Congress Looking at Data Science for Ways to Improve Patent Operations

When Congress passed the sweeping Leahy-Smith America Invents Act (AIA) on September 16, 2011, legislators weren’t concerned about how data analytics might improve efficiencies at one of the Commerce Department’s most data-heavy institutions: the US Patent Office. Patent reformers at the time were instead focused on curtailing patent troll litigation and conforming aspects of US patent law to those of other countries. Consequently, the Patent Office’s trove of pre-classified, pre-labeled, and semi-structured patent application and invention data–information ripe for big data analytics–remained mostly untapped at the time.

Fast forward to 2018 and Congress has finally put patent data in its cross-hairs. Now, Congress wants to see how “advanced data science analytics” techniques, such as artificial intelligence, machine learning, and other methods, could be used to analyze patent data and make policy recommendations. If enacted, the “Building Innovation Growth through Data for Intellectual Property Act” or the “BIG Data for IP Act” of 2018 (S. 2601; sponsored by Sen. Coons and Sen. Hatch) would require an investigation into how data science could help the Patent Office understand its current capabilities and whether its information technology systems need modernizing.

Those objectives, however, may be too narrow.  Silicon Valley tech companies, legal tech entrepreneurs, and even students have already seized upon the opportunities big patent data and machine learning techniques present, and, as a result, have developed interesting and useful capabilities.

Take, for example, the group of Stanford University students who in late 2011 developed a machine learning technique to automatically classify US patent applications based on an application’s written invention description. The students, part of Stanford’s CS229 Machine Learning class, proposed their solution around the same time Senators Leahy, Smith, and the rest of Congress were debating the AIA in the fall of 2011.  More recently, AI technologies used by companies like Cloem, AllPriorArt, AllPriorClaims, RoboReview, Specif.io, and others have shown how patent data and AI can augment traditional patent practitioner’s roles in the legal services industry.

Some of these AI tools may one day reduce much of the work patent practitioners have traditionally performed and could lead to fewer Examiners at the Patent Office whose jobs are to review patent applications for patentability. Indeed, some have imagined a world in which advanced machine learning models conceive inventions and prepare and file a patent application to protect those ideas without further human input.  In the future, advanced machine learning models, trained on the “prior art” patent data, could routinely examine patent applications for patentability, thus eliminating the need for costly and time-consuming inter partes reviews (a trial-like proceeding that has created much uncertainty since enactment of the AIA).

So perhaps Congress’ BIG Data for IP Act should focus less on how advanced data analytics can be used to “improve consistency, detect common sources of error, and improve productivity,” as the bill is currently written, and focus more globally on how patent data, powering new AI models, will disrupt Patent Office operations, the very nature of innovation, and how patent applications are prepared, filed, and examined.

Patenting Artificial Intelligence Technology: 2018 Continues Upward Innovation Trend

If the number of patents issued in the first quarter of 2018 is any indication, artificial intelligence technology companies were busy a few years ago filing patents for machine learning inventions.

According to US Patent and Trademark Office records, the number of US “machine learning” patents issued to US applicants during the first quarter of 2018 rose 17% compared to the same time period in 2017. The number of US “machine learning” patents issued to any applicant (not just US applicants) rose nearly 19% during the same comparative time period. Mostly double-digit increases were also observed in the case of US origin and total US patents mentioning “neural network” or “artificial intelligence.” Topping the list of companies obtaining patents were IBM, Microsoft, Amazon, Google, and Intel.

The latest patent figures include any US issued patent in which “machine learning,” “artificial intelligence,” or “neural network” is mentioned in the patent’s invention description (to the extent those mentions were ancillary to the invention’s disclosed utility, the above figures are over-inclusive). Because patent applications may spend 1-3 years at the US Patent Office (or more, if claiming priority to earlier-filed patent applications), the Q1 2018 numbers are reflective of innovation activity possibly several years ago.

When It’s Your Data But Another’s Stack, Who Owns The Trained AI Model?

Cloud-based machine learning algorithms, made available as a service, have opened up the world of artificial intelligence to companies without the resources to organically develop their own AI models. Tech companies that provide these services promise to help companies extract insights from the company’s unique customer, employee, product, business process, and other data, and to use those insights to improve decisions, recommendations, and predictions without the company having an army of data scientists and full stack developers. Simply open an account, provide data to the service’s algorithms, train and test an algorithm, and then incorporate the final model into the company’s toolbox.

While it seems reasonable to assume a company owns a model it develops with its own data–even one based on an algorithm residing on another’s platform–the practice across the industry is not universal. Why this matters is simple: a company’s model (characterized in part by model parameters, network architecture, and architecture-specific hyperparameters associated with the model) may provide the company with an advantage over competitors. For instance, the company may have unique and proprietary data that its competitors do not have. If a company wants to extract the most value from its data, it should take steps to not only protect its valuable data, but also the models created based on that data.

How does a company know if it has not given away any rights to its own data uploaded to another’s cloud server, and that it owns the models it created based on its data? Conversely, how can a company confirm the cloud-based machine learning service has not reserved any rights to the model and data for its own use? The answer, of course, is likely embedded in multiple terms of service, privacy, and user license agreements that apply to the use of the service. If important provisions are missing, vague, or otherwise unfavorable, a company may want to look at alternative cloud-based platforms.

Consider the following example. Suppose a company wants to develop an AI model to improve an internal production process, one the company has enhanced over the years and that gives it a competitive advantage over others. Maybe its unique data set derives from a trade secret process or reflects expertise that its competitors could not easily replicate. With data in hand, the company enters into an agreement with a cloud-based machine learning service, uploads its data, and builds a unique model from the service’s many AI technologies, such as natural language processing (NLP), computer vision classifiers, and supervised learning tools. Once the best algorithms are selected, the data is used to train them and a model is created. The model can then be used in the company’s operations to improve efficiency and cut costs.

Now let us assume the cloud service provider’s terms of service (TOS) states something like the following hypothetical:

“This agreement does not impliedly or otherwise grant either party any rights in or to the other’s content, or in or to any of the other’s trade secret or rights under intellectual property laws. The parties acknowledge and agree that Company owns all of its existing and future intellectual property and other rights in and concerning its data, the applications or models Company creates using the services, and Company’s project information provided as part of using the service, and Service owns all of its existing and future intellectual property and other rights in and to the services and software downloaded by Company to access the services. Service will not access nor use Company’s data, except as necessary to provide the services to Company.”

These terms would appear to generally protect certain of the company’s rights and interest in its data and any models created using the company’s data, and further the terms indicate the machine learning service will not use the company’s data and the model trained using the data, except to provide the service. That last part–the exception–needs careful attention, because how a company defines the services it performs can be stated broadly.

Now consider the following additional hypothetical TOS:

“Company acknowledges that Service may access Company’s data submitted to the service for the purpose of developing and improving the service, and any other of Service’s current, future, similar, or related services, and Company agrees to grant Service, its licensees, affiliates, assigns, and agents an irrevocable, perpetual right and permission to use Company’s data, because without those rights and permission Service cannot provide or offer the services to Company.”

The company may not be comfortable agreeing to those terms, unless the terms are superseded with other, more favorable terms in another applicable agreement related to using the cloud-based service.

So while AI may be “the new electricity” powering large portions of the tech sector today, data is an important commodity all on its own, and so are the models behind an AI company’s products. So don’t forget to review the fine print before uploading company data to a cloud-based machine learning service.

Not Could, But Should Intelligent Machines Have Rights?

The question of whether intelligent machines of the future could own rights to their creations has been raised in the last few years, including last year at an international conference. Often, the question of rights for machines arises in the context of recognizing pictorial copyrights in images auto-generated by algorithms, but also in the context of invention (patent) rights. The progeny of these intellectual property (IP)-specific machine-rights questions suggests another, broader query: once artificial general intelligence (AGI) is achieved, should there be legal recognition of rights for intelligent machines?

In the context of IP, US laws recognize IP rights as arising from human creativity in the sciences and useful arts as a means to incentivize continuing research and creativity. Incentives flow in part from the economic nature of IP rights. Owning copyrights, for example, gives the owner the right to make and distribute copies of a creative work and to permit others to do so conditioned upon payment of a licensing fee, royalty, or some other consideration. Similarly, patent rights give a patent owner the right to exclude others from using one’s invention unless permission is granted, which again may be conditioned upon payment of a fee. In the US, employees who create copyrightable works or patentable invention are typically obligated to assign all their IP rights to their employer without further compensation beyond a regular salary. In such cases, an employer ultimately owns and benefits from its employee’s creations and discoveries.

But if we’re talking about intelligent machines, the debate about machine rights would have to take into consideration whether an intelligent machine could or would “care” about the economic benefits or have other reasons to innovate and create in the first instance. Aside from the prospect of ensuring an intelligent machine’s rights flow to its owner and are not lost, what purpose would be served by legally recognizing a future intelligent machine has rights at all? And if machine IP or other rights were recognized, what consideration would be appropriate to give future intelligent machines in a bargained-for exchange for their recognized rights, assuming the rights have value?

Some have suggested that the very notion of what it is to have “intelligence” will be protected by recognizing that intelligent machines are capable of owning rights. Put another way, if we do not recognize rights attached to the creative output of an intelligent machine, we undermine or diminish the very same rights associated with natural intelligence-generated creations. So the rights a photographer has in a photograph, or an inventor has to a new drug discovery, would be lessened if the same rights were not recognized in an intelligent machine’s similar endeavors. It seems doubtful, though, that an intelligent machine would care that it was given legally cognizable rights only because natural intelligent beings felt it important to do so to avoid diminishing the value they place on their own “intelligence.”

Despite recent advances in natural language processing (NLP), reinforcement learning (RL), and other artificial intelligence technologies, such as those in the areas of reading comprehension accuracy, computer vision object detection, and others, achieving AGI may be years if not decades away. Thus, machine rights is not a question that is especially pressing right now. On the other hand, one cannot predict when someone will announce a major breakthrough in AI and place this question front and center. So the debate about future intelligent machine rights is likely to continue. Hopefully, the debate will not distract from present-day efforts by many to move the needle toward legally recognizing important individual rights that protect people in the course of their daily interactions with existing and future AI technologies.

Evaluating and Valuing an AI Business: Don’t Forget the IP

After record-breaking funding and deals involving artificial intelligence startups in 2017, it may be tempting to invest in the next AI business or business idea without a close look beyond a company’s data, products, user-base, and talent. Indeed, big tech companies seem willing to acquire, and investors seem happy to invest in, AI startups even before the founders have built anything. Defensible business valuations, however, involve many more factors, all of which need careful consideration during early planning of a new AI business or investing in one. One factor that should never be overlooked is a company’s actual or potential intellectual property rights underpinning its products.

Last year, Andrew Ng (of Coursera and Stanford; formerly Baidu and Google Brain) spoke about a Data-Product-Users model for evaluating whether an AI business is “defensible.” In this model, data holds a prominent position because information extracted from data drives development of products, which involve algorithms and networks trained using the data. Products in turn attract users who engage with the products and generate even more data.

While an AI startup’s data, and its ability to accumulate data, will remain a key valuation factor for investors, excellent products and product ideas are crucial for long-term data generation and growth. Thus, for an AI business to be defensible in today’s hot AI market, its products, more than its data, need to be defensible. One way to accomplish that is through patents.

It can be a challenge, though, to obtain patents for certain AI technologies. That’s partly due to application stack developers and network architects relying on open source software and in-licensed third-party hardware tools with known utilities. Publicly-disclosing information about products too early, and publishing novel problem-solutions related to their development, including describing algorithms and networks and their performance and accuracy, also can hinder a company’s ability to protect product-specific IP rights around the world. US federal court decisions and US Patent and Trademark Office proceedings can also be obstacles to obtaining and defending software-related patents (as discussed here). Even so, seeking patents (as well as carefully conceived brands and associated trademarks for products) is one of the best options for demonstrating to potential investors that a company’s products or product ideas are defensible and can survive in a competitive market.

Patents of course are not just important for AI startups, but also for established tech companies that acquire startups. IBM, for example, reportedly obtained or acquired about 1,400 patents in artificial intelligence in 2017. Amazon, Cisco, Google, and Microsoft were also among the top companies receiving machine learning patents in 2017 (as discussed here).

Patents may never generate direct revenues for an AI business like a company’s products can (unless a company can find willing licensees for its patents). But protecting the IP aspects of a product’s core technology can pay dividends in other ways, and thus adds value. So when brainstorming ideas for your company’s next AI product or considering possible investment targets involving AI technologies, don’t forget to consider whether the idea or investment opportunity has any IP associated with the AI.

Patenting Artificial Intelligence: Innovation Spike Follows Broader Market Trend

If you received a US patent for a machine learning invention recently, count yourself among a record number of innovators named on artificial intelligence technology patents issued in 2017. There’s also good chance you worked for one of the top companies earning patents for machine learning, neural network, and other AI technologies, namely IBM, Amazon, Cisco, Google, and Microsoft, according to public patent records (available through mid-December). This year’s increase in the number of issued patents reflects similar record increases in the level of investment dollars flowing to AI start-ups and the number of AI tech sector M&A deals in 2017.

As the chart indicates, US patents directed to “machine learning” jumped over 20% in 2017 compared to 2016, and that follows an even larger estimated 38% annual increase between 2015 and 2016. Even discounting the patents that merely mention machine learning in passing, the numbers are still quite impressive, especially given the US Supreme Court’s 2014 Alice Corp. Pty Ltd. v. CLS Bank decision, which led to the invalidation of many software and business method patents and likely also put the brakes on software-related patent application filings (as explained here) beginning in 2014. So the recent jump in issued patents for “machine learning,” “artificial intelligence,” and “neural network” inventions suggests that specific applications of those technologies remain patentable despite Alice.

A jump in issued patents in a highly competitive, increasingly crowded market segment, could lead to an uptick in patent-related infringement. Already, estimates by some suggest that 35% more companies expect to face IP litigation in 2018 compared to 2017.

Artificial Intelligence Won’t Achieve Legal Inventorship Status Anytime Soon

Imagine a deposition in which an inventor is questioned about her conception and reduction to practice of an invention directed to a chemical product worth billions of dollars to her company. Testimony reveals how artificial intelligence software, assessing huge amounts of data, identified the patented compound and the compound’s new uses in helping combat disease. The inventor states that she simply performed tests confirming the compound’s qualities and its utility, which the software had already determined. The attorney taking the deposition moves to invalidate the patent on the basis that the patent does not identify the true inventor. The true inventor, the attorney argues, was the company’s AI software.

Seem farfetched? Maybe not in today’s AI world. AI tools can spot cancer and other problems in diagnostic images, as well as identify patient-specific treatments. AI software can identify workable drug combinations for effectively combating pests. AI can predict biological events emerging in hotspots on the other side of the world, even before they’re reported by local media and officials. And lawyers are becoming more aware of AI through use of machine learning tools to predict the relevance of case law, answer queries about how a judge might respond to a particular set of facts, and assess the strength of contracts, among other tools. So while the above deposition scenario is hypothetical, it seems far from unrealistic.

One thing is for sure, however; an AI program will not be named as an inventor or joint inventor on a patent any time soon. At least not until Congress amends US patent laws to broaden the definition of “inventor” and the Supreme Court clarifies what “conception” of an invention means in a world filled with artificially-intelligent technologies.

That’s because US patent laws are intended to protect the natural intellectual output of humans, not the artificial intelligence of algorithms. Indeed, Congress left little wiggle room when it defined “inventor” to mean an “individual,” or in the case of a joint invention, the “individuals” collectively who invent or discover the subject matter of an invention. And the Supreme Court has endorsed a human-centric notion of inventorship. This has led courts overseeing patent disputes to repeatedly remind us that “conception” is the touchstone of inventorship, where conception is defined as the “formation in the mind of the inventor, of a definite and permanent idea of the complete and operative invention, as it is hereafter to be applied in practice.”

But consider this. What if “in the mind of” were struck from the definition of “conception” and inventorship? Under that revised definition, an AI system might indeed be viewed as conceiving an invention.

By way of example, let’s say the same AI software and the researcher from the above deposition scenario were participants behind the partition in a classic Turing Test. Would an interrogator be able to distinguish the AI inventor from the natural intelligence inventor if the test for conception of the chemical compound invention is reduced to examining whether the chemical compound idea was “definite” (not vague), “permanent” (fixed), “complete,” “operative” (it works as conceived), and has a practical application (real world utility)? If you were the interrogator in this Turing Test, would you choose the AI software or the researcher who did the follow-up confirmatory testing?

Those who follow patent law may see the irony of legally recognizing AI software as an “inventor” if it “conceives” an invention, when the very same software would likely face an uphill battle being patented by its developers because of the apparent “abstract” nature of many software algorithms.

In any case, for now the question of whether inventorship and inventions should be assessed based on their natural or artificial origin may merely be an academic one. But that may need to change when artificial intelligence development produces artificial general intelligence (AGI) that is capable of performing the same intellectual tasks that a human can.

Marketing “Artificial Intelligence” Needs Careful Planning to Avoid Trademark Troubles

As the market for all things artificial intelligence continues heating up, companies are looking for ways to align their products, services, and entire brands with “artificial intelligence” designations and phrases common in the surging artificial intelligence industry, including variants such as “AI,” “deep,” “neural,” and others. Reminiscent of the dot.com era of the early 2000’s, when companies rushed to market with “i-” or “e-” prefixes or appended “.com” names, today’s artificial intelligence startups are finding traction with artificial intelligence-related terms and corresponding “.AI” domains. The proliferation of AI marketing, however, may lead to brand and domain disputes. But a carefully-planned intellectual property strategy may help avoid potential risks down the road, as the recent case Stella.ai, Inc. v. Stellar A.I, Inc., filed in the U.S. District Court for the Northern District of California, demonstrates.

Artificial Intelligence startups face plenty of challenges getting their businesses up and going. The last things they want to worry about is unexpected trademark litigation involving their “AI” brand and domain names. Fortunately, some practical steps taken early may help reduce the risk of such problems.

According to court filings, New York City-based Stella.AI, Inc., provider of a jobs matching website, claims that its “stella.AI” website domain has been in use since March 2016, and its STELLA trademark since February 2016 (its U.S. federal trademark applications was reportedly published for opposition in April 2016 by the US Patent and Trademark Office). Palo Alto-based talent and employment agency Stellar A.I., formerly JobGenie, obtained its “stellar.ai” domain and sought trademark status for STELLAR.AI in January 2017, a move, Stella.ai claims, was prompted after JobGenie learned of Stella.AI, Inc.’s domain. Stella.AI’s complaint alleges unfair competition and false designation of origin due to a confusingly-similar mark and domain name. It sought monetary damages and the transfer of the stellar.ai domain.

In its answer to the complaint, Stellar A.I. says that it created, used, and marketed its services under the STELLAR.AI mark in good faith without prior knowledge of Stella.AI, Inc.’s mark, and in any case, any infringement of the STELLA mark was unintentional.

Artificial Intelligence startups face plenty of challenges getting their businesses up and going. The last things they want to worry about is unexpected trademark litigation involving their “AI” brand and domain names. Fortunately, some practical steps taken early may help reduce the risk of such problems.

As a start, marketers should consider thoroughly searching for conflicting federal, state, and common law uses of a planned company, product, or service name, and they should also consider evaluating corresponding domains as part of an early branding strategy. Trademark searches often reveal other, potentially confusingly-similar, uses of a trademark. Plenty of search firms offer search services, and they will return a list of trademarks that might present problems. If you want to conduct your own search, a good place to start might be the US Patent and Trademark Office’s TESS database, which can be searched to identify federal trademark registrations and pending trademark applications. Evaluating the search results should be done with the assistance of the company’s intellectual property attorney.

It is also good practice to look beyond obtaining a single top-level domain for a company and its brands. For example, if “xyzco.ai” is in play as a possible company “AI” domain name, also consider “xyzco.com” and others top-level domains to prevent someone else from getting their hands on your name. Moreover, consider obtaining domains embodying possible shortcuts and misspellings that prospective customers might use (i.e., “xzyco.ai” transposes two letters).

Marketers would be wise to also exercise caution when using competitor’s marks on their company website, although making legitimate comparisons between competing products remains fair use even when the competing products are identified using their trademarks. In such situation, comparisons should clearly state that the marketer’s product is not affiliated with its competitor’s product, and website links to competitor’s products should be avoided.

While startups often focus limited resources on protecting their technology by filing patent applications (or by implementing a comprehensive trade secret policy), a startup’s intellectual property strategy should also consider trademark issues to avoid having to re-brand down the road, as Stellar A.I. did (their new name and domain are now “Stellares” and “stellares.ai,” respectively).

Federal Circuit: AI, IoT, and Robotics in “Danger” Due to Uncertainty Surrounding Patent Abstraction Test

In Purepredictive, Inc. v. H2O.ai, Inc., the U.S. District Court for the Northern District of California (J. Orrick) granted Mountain View-based H2O.ai’s motion to dismiss a patent infringement complaint. In doing so, the court found that the claims of asserted U.S. patent 8,880,446 were invalid on the grounds that they “are directed to the abstract concept of the manipulation of mathematical functions and make use of computers only as tools, rather than provide a specific improvement on a computer-related technology.”

Decisions like this hardly make news these days, what with the frequency by which software patents are being invalidated by district courts across the country following the Supreme Court’s 2014 Alice Corp. Pty Ltd. v. CLS Bank decision. Perhaps that is why the U.S. Court of Appeals for the Federal Circuit, the specialized appeals court for patent cases based in Washington, DC, chose a recent case to publicly acknowledge that “great uncertainty yet remains” concerning Alice’s patent-eligibility test, despite the large number of post-Alice cases that have “attempted to provide practical guidance.”  Calling the uncertainty “dangerous” for some of today’s “most important inventions in computing” (specifically identifying medical diagnostics, artificial intelligence (AI), the Internet of Things (IoT), and robotics), the Federal Circuit expressed concern that perhaps Alice has gone too far, a belief shared by others, especially smaller technology companies whose value is tied to their software intellectual property.

Utah-based Purepredictive says its ‘446 patent involves “AI driving machine learning ensembling.” The district court characterized the patent as being directed to a software method that performs “predictive analytics” in three steps. In the method’s first step, the court said, it receives data and generates “learned functions,” or, for example, regressions from that data. Second, it evaluates the effectiveness of those learned functions at making accurate predictions based on the test data. Finally, it selects the most effective learned functions and creates a rule set for additional data input. This method, the district court found, is merely “directed to a mental process” performed by a computer, and “the abstract concept of using mathematical algorithms to perform predictive analytics” by collecting and analyzing information.

Alice critics have long pointed to the subjective nature of Alice’s patent-eligibility test. Under Alice, for subject matter of a patent claim to be patent eligible under 35 U.S.C. § 101, it may not be “directed to” a patent-ineligible concept, i.e., a law of nature, natural phenomenon, or abstract idea. If it is, however, it may nevertheless be patentable subject matter if the particular elements of the claim, considered both individually and as an ordered combination, add enough to transform the nature of the claim into a patent-eligible application. This two-part test has led to the invalidation of many software patents as “abstract,” and presents an obstacle for inventors of new software tools seeking patent protection for their inventions.

In the Purepredictive case, the district court found that the claim’s method “are mathematical processes that not only could be performed by humans but also go to the general abstract concept of predictive analytics rather than any specific application.” The “could be performed by humans” query would seem problematic for many software-based patent claims, including those directed to AI algorithms, despite the recognition that humans could never perform the same feat as many AI algorithms in a lifetime due to the enormous domain space these algorithms are tasked with evaluating.

In any event, while Alice’s abstract test will continue to pose challenges to those seeking patents, time will tell whether it will have “dangerous” impacts on the burgeoning AI, IoT, and robotics industries suggested by the Federal Circuit.

Sources:

Purepredictive, Inc. v. H2O.AI, Inc., slip op., No. 17-cv-03049-WHO (N.D. Cal. Aug. 29, 2017).

Smart Systems Innovations, LLC v. Chicago Transit Authority, slip. op. No. 2016-1233 (Fed. Cir. Oct. 18, 2017) (citing Alice Corp. Pty Ltd. v. CLS Bank, 134 S. Ct. 2347, 2354-55 (2014)).