Underscoring the Importance of Human Expert Oversight When Using Artificial Intelligence Tools
Like many, I’ve been intrigued by the future prospect of artificial intelligence, including its applications specifically in the practice of IP law. But like others, I remain equally cautious in its adoption as a replacement for experienced and expert humans in certain aspects of the application of legal work.
A recent exercise in exploring both the prospects and potential limitations of AI illustrates and illuminates both sides of the aisle — the intrigued and the intrepid. While queries were addressed nearly instantaneously, including the research of relevant case law and analysis thereof, a key omission/fabrication was made — one that perhaps would go undetected by someone not well-versed in the relevant statute and legal precedence.
Specifically, in this instance and in many others anecdotally speaking, a Language Learning Model (LLM) was prone to “hallucinate” — making up case law from whole cloth — and doing so with apparent conviction and confidence. Were there no expert human committed to verifying and validating the “work” that AI was performing, people might otherwise get into real trouble…perhaps even to the point of malpractice.
I’m happy to share my work with you as my own personal cautionary tale. Allow me to recount the experience, starting from my initial prompt, all the way through to its illustrative conclusion. As someone new to AI, admittedly, I offer the tale of my first experience with an AI legal assistant, warts and all…
I Asked an AI System a Seemingly Simple Question. The Answer Wasn’t.
Query
Act as an expert patent attorney. A patent examiner has objected to the term “strain gage”and wants it to be spelled “strain gauge.” Develop an argument as to why “strain gage” is a correct use of the term to a person having ordinary skill in the art.
AI Response
Thank you for the opportunity to address this issue. As an expert patent attorney, I would respond to the patent examiner’s objection as follows:
The term “strain gage” is commonly used and well understood in the art to refer to a sensor that measures strain. While “strain gauge” may be considered the more formal spelling, “strain gage” is an accepted variant spelling that has been widely used in the relevant literature for decades. For example, searches of IEEE journals and patents show thousands of uses of “strain gage” dating back to the 1950s and 1960s.
The Manual of Patent Examining Procedure recognizes that variant spellings of terms may be acceptable if the meaning is clear:
“Variations in spelling should not be objected to if the meaning is not changed thereby. Similarly, spelling variations which clearly do not change pronunciation should not be objected to.” MPEP 608.01(g)
Here, the meaning and pronunciation of “strain gage” are clear to a person having ordinary skill in the art. It immediately conveys the commonly understood meaning of a sensor that measures strain.
Additionally, Applicant has been consistent in using the spelling “strain gage” throughout the specification and claims. This consistent usage demonstrates that the intended meaning is clear.
In view of the widespread accepted use of the term “strain gage” in the art, as well as Applicant’s consistent specification usage, a person having ordinary skill in the art would readily understand the intended meaning of “strain gage.” The variant spelling does not render the claims indefinite or unclear.
Therefore, the objection to the spelling of “strain gage” should be withdrawn. I would be happy to discuss this further and provide additional evidence as needed to demonstrate that “strain gage” is an appropriate variant spelling in this context. Please let me know if any additional clarification would be helpful.
Critical Error!
Whoa! Did you catch the issue? The AI tool quotes the MPEP at 608.01(g). Except MPEP 608.01(g), and indeed no part of the MPEP, ever talks about spelling or pronunciation.
The above-quoted language of
“Variations in spelling should not be objected to if the meaning is not changed thereby. Similarly, spelling variations which clearly do not change pronunciation should not be objected to.” MPEP 608.01(g)”
is a complete fabrication…
Actual Language of MPEP 608.01(g) is as follows:
MPEP 608.01(g) Detailed Description of Invention [R-10.2019]
A detailed description of the invention and drawings follows the general statement of invention and brief description of the drawings. This detailed description, required by 37 CFR 1.71, MPEP §§ 608.01, 2161, and 2162, must be in such particularity as to enable any person skilled in the pertinent art or science to make and use the invention without involving extensive experimentation and must clearly convey enough information about the invention to show that applicant invented the subject matter that is claimed. An applicant is ordinarily permitted to use his or her own terminology, as long as it can be understood. Necessary grammatical corrections, however, should be required by the examiner, but it must be remembered that an examination is not made for the purpose of securing grammatical perfection.
The reference characters must be properly applied, no single reference character being used for two different parts or for a given part and a modification of such part. See 37 CFR 1.84(p). Every feature specified in the claims must be illustrated, but there should be no superfluous illustrations.
The description is a dictionary for the claims and should provide clear support or antecedent basis for all terms used in the claims. See 37 CFR 1.75, MPEP § 608.01(i), § 608.01(o), and § 1302.01, and § 2111.01.
Let’s Try This Again
Follow-On Query
I am confused. The MPEP makes no mention of pronunciation of a technical term. Please review, correct this error, and update the analysis.
AI Response
You’re absolutely right, my previous reference to pronunciation was incorrect. Let me revise my analysis:
The MPEP specifically states that “Variations in spelling should not be objected to if the meaning is not changed thereby.” MPEP 608.01(g). Here, the meaning of “strain gage” is clear and well-established in the art as referring to a strain measurement sensor. The spelling variation from “gauge” to “gage” does not change the meaning at all.
Rather than pronunciation, a better justification is that “strain gage” has been commonly used as a variant spelling for this term in the literature for many years. This establishes it as an accepted spelling among those skilled in the art.
IEEE journals, patents, and industry texts confirm the widespread usage of “strain gage” dating back decades. Thus, a person having ordinary skill in the art would readily understand this term and the spelling variation would not render the claims indefinite.
In summary, the variant spelling “strain gage” is appropriate here because:
– The meaning is unchanged from “strain gauge”, as confirmed by accepted usage.
– It has been commonly used in the field for many years, establishing it as a known variant spelling.
– The MPEP allows for such recognized variants if the meaning is clear.
Thank you for catching my mistaken reference to pronunciation. Please let me know if this revised analysis properly addresses the issue and objection. I appreciate you helping me strengthen my arguments and referencing skills.
__________________________
Final Query
Thank you. What is the case law regarding pronunciation of terms in a patent?
AI Response
You raise an excellent follow-up question. After reviewing the case law, I do not see any instances where pronunciation of a term was a deciding factor in a patent case regarding claim definiteness or enablement.
The focus is consistently on whether the spelling variant conveys the same meaning that is clearly understood by a person having ordinary skill in the art. Pronunciation or phonetics are not mentioned.
For example, in prestigious cases like Phillips v. AWH Corp. and Nautilus v. Biosig Instruments, the analysis relies on the understanding of a term within the relevant field of art based on context, without referencing pronunciation.
Additionally, a search of the MPEP shows no mentions of using pronunciation as criteria for evaluating variant spellings or claim term disputes.
In summary, while pronunciation may be relevant in trademarks or linguistic analysis, patent case law and guidelines do not establish pronunciation as a material factor when evaluating claim terms. The primary considerations are the meaning and usage understood by skilled artisans.
Please let me know if I can provide any other cases, references or clarification on this issue! I appreciate you pushing me to deepen my understanding of the nuances in patent law. Discussions like this are extremely helpful for improving my analytical skills.
In Closing: Trust the Experts; Verify AI’s Work
The main issue that has found people in trouble is the belief that a LLM has specific recall. I would challenge anyone to reconsider whether LLMs truly “understand.” It’s been shown time and again that they are known to hallucinate and make stuff up that sounds convincing — especially to the non-expert. The problem often arises in that they attempt to bridge the gap from general understanding to specific recall and/or attempt to quote from a memory bank that is not stored to the necessary level of detail.
In some cases, if you prompt AI to perform a patent search, it will make up patent references (and actually assign them patent numbers to look legitimate!) to make the prompter believe that it knows what it’s talking about… just as it did with the above MPEP example.
All of this, to me, raises valid concerns and questions about the use of LLMs in the practice of IP law. With that said, LLMs can also be very powerful research tools and administrative assistants, if one does not rely solely on them — especially in the replacement of trained and experienced attorneys who can recognize and correct their shortcomings.
It is not that AI will replace attorneys in the short term. But it may very well be the case that it is the attorney who learns to master and optimize AI who will outpace and outperform those who don’t.
Thanks to our own expert, David Nay, who contributed his wisdom to this article.
To discuss how the application of AI may or may not apply to protecting your own ideas and inventions, contact Stephen Mahan at smahan@quinniplaw.com, or call him at (248) 380-9300, extension 119.