Software Litigation Consulting

Andrew Schulman
Consulting Technical Expert & Attorney

NotebookLM-generated “comprehensive briefing” on rough notes for “Patent Litigation” book

I’m using Google’s NotebookLM to help me finish my long-delayed book on Patent Litigation (see here). NotebookLM takes your documents, and automatically generates several new summary documents, as well as audio.

One of the documents NotebookLM generates is a “Comprehensive Briefing,” shown below with heavy commentary from me in [square-bracketed bold italic] like this. Especially given the generally sloppy state of the notes I provided, NotebookLM did a remarkable job producing this document. The quotations in NotebookLM’s briefing doc come from my notes, and tend to focus on my key catch-phrases such as “business end” of the patent (claims), “The Next One” (see below), “negative IP”, etc., which are often not intuitive. A few, like “dollop of state power,” come from sources I cited, as noted below.]

In a few places, NotebookLM’s summary is plainly wrong (e.g. the description below the NPEs/trolls “specialize solely in patent licensing”). One role I can see for writing with an AI tool such as NotebookLM is in expecting it to generate some misunderstandings and confusions, which might also be typical of what a human reader would think when reading my material, and then trying to explicitly address those.]

While I’m mainly interested in actually making headway with my book, and using NBLM as part of that, NBLM’s “comprehensive briefing” also provides a useful example of AI summarization, which is a subject I’ve discussed elsewhere on this site.

With a RAG such as NotebookLM, issues include the following (this list is based on the RAG chapter in Kamath et al., Large Language Models: A Deep Dive Bridging Theory and Practice (2024), which also covers summarization with RLHF):

  • recall (did it get everything important?) as well as precision (is it correct?);
  • a desire for “interestingness” (hoping the summary will surface something unexpected yet grounded in the source docs);
  • specifics over generalities;
  • not mimicking the sources verbatim;
  • not mimicking the prompt verbatim;
  • “faithfulness” both to the sources and to the prompt;
  • “missing middle” problems with large source docs exceeding the context window size.

Since I’m intimately familiar with the docs here (I wrote them, such as they are), I can assess whether NBLM “got it”. Interestingly, the NBLM-generated audio in some ways seems better at capturing nuance than the text below — why?

My comments below often simply expand on how NotebookLM (NBLM) summarized my notes:

  • The NBLM original summary is often more understandable than my own summary of my material would have been. This is one example of how some “futzing” with LLMs is useful, even when it ends up taking more time than “I’ll just do it myself” would have been.
  • I’d like to get NBLM to take my commented version below, and use my my bold-italic comments as examples for further elaboration: “do what I did but keep going.” Feeding in comments/critiques as a form of user-supplied RLHF (reinforcement learning with human feedback)?
  • If an NBLM-generated doc is deleted and then regenerated, will the new version be substantially different (in the way AI chatbot answers are different each time, due to LLM “temperature” setting)?
  • NBLM does NOT train on your uploaded/linked sources (see Steven Johnson article introducing NotebookLM). It is a RAG (retrieval augmented generation): in effect, your sources are added to the prompt, not used to train the underlying model’s weights. So on the one hand, no worries about NBLM being trained on your private docs. But on the other hand, like some other authors, I want to “write for LLMs” (LLMs are part of my target audience).

I asked NBLM how many pages or words there were in my several dozen docs/links, and (perhaps not surprisingly) it couldn’t tell me, apart from pointing to this or that place where one of my docs explicitly noted its own length. NotebookLM answers questions from the contents of your docs, seemingly without any “meta” understanding about the docs. It also limits itself to the literal text in your docs, without using metadata such as headers, TOCs, etc. that would reflect the structure of your docs.

With that preface, below is what NotebookLM automatically generated from my rough notes, with my comments in bold italics.

Patent Litigation: A Comprehensive Briefing

This briefing provides a detailed overview of key themes, concepts, and factual considerations in patent litigation, drawing from various source materials. It aims to clarify the nature of patents as a form of property, their role in incentivizing innovation, the mechanics of infringement and invalidity, and the strategic landscape of patent disputes.

[Had I written this, I would have referred to “patents as both a somewhat odd and also somewhat typical form of property”, and referred to “their possible role in incentivizing (or perhaps sometimes not) innovation”, and not to the mechanics of infringement/invalidity as such, but rather to the mechanics of finding and providing evidence of infringement/invalidity.] 

1. Patents as Property: Boundaries, Exclusion, and Unstable Nature

A patent is fundamentally a form of intellectual property (IP), distinct from traditional real or personal property, yet sharing key attributes. It grants a temporary exclusive right over an invention, functioning as a right to exclude others, rather than a permit to practice the invention oneself. [Why only “fundamentally”? It’s not like there are scenarios in which a patent is not IP, right?]

  • Boundaries and Scope: Patents define “boundaries” within the world of technology, similar to a title deed to land. These boundaries represent the “scope” of the invention – how large it is and its technological location. This concept is crucial for determining infringement and invalidity. “Patent claims are made of ‘limitations’; FIGURES: map/stakes, ‘business end’ p.20.” These “limitations” act as boundary markers, defining the property by setting its outer limits. “The real-property term is ‘metes and bounds,’ and this term is also used within patent law.”
  • Right to Exclude, Not to Use: A patent “is not a permit to use technology, but rather is a right to exclude others from doing so.” This means that even with a patent, an inventor might still need a license from a prior patent holder, especially in fields with “dense ‘thicket’ of patents.” This can lead to “blocking patents” and “deadly embrace” scenarios, often resolved through cross-licensing.
  • Not Self-Acting: Unlike a self-enforcing right, “Patents are not self-acting, and the ‘patent police’ are not on their way.” Enforcement is the responsibility of the patent owner, who has a “right to sue for infringement.” This “dollop of federal power” is placed in the hands of private citizens. [Is there any such thing as a “self-enforcing right”, anywhere in the law? (“No right without a remedy.”) The “right to sue for infringement” is based not only on ownership of a patent, but also on presenting a “plausible” showing that evidence may be found that the defendant infringed. The “dollop of federal power” phrase comes from Merges, American Patent Law: A Business and Economic History (2023). Reference could have been made to notion that “getting a patent is just buying a lawsuit,” and the sense in which that is literally true.]
  • Unstable Property: A seemingly paradoxical aspect of patents is their inherent “shaky/uncertain as well as temporary” nature. A patent, “granted by a government agency (in the U.S., the PTO) after a lengthy examination, might nonetheless be invalid.” This “wiping out” of a patent is a significant difference from other forms of property. The validity and scope of a patent are “never irrevocably vests” and are often only truly “tested in expensive ‘crucible of litigation’.” This uncertainty can be a “source of strength” through its “scarecrow” or “penumbra” effect, deterring competitors, but also a weakness due to its inherent unpredictability. [There have been calls to make patent validity irrevocable after some time; see e.g. “settled expectations” doctrine at the PTAB. Is revocability really so different from e.g. real property (on the one hand, eminent domain; on the other hand, quiet title and title insurance)?]

2. The Role of Patent Claims and Limitations

Patent claims are the operational core of a patent, serving as the primary tool for determining infringement and invalidity. They are meticulously constructed and interpreted. [Better to not use the word “constructed” here, since that is a term of art referring to claim interpretation. But good point that a patent claim is “constructed” in the more typical sense of that word: not merely written down once and for all time, but carefully built up and modified in a back-and-forth process between the patent applicant and the patent office, and possibly even revised post-grant.]

  • “Business End” of a Patent: “Patent claims are the ‘business end’ of a patent, i.e., what make it actionable and operative.” They are not typically pictures of the invention but rather “devices for testing patent infringement and invalidity.” [Not a picture: i.e., not a verbatim reflection of what was in the inventor’s mind; but note that in the book’s first major example (Claim 11 in US 7,421,032 in Caltech v. Apple/Broadcom) one of the claims, atypically, is little more than a literal picture, of a “Tanner graph.”]
  • Made of Limitations: Claims are largely composed of “limitations,” which are “selected elements or steps implementing an invention.” These are not exhaustive “parts lists” but a “subset of elements of an invention necessary to differentiate/distinguish the claim from earlier technology (‘prior art’), while still trying to leave a wide scope of infringement.”
  • Claim Scope: The “length of a patent claim is a rough indication of claim scope… longer claims generally… protect against less infringement than do shorter claims.” Counter-intuitively, “the patent owner wants the claim to be infringed — no value without likelihood of someone else practicing the patent.” Adding limitations generally reduces scope but increases validity likelihood. [All things being equal, Less Is More.]
  • Claim Construction (Markman): Interpreting the meaning of terms within a patent claim is a critical phase known as “claim construction.” This process “uses the patent specification (and a hierarchy of other sources) to interpret the meaning of terminology in a claim.” It is a complex process often involving “lawyer games” of parsing words, even small, seemingly insignificant ones, to establish the precise meaning for legal analysis. [Note that claim construction is intended to be independent of the accused product, yet also occurs in the context of a particular accused product or piece of asserted prior art. The potential dependence of case outcome on parsing of small words is a source of risk.]
  • Mapping Claims to Products/Prior Art: A central activity in litigation involves a “limitation-by-limitations comparison of a patent claim with an accused product” or prior art. This structured comparison, often presented in “claim charts,” is essential for “showing infringement” or invalidity. [Limitation-by-limitation means not a holistic exercise.]

3. Infringement and Invalidity: The Core of Litigation

Patent litigation revolves around demonstrating that an accused product or method “infringes” a patent, or conversely, that the patent itself is “invalid.” These two concepts are often seen as “sort-of mirror image.” [Mirror image: “That which infringes, if after, thereby anticipates/invalidates, if before.” But only a “sort-of” mirror image, because infringement can be shown merely with preponderance of the evidence, whereas invalidity in litigation requires more: clear and convincing evidence.]

  • Infringement: Defined by federal law (35 USC 271), infringement includes making, using, selling, offering for sale, or importing a patented invention during its temporary life.
  • Literal Infringement: Occurs when every limitation in a patent claim is present in the accused product or method. [Note that “literal” does not mean verbatim. The NotebookLM-generated audio has a good explanation of how defendant may use nomenclature in describing the accused product that is entirely different from language of the plaintiff’s patent claim, yet there can still be literal infringement.]
  • Doctrine of Equivalents (DoE): Infringement can also occur if the accused product performs “the same work (F), in substantially the same way (W), to achieve substantially the same result (R)” as the claimed invention, even if differing in “name, form, or shape.” [This unfortunately suggests a holistic approach to equivalence, looking at the entire accused product. Instead, the F/W/R test is on an element-by-element basis.]
  • Indirect Infringement: Includes inducing or contributing to another’s direct infringement.
  • Willful Infringement: Involves knowledge and intent, potentially leading to treble damages.
  • Invalidity: Even a granted patent can be challenged and found invalid. Key grounds for invalidity (primarily under 35 USC 101, 102, 103, and 112) include:
    • Anticipation (Lack of Novelty): The invention was already disclosed in prior art (35 USC 102). “If infringe if after, invalidates if before.”
    • Obviousness: The invention, even if not directly anticipated, would have been “obvious” to a “person having ordinary skill in the art” (PHOSITA) at the time of invention (35 USC 103). This often involves assessing “motivation to combine” existing prior art. [Need a little more on anticipation basis generally in a single piece of prior art, whereas obviousness based on combination of multiple pieces of prior art.]
    • Subject Matter Eligibility: The invention does not fall into a patentable category (35 USC 101), as highlighted by decisions like Alice, Bilski, and Mayo.
    • Lack of Enablement: The patent specification does not sufficiently describe how to make and use the invention without “undue experimentation” (35 USC 112). This disclosure is theoretically the “payment” the inventor makes for the patent. [Of course, patent owner makes actual monetary payments, including to maintain a granted patent over time, but this is referring to what the inventor must provide in exchange for the patent: disclosure of how to make and use the full scope of the invention.]

4. Incentives, Progress, and the Patent System’s Dual Nature

The patent system is designed to “promote progress” by incentivizing invention and disclosure, yet it operates with inherent contradictions and complexities. [“Promote progress of science and the useful arts” of course comes from the US Constitution. The reference to “inherent” contradictions is based on the notes I provided, but it’s not really explained here. The NotebookLM-provided audio is good on contradictions, picking up for example on my notes that try to show how a system based on exclusion (ownership) can nonetheless possibly promote diffusion.]

  • Incentive to Invent and Disclose: The fundamental premise is that a “temporary monopoly” encourages the creation and sharing of new knowledge. This “incentive from pats should be disincentive to TS” (trade secrets), pushing information into the public domain through patent disclosures. [Should probably explain this isn’t a “monopoly” in the strict economic sense (ability to dictate prices within a certain market structure).]
  • “The Next One”: Incentives are largely “re: the future,” encouraging continued innovation and “follow-on improvements.” The system aims to “goad to contrive small move away from prior art,” leading to a “treadmill” or “arms race” of innovation. [Make clearer that patents are primarily not intended as a reward for a job well done (i.e. re: the past). The “contrive small move” came from my notes specifically re: obviousness doctrine pushing inventors just a bit further away from prior than anticipation doctrine does; that point in turn from Scotchmer, Innovation and Incentives (2006).]
  • Diffusion and Spillover: While granting exclusivity, the system also implicitly encourages diffusion of knowledge through its disclosure requirement and the possibility of “design-around” or non-infringing alternatives (NIAs). This “leaky” nature, where knowledge “spills over,” is considered necessary for economic growth. [Necessity for “leaks” (incomplete appropriation by the inventor) comes from Paul Romer, via Warsh, Knowledge and the Wealth of Nations: A Story of Economic Discovery (2007).]
  • “Negative IP”: Patents can serve a “negative” function by creating prior art that blocks competitors from patenting the same technology, even if the patent itself is never actively asserted [or even if patent not granted, but patent application published]. This “poisoning the well” [from perspective of the owner of the patent or patent application] expands the public domain.
  • Channeling Inventive Activity: The patent system, through its structure and requirements (especially the “claim format”), “pushes inventive thinking/direction” and can “routinize invention.” It encourages inventors to “think in terms of claims,” leading to both generalization and particularization. [These I think are some of the more important points I try to make in the book, and it would be good to expand this into several bullet points.
    • For example, patent claims are a form of “reification”, providing a way to work with inventions at a remove from a particular embodiment of the invention.
    • An abstraction, but one that is still concrete (at least to the extent of so-called “constructive reduction to practice”).
    • Oren Bracha’s book Owning Ideas: The Intellectual Origins of American Intellectual Property, 1790-1909 pp. 301-6 is good on the shift to “peripheral claiming,” suggesting its current role in creating reified handles to technology. (Does this mean that earlier patent systems, with so-called “central claiming” and physical models, or without claiming at all, did not “channel inventive activity”? or that they did so in a different way?)
    • That role is emphasized in the “Litigation Machines” and “Text Machines” chapters of Pottage & Sherman, Figures of Invention: A History of Modern Patent Law (2010).
    • A goal of my book is to combine coverage of topics such as this on the one hand, with in-depth focus on e.g. claim charts, expert reports, and source code as evidence, on the other hand. An ideal page in the book might footnote both to Pottage & Sherman, and to a solid practitioner’s guide like Robert Fish’s White Space Patenting: Patenting Ideas Not Just Inventions (2023).]
  • Raising Prices: While a “monopoly,” patents are not just about raising prices. However, it is acknowledged that “the very goal is to (temp) give P ability to raise prices above what they otherwise would be w/o patents, or alternatively to increase P’s customer base.” This is framed as a sacrifice now for future progress.
  • [The NBLM summary is missing an important issue raised in my notes: to what extent is the patent system “self-referential”? I asked NBLM, and it provided a useful answer, though without any specific examples from my notes.]
  • “Schizo Mishmash”: The patent system is acknowledged as a “schizo mishmash that pushes/pulls in multiple directions,” a product of evolving statutes and common law rather than a single coherent design. Despite this, it functions as a “system” that aims to impose some “regularity to invention.” [I thought that NotebookLM came up with this “schizo mishmash” characterization, as I didn’t remember writing it, and I’m not sure I totally agree with it, but yes, it’s in my original notes. See below, which also shows an example of the “mishmash” :-) of my notes that NotebookLM had to work with.]An example of my very rough notes that NotebookLM had to deal with, to generate the summaries, audio, etc. discussed here

5. Patent Litigation: Strategy, Process, and Practicalities

Patent litigation is a complex, often expensive process, serving both private goals for patent owners and a crucial public role in testing the patent system’s boundaries. [The public/social role of patent litigation (is litigation a blemish on the patent system, and/or a crucial part of it?) is a major topic of the book, and should have been covered in the “Incentives, Progress, and the Patent System’s Dual Nature” section above, or should be covered here, with the subhead changed to reflect this. And of course, besides a given lawsuit helping test the overall system’s boundaries (through e.g. the development of case law that establishes the meaning of raw statutory language), a lawsuit also specifically tests the boundaries of the patents at suit.]

  • Why Litigate?: Private goals for patent owners include protecting products, generating revenue streams through injunctions or monetary damages (lost profits, reasonable royalties), signaling to investors, enhancing reputation, creating bargaining chips for cross-licensing, deterring competitors, and increasing company valuation.
  • “Just Buying a Lawsuit”: A patent, in essence, provides “the power to back up the federal right in court.” It grants the owner the right to initiate a “plausible lawsuit,” which in turn provides “a right to go rooting around in the defendant’s closets for evidence of infringement.” [See “Discovery” below.]
  • Plausibility and Claim Charts: Bringing a case requires “factually plausible infringement contentions,” based on “reasonable pre-filing investigation.” “Local Patent Rules (LPRs) in key federal districts mandate how litigations [sic; litigants] set forth their infringement and invalidity contentions and responses,” typically expressed as “claim charts.” These charts are “workhorse of patent litigation” for marshalling facts and pinpointing infringement.
  • Discovery: This phase allows each side to “obtain information, including confidential information, held by the other side,” often including source code or other technical data. This information is crucial for “mapping claims to products/services.”
  • Experts: Technical and economic experts play a vital role in litigation, assisting with claim construction, infringement/invalidity analysis, and damages calculations. Their opinions must be well-supported and avoid “conclusory” statements. [This bare-bones summary of expert role doesn’t reflect the bulk of the book’s planned coverage of experts in patent litigation, because the NotebookLM summary was generated before linking to some material at my site, especially “Expert witnesses & non-testifying consultants in software patent litigation” in notes for forthcoming book on Source Code in Patent Litigation. “Opinions must be well-supported” is a placeholder for extensive coverage of expert methodology and factual basis underlying expert opinions under FRE 702 , and of expert reports under FRCP 26(a)(2).]
  • Remedies: If infringement is found, remedies can include injunctions (to stop infringement) and monetary damages (lost profits or reasonable royalties). The concept of “Smallest Saleable Patent-Practicing Unit (SSPPU)” is relevant in determining damages when a patent covers only a small part of a larger product. [This gives too much emphasis to SSPPU, and too little to lost profits (Panduit factors) and reasonable royalties (Georgia-Pacific factors); the NotebookLM-generated audio discusses G-P factors.]
  • Alternatives and Non-Litigation Uses: While litigation is central, patents also have significant non-litigation uses, such as signaling to VCs, enhancing company reputation, facilitating cross-licensing, and acting as a deterrent (e.g., through “mutually assured destruction” or the “FUD effect”). Some companies, known as Non-Practicing Entities (NPEs) or “trolls,” specialize solely in patent licensing, operating in the “knowledge economy” as a “pure play.” [Saying that NPEs/trolls “specialize solely in patent licensing” misses the mark.]
  • “Competition by Other Means”: Patent litigation is viewed as a form of “competition by other (non-market) means.” It’s a strategic battle, where the litigation itself can be leverage for negotiation and settlement, with most cases never reaching trial. [“Strategic battle” is much too vague.]

In conclusion, patent litigation is a nuanced and dynamic field. It is deeply rooted in the legal construct of patents as a unique form of property, designed to incentivize innovation through a carefully balanced, albeit sometimes contradictory, system of rights and responsibilities. The “name of the game is the claim,” as the precise wording and interpretation of patent claims and their limitations dictate the outcomes of these often high-stakes disputes. [“Unique form of property” misses the book’s attempted emphasis on patents as having some fairly standard attributes of property, especially boundaries. Wording/interpretation do NOT “dictate” the outcome of patent litigation, even though they are said to often be dispositive. The outcome may also be dictated by the facts of infringement or invalidity, by strategic/tactical considerations, etc.]

[I want to think a little more about which key themes NotebookLM failed to extract from my mass of notes. Above I note that it didn’t really get into the social role of patent litigation (is each lawsuit just a blemish on an otherwise-pure patent system, or is the possibility of litigation crucial to the system itself, in particular because litigating the validity of patents (as happens when P’s infringement suit motivates D’s efforts to have the patent invalidated) enables what is called “rational ignorance” at the patent office (spending only 20 hours per patent, for the vast majority of patents that will never be litigated)? And it missed my notes on the arguably “self-referential” nature of the patent system, though when explicitly asked about that, it generated a decent summary. Other themes missing here, but central to the book, are the relation between procedure and substance, the structure/function relationship in the patent system, and the production of big results from seemingly-small inputs (again, the NBLM audio covers this, using the Caltech v. Broadcom/Apple example from my notes for chapter 1 of the book).]

 

 

Print Friendly, PDF & Email