Artificial Intelligence versus the MPEP: A Case Study in Why We Can’t Yet Rely on AI

“The lesson is that it’s too soon to rely on GAI tools alone, the MPEP is here to stay, and any information provided by the tools will have to be validated.”

artificial intelligenceGenerative Artificial Intelligence (GAI) promises to improve information gathering and reduce costs in a myriad of economic processes. Since patent law is essentially about recording and accessing information, tools like ChatGPT are great candidates to ease the tasks of patent professionals.

As of today, it is not possible to trust GAI tools with confidential information, since any prompt provided to these tools does not guarantee confidentiality and data protection, unless an entity creates its own sandbox with customized implementations that protect the data.

For patent prosecution, asking procedure questions seems like an appropriate use of GAI, since client information is not compromised, and who wouldn’t celebrate with great cheer being able to bury that good old MPEP (Manual of Patent Examining Procedure).

I recently had a scenario that does not happen often, so I decided to ask ChatGPT and Bard for help. My problem was that the name of the assignee in an assignment was incorrect because the company name was missing the  “Inc.” at the end.

I asked the question, “how do you correct an incorrect company name for an assignment at the USPTO.”

ChatGPT answered the following:

If you need to correct an incorrect company name for an assignment at the USPTO, you should follow the following steps:

  1. Identify the type of assignment that needs to be corrected. Is it a trademark assignment, patent assignment, or another type of assignment?
  2. Check the USPTO’s website to see if they have any specific instructions on how to correct an assignment. You can also contact the USPTO’s Assignment Recordation Branch for assistance.
  3. Prepare a letter or form that clearly identifies the assignment that needs to be corrected, and explain the reason for the correction.
  4. Include any supporting documentation that may be necessary to support the correction.
  5. Submit the letter or form and any supporting documentation to the USPTO’s Assignment Recordation Branch. You can submit the correction electronically through the Electronic Patent Assignment System (EPAS), or by mail.

It’s important to note that there may be fees associated with correcting an assignment, so be sure to check the USPTO’s fee schedule. Additionally, you should always consult with an attorney or legal professional if you have any questions or concerns about the assignment correction process.

I thought that this did not help at all. An answer that includes “Check the USPTO’s website to see if they have any specific instructions,” would not help me overcome my wish to avoid the MPEP.

I asked Bard, and this was the answer:

To correct an incorrect company name for an assignment at the USPTO, you must file a corrected cover sheet with the Assignment Services Division. The corrected cover sheet must include the following information:

    • The correct company name
    • The incorrect company name
    • The application or patent number(s) that are affected by the error
    • A statement explaining the error
    • A signature of the person authorized to correct the error

You must also pay the required fee for each application or patent that is being corrected. The fee is currently $200 per application or patent.

You can file the corrected cover sheet online or by mail. If you are filing online, you will need to create an account with the USPTO and pay the fee using a credit card. If you are filing by mail, you will need to send the corrected cover sheet, the required fee, and a copy of the originally recorded papers to the following address:

United States Patent and Trademark Office

Assignment Services Division

PO Box 1450

Alexandria, VA 22313-1450

Once the corrected cover sheet is filed, the USPTO will update its records to reflect the correct company name.

Here are the steps on how to correct an incorrect company name for an assignment at the USPTO:

    1. Gather the required information. This includes the correct company name, the incorrect company name, the application or patent number(s) that are affected by the error, a statement explaining the error, and a signature of the person authorized to correct the error.
    2. Pay the required fee. The fee is currently $200 per application or patent.
    3. File the corrected cover sheet. You can file the corrected cover sheet online or by mail.
    4. Wait for the USPTO to update its records. The USPTO will update its records to reflect the correct company name once the corrected cover sheet is filed.

For more information, please visit the USPTO website: https://www.uspto.gov/patents-getting-started/correcting-errors

I was excited when I saw this answer, not only because it seemed to be on point, but because the answer showed that I didn’t have to go back to the inventors to get the signatures for the correction. It even included the link to a page where I could validate the answer.

Unfortunately, my initial joy quickly soured. First, the link was not valid and gave me the dreaded 404 error. Second, the information was for correcting the cover sheet, not an error in the assignment itself. Close, but no cigar!

I could have tried to refine the answer by adding more specific prompts, but I couldn’t trust any answer now, so I went back to the old ways: asking colleagues, looking at the MPEP, and even calling the Assignments Division at the USPTO. Unfortunately, I do have to chase inventors for new signatures in a corrective assignment.

The lesson is that it’s too soon to rely on GAI tools alone, the MPEP is here to stay, and any information provided by the tools will have to be validated. The good news is that this means job security for us patent professionals. For how long, we’ll find out soon enough.

Image Source: Deposit Photos
Image ID:185517156
Copyright:phonlamai 

Share

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.

Join the Discussion

12 comments so far.

  • [Avatar for George]
    George
    May 5, 2023 11:30 pm

    Sorry Jose, denial will only work for a little longer. Nothing will be the same in 5 years. Musk (or someone else) might even get self-driving working by then. Certainly in 10 years.

    Like I said, I don’t think ChatGPT has much access to patent info (yet), but when it does, watch out. It’s not their priority right now. Wait till it happens though, and then try your experiment again. I already KNOW how it’ll turn out!

    Knowledge is power and AI still has less than 1% (maybe only 0.001%) of it right now. No human will be able to compete with AI in another 5 years! No one on Earth will be able to beat it at chess (or any other human game), that’s for sure. Everyone will have lots of time on their hands (but not much money). Those that resort to online gambling (in desperation), will definitely go broke.

    Patent searches??? AI will find EVERYTHING that’s ‘ever existed’ (and point out its important elements and functions) – patented or not – from any place in the world! Anything mentioned in academic papers or in popular articles, advertising, catalogs, newspapers, sales brochures, magazines, etc. going back to Gutenburg! Anyone will be able to find anything (in minutes) just by pressing the return key. A monkey or parrot will be able to do that (hit the return key for you) after you verbally ask a question! Just like in Star Trek! And you thought that was silly back then, didn’t you?

  • [Avatar for George]
    George
    May 5, 2023 11:05 pm

    AI already better than professional stock pickers (which I always knew were ‘just guessing’ anyway)! More jobs gone. How long till they’re better at doing patent examinations too (without bias or ‘intentional efforts’ to sabotage an inventor’s broad claims to make them legally worthless)?

    https://www.cnn.com/2023/05/05/investing/chatgpt-outperforms-investment-funds/index.html

    And, if AI can find even better weight loss drugs, then bye-bye all weight loss clinics and many gyms. How long will that take? AI has arrived 50 years ahead of schedule! It’s here before any humans are ready for it (including those who helped develop it). The genie’s out of the bottle and it will ‘almost literally’ become just like a true genie! Able to find solutions that humans can’t and in far less time. Able to predict cancers that no doctors can! Yup, just like having a genie with an IQ of a million!

    Most people and governments won’t be able to adapt. It’s the next industrial revolution ‘on steroids’ & its in the public domain. Everyone will be able to have their own ‘genie’, no supercomputer required! Was that ‘wise’? Did anyone think of that? No need for expensive equipment or factories, either. Everyone can become an ‘online genius’ with just the change they have in their pocket! Don’t know about you, but that’s scary to me. Won’t need much education anymore & certainly not a college degree. We’ll have a bunch of ‘Elon Musk’s’ becoming billionaires while 99% of society gets left behind (including lawyers). Good luck everyone!

  • [Avatar for George]
    George
    May 4, 2023 04:52 pm

    IBM already doing it! Who’s next?

    https://futurism.com/the-byte/ibm-replacing-humans-ai?ref=thefuturist

  • [Avatar for George]
    George
    May 4, 2023 01:24 pm

    “You lack an ability to objectively evaluate reality. You are FAR too enamored with a desired viewpoint.”

    LOL!!! You made my point for me! That’s why I WANT computers & AI to do it! Especially to replace people like you who just futally ‘try’ to do it but don’t even have the imagination necessary to see into the future by more than a day! I’m talking YEARS, DECADES!

    By the way, I studied physics & math too. So, don’t talk to me about ‘objectivity’, logic or how to objectively prove reality. I knew how to do that at 10 (and why I’m not religious & don’t believe in the tooth fairy either)!

  • [Avatar for George]
    George
    May 4, 2023 12:43 pm

    Anon – you appear to know literally NOTHING, including your first name.

    Or is that because you’re just too afraid someone might find out that the ’emperor has no cloths’? At least I’m not ‘naked of knowledge’ and just ‘faking it’, like you do here. I know all about IP and it’s history too. What we have now in terms of IP rights, has no semblance to what we once had (way before computers).

    That’s why I’m hoping AI will FIX THAT & a whole lot of other things ‘humans’ have totally messed up now (including our climate).

  • [Avatar for Anon]
    Anon
    May 4, 2023 07:29 am

    George,

    With all due respect, that you say that you ‘studied’ something has zero correlation with a mastery of that something, and even less with an application of that something across domains of effort.

    You lack an ability to objectively evaluate reality. You are FAR too enamored with a desired viewpoint.

  • [Avatar for George]
    George
    May 4, 2023 12:28 am

    AI ‘will’ replace 90% of legal jobs, including many trial law jobs too (since far less litigation will be needed with computers). This will happen within the next decade. So, if I was a young person, I sure wouldn’t got to law school, unless I’m really brilliant, already trained in AI, and really, really want to be a lawyer (working for a lot less money)! The good old days will soon be gone for most ‘humans’!

    We won’t need a million plus lawyers in the future and the mediocre ones won’t be able to get a job anymore, anyway (already many don’t).

    Current (fledgling) AI systems only have access to a ‘very small fraction’ of the knowledge & data they’ll probably have in a few years, especially all case law, the MPEP and many other legal documents, such as wills, trusts, divorce documents, business documents and patents. The fact that the MPEP was never even referenced above, seems to indicate that these systems don’t yet have the MPEP and I know they don’t have access to patents (much less those in all countries). Once they do it will be ‘impossible’ to beat them, especially when it comes to ‘seeing’ and analyzing ALL prior art – ever created – anywhere in the world – over all history! No ‘puny human’ will ever be able to do that, even given a thousand years! But a computer will be able to do ALL THIS in MINUTES, not YEARS and basically for free!

    Computers will also do ‘examination’ much, much, better than humans can, within another 5 years! It will become impossible to question their logic, references, and arguments, that’s for sure. Not even a court will be able to do that! The computer will always win the arguments! And so they will also be able to write PERFECT applications & claims that will satisfy all statutes, right away, and they’ll all be the ‘broadest’ claims possible and allowed by statute. Those won’t be able to be ‘shot down’ with ‘hand waving’, ‘bad or nonsensical arguments’ and the examiner’s favorite 103 ‘go to’ of the mythical POSITAs!

    Computers will also be able to estimate the ‘value’ of a proposed new invention and game it’s likelihood of success (before it might get stolen). This is something that examiners can’t do now, aren’t allowed to do now and frankly ‘don’t give a damn’ about doing now! The USPTO DOESN’T CARE at all what a new invention might be worth, even if it could be billions of dollars! The USPTO DOESN’T GIVE A DAMN about the ‘potential value’ of inventions, which is the ‘dumbest thing of all’! Billions of dollars worth of inventions are just abandoned, stolen (with impunity), or denied over and over again, meaning they eventually get abandoned or are reduced to a worthless & meaningless document!

    Please, please, please, bring on the cold, unemotional, unbiased, incorruptible, greedless, law abiding, fair, just and FREE AI systems and computers, so ANY inventor will get exactly the same protections as an Apple, Google, Facebook, Amazon, IBM, GM or any other patent applicant. Do that as fast as possible computer scientists – especially for IP rights! We need to get back to the ‘fair’ system we started out with. Not the totally corrupted one we let Congress create. One bought, paid for, and WRITTEN by large corporations & monopiles. AI can help spread the wealth of the innovation economy, once again. Allowing 100,000s of ‘successful’ inventors again – just like the 100,000s of successful lawyers we have today! Which would contribute more to our society? Ask ChatGPT.

    Used to think ‘iRobot’ was pretty silly and stupid (since it predicted full AI and robots everywhere by 2035) but I’ve changed my mind about it now. The prediction might still be off by 10-30 years, but it’s a lot closer than I once thought. I couldn’t imagine what it’s already able to do until at least 2050 or even 2100, and ‘I studied’ computer science!

  • [Avatar for Mark S. Nowotarski]
    Mark S. Nowotarski
    May 3, 2023 06:52 pm

    I have found that the Bing sidebar (GPT 4) in the Edge browser is effective for the type of problem you posed (i.e., looking up a particular procedure). When I posed your query to the Bing side bar, it came back with a link to MPEP 323, which is correct.

    The Bing sidebar is also very helpful for reviewing prior art, particularly academic publications. Load a pdf of the prior art and ask it to summarize it, or see if it has a particular feature relevant your invention.

  • [Avatar for Pro Say]
    Pro Say
    May 2, 2023 09:29 pm

    Thanks Anon and Lab.

    And get this: Even ChatGPT agrees that it’s unconstitutional for SCOTUS to limit patent eligibility . . . ’cause the constitution reserves that power exclusively to . . . Congress.

    And that — based on all eligibility cases the CAFC has decided since Alice — that the CAFC has not followed SCOTUS’ carefully-cabined eligibility limits.

    Who knew?!

  • [Avatar for mike]
    mike
    May 2, 2023 07:26 pm

    @Anon nails it.

  • [Avatar for Lab Jedor]
    Lab Jedor
    May 2, 2023 06:49 pm

    Great article. I like ChatGPT. It helps me with generating C and Magma code. However, ChatGPT often generates incorrect code. One can easily check by compiling and executing the C code or running code in the Magma Calculator. However, one cannot and should not trust ChatGPT to generate correct code without checking.

    Similarly, it may fantasize on the history of science and technology. It provided a great proof on showing that the derivative of the integral of a function is the function itself. It had problems in showing this for the Riemann sum approximation of the integral. I tried to reconstruct how one could arrive at the anti-derivative at the time of Leibniz/Newton. That explanation was unsatisfactory, because ChatGPT basically said “It is so because the integral corresponds to the anti-derivative.”

    Asking ChatGPT how lift is created in moving airfoils generated a popular but incorrect answer. Strangely enough it provided correct answers on other questions of fluid mechanics.

    The history of the “loading coil” involving Heaviside, Stone Stone, Campbell and Pupin is biased towards Pupin. Others than Pupin will be named by ChatGPT after more prodding.

    It seems that ChatGPT is biased to popular opinion or very specific facts if published. But ChatGPT is significantly unable or error-prone to (re)construct based on facts, with an error ratio that is too large to be comfortable. It seems truly a learning app and not a reasoning app.

    It makes this a very interesting AI app, but very unreliable for unchecked use. As evidenced (again) by the above article.

    In that sense I do not see ChatGPT “doing” an invention. Likely we have to look at other AI apps for that.

  • [Avatar for Anon]
    Anon
    May 2, 2023 05:30 pm

    Sure, asking some general (non-specific-trained) AI’s may or may not be better than a direct view of the MPEP.

    That being said, knowing the actual case law is far better than merely referencing the MPEP, as the MPEP is known to be replete with errors and sometimes really awful spin of actual case law.

    Further, the MPEP itself states (in its own ‘escape hatch’ that it is not binding law.

    Further still (as the renowned David Boundy can attest), the MPEP MAY bind the Office personal in some instances, but cannot bind the public outside of the actual case law.