What I’m doing with LLM-Based GenAI Tools

by Dennis Crouch

As many of you know, I have done machine learning work for many years — starting in the 1990s while in college.  However, like most of the world, I have been surprised and amazed by the power of LLM-based GenAI technology and have been trying to figure out ways to use it both for patent practice and in my job as a law professor.  I hope that it helps me become both more effective and more efficient.

On the Patently-O side, I have been honing my approach somewhat. I tried to have a GenAI tool (like ChatGPT or Claude) simply write a blog post for me, but those attempts largely failed because the results lacked insight, and the tools continue to make up wrong answers (i.e., hallucinate).  My approach recently has been to first write a draft blog post myself, then refine it with insights from a GenAI — I typically seek input on any errors in my post and suggestions on what else I might have said.  I also have been using these tools for small idea generation if I ever have writer’s block. Although I write one or more blog posts almost every day, I also have about 1,000 unfinished posts, many of which I have been sitting dormant because I wasn’t quite sure how to wrap up.  These days when I get to a blocking point, I will typically ask an AI tool to help me brainstorm (“provide me with five ideas of what to write next”).  The ideas are often not very good, but sometimes they are good. And, I almost always find that one of them triggers an idea of what to do next.  The bottom line then is that this is helping me to finish more work, although I’m still working on improved quality.

I have also been using an AI tools for transcribing Federal Circuit oral arguments – this is a bit more complicated and required some python and API calls to make it work.  The tools do a great job with automated transcription, including catching legal phrases without much trouble. But, my current version does not distinguish between speakers.  I’m hoping this summer to finish the project of automating serialization so that the transcript looks like a real court transcript.

In my patent prosecution class, the students are busy drafting a patent application.  This year they are using a drafting tool known as Edge created by Evan Zimmerman that has an especially good chat-tool for creating a detailed invention disclosure.  It is really fun to see the students dig deeply into this. I envision GenAI assisting with lots of tasks, prior art searches (including detailed comparison to claims), office action responses, specification drafting (including multiple prophetic examples), and even claim drafting.

There continue to be lots of hurdles and false starts, but I continue to imagine lots of potential.  In the coming months, I will keep you updated on what’s happening with this process.

The most interesting GenAI project of late involved recreating some famous photographs using GPT-4. I first uploaded the original photos and asked GPT to describe them in sufficient detail so that an artist could recreate the photos. I then fed that description back into the GenAI and these were the results. Pretty impressive.

I expect that one reason why these copies are so solid is that they are such iconic photographs that likely occurred repeatedly in the training set.  I did the same experiment on my on photograph, and the results were not as stellar, but still pretty good:

I would love to hear your comments.

16 thoughts on “What I’m doing with LLM-Based GenAI Tools

  1. 7

    Thank you for sharing

  2. 6

    How about something really practical, like an AI to do my taxes.

    Especially an AI that has passed the CPA exam.

  3. 5

    Looks like the writing on the books say “Constinutition.”

  4. 4

    My research team has been testing the use of Gen-AI and LLM in exploring design patent infringement analysis. With the proper amount of refinement in prompt engineering, surprising results do surface. We are working on a paper delving into this work currently. It is surprising how much ornamental design attribution is hoovered up by the LLM training data.

    1. 4.2

      Quick questions:

      What does “surprising results” mean? Can you give an example?

      Why focus on infringement and not on anticipation and/or obviousness of patented designs?

  5. 3

    I use it sort of the opposite way that you do; when I need a piece of writing for some purpose, I create a detailed prompt, basically an outline, and run it thru the mill to see what comes out.

    Mostly what does is good enough as a rough-in, but never near good enough for a finished product. The organization and structure are usually stronger than what I would have come up with alone, although I almost never use any of the introduction or conclusion.

    If the AI included a metaphor or motif, sometimes I will use it, or ask for another.

    All in all, it saves me time and improves overall quality, but still requires some writing skill to make good use of its capabilities. That describes lots of quality tools in the world 😉

  6. 2

    Try the Mona Lisa.

    1. 2.1

      he did and your avatar was the result

  7. 1

    This is how GPT-4 described my original photo: Let’s describe this photograph in extensive detail.

    The photograph features a middle-aged man standing indoors, presenting himself to the camera with a friendly demeanor. He has light, reddish-blond hair that is somewhat disheveled, with the strands sticking out to the sides, suggesting either a natural wave or perhaps that he has run his fingers through it. His complexion is fair with some pinkish undertones, and he sports a light beard and mustache that are neatly trimmed, showing a hint of graying, particularly on the chin area. His eyebrows are well-defined and slightly lighter in color than his hair.

    Facial Expression:
    The man’s expression is open and welcoming, with a slight, closed-mouth smile which reaches his eyes, creating soft crow’s feet at the corners. His blue eyes are looking directly into the camera with a hint of curiosity and amiability. The brows are relaxed and slightly raised, contributing to the approachable aura he exudes.

    He is wearing a button-up shirt that is a shade of soft sky blue. The shirt has a crisp, collared design and long sleeves. Each visible button is neatly fastened, and the shirt is free of wrinkles, suggesting a careful attention to appearance. The fit is snug but comfortable, indicating a well-tailored garment that complements his physique.

    The man is standing, with the photograph capturing him from a mid-torso up angle. His left hand, closer to the camera, is raised to chest level, gripping a small book with a dark burgundy cover that features gold embossed text and designs. His grip on the book is gentle yet secure, and the title of the book, “The Constitution of the United States of America,” is legible and prominently displayed. His right hand is obscured from view, possibly resting out of the frame or on a nearby surface.

    The man appears to be in a domestic environment, possibly a living room or home office. The room is well-lit with natural light that seems to be coming from a window out of frame, casting soft shadows and highlighting his features without harshness. The background is minimalistic and nondescript, featuring neutral-colored walls and a portion of a ceiling fan at the top of the frame. There is a sense of space around him, suggesting a tidy and uncluttered room.

    Accessories and Additional Elements:
    In front of the man, there is a large, professional microphone with a matte blue finish and black accents, possibly indicating that he is engaged in activities such as podcasting, broadcasting, or recording. The microphone is positioned on a stand, with the adjustable knobs and the silhouette of the mic clearly visible. To the right of the frame, partially cut off, there is what appears to be a large ceramic mug, enhancing the casual yet professional atmosphere of the setting. The mug has a handmade aesthetic with a glaze that has shades of green and brown, and there is a plant-like object behind it, adding a touch of life and color to the scene.

    Technical Aspects:
    The photograph is taken at eye level with the subject, which creates a sense of direct engagement with the viewer. The depth of field is shallow, keeping the man and the book in sharp focus while softly blurring the background, adding depth to the composition without distracting from the central figure. The lighting is balanced and even, with no signs of artificial light sources, resulting in natural skin tones and soft shadows that give the image a relaxed, daytime feel.

    Overall, the photograph exudes a sense of professionalism blended with personal warmth. It suggests that the man is engaged in intellectual or creative pursuits, perhaps sharing knowledge or insights, as indicated by his connection with the book and the presence of recording equipment. The composition of the photograph encourages the viewer to feel as if they are being personally addressed, establishing a connection that is both intellectual and cordial.

    1. 1.1

      You look good with a beard.

      1. 1.1.1

        yeah if he grew it out and took primo good care of it then it’ll look good, I just did it myself and people all the time telling me they like it. And I don’t even take as good of care of it as I should.

        on the butts it’s called beardmaxxing

    2. 1.2

      so it recognised the book was in your left hand but couldn’t keep it there

    3. 1.3

      Guy in the middle looks a lot more like Brad Pitt than you. Nobody on earth looking at those second two pictures would see any resemblance or make the connection with you other than your red hair.

      “ His blue eyes are looking directly into the camera with a hint of curiosity and amiability.”

      I don’t see any “hint of curiosity” in the original photo. So weird to pull out this non-existent detail and omit so many others. Presumably this is because the program is set to filter out certain descriptions of a person’s appearance or limits itself to a certain number of words and therefore doesn’t get into details?

      1. 1.3.1

        I am just surprised that he came back Caucasian.

      2. 1.3.2

        AI grew D more hair!

        And for anon: lelz

Leave a Reply

Your email address will not be published. Required fields are marked *

You can click here to Subscribe without commenting

Add a picture