s say the U.S. for exemplary purposes,
Next, an examination process occurs. Patent examiners that are employed by the U.S. government negotiate with the patent applicants to determine whether an invention is novel, non-obvious, and useful (among other things).
To determine whether a patent application is novel or non-obvious, the patent examiner refers to prior art, or literature that is published prior to the filing of the patent application.
In some cases, the patent examination process results in a granted patent which provides the owner of the patent with the right to stop others from making, using, or selling products that “infringe” the patented claims. In other cases the patent application is rejected.
The granting of a patent results in a limited (20 year) exclusive right to stop others from making, using, or selling the patented invention. This government granted right is provided in exchange for a public disclosure of the technology which must enable others that are skilled in the art to practice or duplicate the technology disclosed in the patent.
Once the 20 year term (from the date of filing) expires, the material in the patent can no longer be enforced against others and becomes part of the public domain (anyone can use the technology disclosed without having to pay the patent owner).
In a nutshell, the system is based on the idea that the issuance of the patent provides the applicant with a limited-term monopoly, in exchange, the applicant is required to teach others how the technology works by disseminating the ideas, which occurs when the patent application publishes.
The publication and teaching of the invention is theorized to incentivize other innovators from competing organizations to find new (novel and non-obvious) and ideally more efficient ways to solve whatever problem the technology solves. One way to outwit a competitor is to design around the competitor’s published patent.
Alternatively, the system inspires innovators to find an entirely new solutions.
In either case, the goal is to avoid infringing the patent claims of a competitor which can result in having to pay a competitor (a licensing fee, product sales royalty, or to defend a law suit) to make, use, or sell the patented technology.
Prior Art:
Prior art is at the core of the patent (and innovation) system. In order for something to be patentable, it must be determined to be novel and non-obvious in light of the prior art. Patent examiners do their best to identify all of the relevant prior art related to a patent filing. Due to the backlog of patent applications, a general lack of examiners, and an abundance of new patent application’s filed along with myriad other pressures that patent examiners face, examiners have less time to do an thorough search for prior art. Many patents are granted without an exhaustive prior art search.
Potential licensees or infringers will often perform a “scorched earth” search to identify any prior art that exists that wasn’t found by the examiner to invalidate a granted patent. Because there are millions of published documents available on the internet and in libraries across the world, this can serve as an effective strategy to defang a competitors patents. And, ultimately, a powerful tool to avoid paying a license fee or succumbing to a guilty charge of infringement.
Enter Megalo DB, “solving all of our innovation needs”:
Time for a bit of reality suspension.
Generative AI solutions are extremely powerful tools to generate multiple “versions” of text based on a prompt. A somewhat recent (2018) publication estimated that Google Scholar contains 389 million documents.
Now, imagine if an entity were to train a generative AI on the world’s published patents and scientific literature, this would include all non-fiction books, PhD dissertations, product manuals and other electronic versions of scientific documents. Then imagine if that entity asked the AI to iterate on each document thousands or millions of times. And then iterate on possible combinations of documents thousands or millions of times more.
The purpose would be to generate a Megalo database (Megalo DB) millions of times larger than the worlds current published literature database.
This hypothetical Megalo DB would contain trillions of documents discussing unique and combinatorial ways to solve problems far beyond what the human race could conceive, all based on ideas conceived by the human race.
To some degree this is AI and Machine Learning are already being applied to solve difficult problems. For example, in a recently published article on a National Renewable Energy Lab (NREL) project aimed at discovering plastic eating enzymes to recycle Polyethylene Teraphthalate (PET), developed by Japheth Gado and Erika Erickson, the author wrote:
“Advances in bioinformatics and machine learning had already made it possible to prospect vast databases of existing enzyme sequences for varieties active on crystalline PET.
…Gado built a statistical model to learn the biological rules of known plastic-deconstructing enzymes. The model assigned probabilities to the unique composition of enzymes studied to date. Gado also built a companion machine-learning model to predict the heat tolerance of enzymes, important for industrial applications.
Together, the two computational models let Gado and colleagues project into the unknown. In less than an hour, they screened over 250 million proteins to create a short list of promising candidates. Further testing confirmed that 36 were able to deconstruct PET, and 24 were previously undescribed in scientific literature.”
Megalo DB would take this type of experiment several steps forward and contain documents that describe *all* candidates, thus serving as prior art to future innovators querying the database.
Megalo DB would contain the worlds largest collection of innovative ideas. These ideas could focus on everything from small iterations of existing research fields (adding new processes, materials, chemicals, etc…) to considering combinations of research fields that have yet to be contemplated by humanity.
Ironically, most of the ideas generated by Megalo DB would likely not be possible to actually implement, but as long as it produced some useful ideas for solving some of the worlds’ most intractable problems, it wouldn’t matter.
To consider the implications that Megalo DB might have on the patent and larger innovation system, let’s consider a few potential business models Megalo DB could enable:
It could be used as an internal research tool. Query the tool for new ideas related to anything. For example, solving climate change. Focused on a specific technology such as Direct Air Capture (DAC), you could ask it to provide iterations on known technologies and the tool could hypothetically produce hundreds, thousands or even millions of research ideas and combinations never previously considered.
Megalo DB could be made available to others, for a fee. So for example, let’s say your startup wanted to develop a new material for producing a water resistant adhesive, the database could spin up thousands of ideas about what materials to test which to combine and so forth. Although many of the suggested solutions might not be feasible, the mere existence of these creative, non-intuitive, hypothetical solutions would be valuable.
And finally, what I believe to be the most disruptive model, imagine if an entity, under the auspicious of “serving the public good” decided to publish Megalo DB, making it freely available for any one to query. This could be a company like Alphabet, Apple, Amazon, or Microsoft, or a consortium that includes all four tech giants. Alternatively, it could be an open-source advocacy group such as the Electronic Frontier Foundation.
What am I actually proposing?
Simply stated, a massive AI generated database containing hundreds of billions or even trillions of scientific ideas. The database would not only have solutions to many (all?) of the worlds scientific problems, it would also serve (if published for all to access) as prior art to every human generated idea. Ultimately, this would make it extremely difficult (impossible?) for anyone to get a patent granted after the database has been published.
What impact would this massive, publicly available database have on innovation? How might it impact the patent system?
I asked a few colleagues what they thought:
First, I went to one of my more optimistic colleagues, a patent attorney and pharmaceutical expert. I explained the idea and told him that I that the impact on pharma patents could be very powerful as small molecules patents are one of a kind. Typically, there is one specific molecule that treats a disease, therefore, with a mega published database that contained every small molecule conceivable, applied to every disease, any new small molecule would be impossible to patent. Without the ability to stop others, what motivation or incentive do pharma companies have to continue to spend money developing new drugs? What happens to small companies, without the ability to stop others, do big companies buy them or simply copy them once they develop a successful product? What is the role of the organization that publishes this massive prior-art database? Would it likely be a large for profit company or a smaller not for profit company focused on “leveling the playing field”.
My colleague, was skeptical that Megalo DB would have an actual impact. In fact, he suggested that this type of “exhaustive” prior art database would have a limited impact. He argued that because the scientific literature in Megalo DB was created by a machine, patent lawyers and judges would argue the the simple existence of the information does not deem an innovation unpatentable as it requires humans to select the technologies that could actually work. He later admitted that if the AI were able to take an additional step of scoring the documents, thereby identifying those with the most potential, it could be a problem for the overall patent system.
Another colleague who is an IP software developer, but admittedly not an AI expert, emphasized the importance of the prompt when querying such a database. He argued that the prompt, or how the question is asked, matters most to the result that the AI returns. I asked, what if the prompt were: “in light of the existing corpus of published scientific literature and patents (on which you’ve been trained), what iterations could be developed to improve the current state of __________________(fill in the blank) technology”. He agreed, this could be a problem but was unconvinced that it would actually be possible to create Megalo DB.
Another colleague who is more of an IP-centric futurist suggested that although Megalo DB would immediately deem new ideas unpatentable, the game would adapt to whoever reduced the idea(s) to practice. In other words, the “flash of genius” would not be the critical part of how we define innovation, as Megalo DB would have already solved for that. The “innovation” would be given to whomever was able to select the idea that would work and build it to show that it works. Of course, his next statement was, “that works fine until the 3D printers or robots themselves are building and implementing the technologies.”
Several others I spoke with understood the immediate implications to the patent system that I envisioned when I thought of the idea.
Namely, that the patent system as we know it would disappear as soon as all current granted patents expired and all pending patent applications that were (eventually granted) expired.
Ostensibly, this would take approximately 20 to 30 years from the publication of Megalo DB. In short, every patent filed after the publication of Megalo DB would unequivocally be deemed unpatentable as the ideas would be either 1. anticipated by prior art (all elements of the technology disclosed in a single document), or 2. held as obvious in light of the combination of two or more documents.
Thus, in this scenario the patent system dies a slow death over the next 20 years as the currently granted patents and the backlog of patent applications pending at the time of this database being published get examined and eventually reach their 20 year expiration.
Patent attorneys are out of work, patent litigators are out of work, patent examiners are out of work, nobody can use a patent to stop others, innovation goes under ground, Trade secrets which have no time limit flourish - except that most of them are published in this massive ai generated database and so can be copied by any organization at their whim.
I suppose if Megalo DB were to come to fruition it would impact all types of other people and their jobs as well including researchers, writers, teachers, and likely many others.
But what does this do to innovation more generally?
Some (especially pro-patent people) argue that Megalo DB kills the spirit of innovation because there is zero incentive to innovate, thus people just sit back and drink Slurpies in their autonomous Lazy-Boy chairs, (thank you Wall-E).
Others (anti-patent, open source folks) argue that in a Megalo DB world, innovation is enhanced, as the “instructions” now exist to solve most (if not all??) of the world’s problems. The question only becomes how quickly can we build (and fund) these innovations.
Obviously, we don’t know the answer. The purpose of this essay was to simply advance the idea in the hopes of creating a discussion. If nothing else, it seems clear as day to me, that the patent system as it currently works will be incapable of keeping pace which how quickly technology is evolving (especially AI tech). And, the innovation system will continue to flourish as humans are highly focused on answering all of their questions.
In reality, it is unlikely that Megalo DB appears anytime soon. Most likely a large organization would keep something like this for internal use - at least in the short to medium term. That said, even if the AI technology is advanced enough to iterate on millions of scientific journal articles, millions of times each, there are significant questions regarding the storage, hardware, accessibility and a host of other technical issues required to scale Megalo DB. I recognize that this essay is as much a Sci-Fi style thought experiment as it is likely to become a reality.
That said, I think it is an important contribution to the broader discussion of the impact of AI on business, technology and law.
I’d love to hear your ideas and opinions this topic. Feel free to send me an email directly or post your ideas in the comments section below.
In any case, don’t forget to ignore the confusion.