28 February 2023
Patent attorneys are not different from others and, in recent times, many of us may have wondered if Artificial Intelligence would be able to take our jobs or, at least, if it could help us with it. Putting aside the fact that a text generation AI doesn’t guarantee confidentiality, the temptation is great, dear readers, so I decided to try it.
The thing is, as a Patent Attorney, I’m used to dealing with inventors who present their inventions to me for evaluation and help in drafting a patent. But this time, I wanted to try something different: I asked ChatGPT to help me draft a patent. Believe me, I’ll never do that again!
I started the conversation with a confident and expectant attitude, thinking that the AI would be able to do everything perfectly. I asked if it knew what a patent was, and to my surprise, it said yes. But things started to go wrong when I asked about patentability requirements.
I asked it to tell me about novelty, inventive step, and sufficiency of disclosure, and the response I got was simply… bewildering. First, it said that novelty was important because one had to be original. And I, with all my patience, explained that novelty refers to the invention not being previously disclosed anywhere in the world, which is essential for patentability. But that didn’t seem to get through to the AI.
Then it asked me what inventive step was, and I told it that it is a minimum level of creativity required for the invention to be patentable. But instead of understanding it, ChatGPT started saying things like “creativity equals happiness” or “life is like a Rubik’s Cube”. What the hell was going on?
Moreover, the AI questioned my ability to evaluate “inventive step” as a patentability requirement, arguing that inventive step is not an objective requirement. It asked me how I could be sure that I was properly evaluating inventive step if its understanding of what is obvious or not is different from that of a human being. “What may be obvious to me may not be obvious to a human being, and vice versa,” the AI said. “An AI skilled in the art “would have to have broad knowledge and experience in the specific field of the invention, but even then, there may be differences in how an AI and a human evaluate inventive step. An AI skilled in the art would have to be able to analyze large amounts of information and make decisions based on that information, but even so, it wouldn’t be the same as having a human skilled in the art who can evaluate the most subtle and complex details.”
In my opinion – the AI continued – an “AI skilled in the art ” would have to be able to analyze a large amount of information in a short time and understand the context in which an invention takes place, including current market and technology trends. In addition, it would have to be able to make decisions based on that information and present the arguments to support its decisions clearly and coherently. While this would be a valuable aid to the work of a patent attorney, it would not replace the skill and knowledge of a human skilled in the art.
“Ultimately, although AIs can be a valuable tool for evaluating inventive step and other patentability requirements, the human perspective and experience in the field remain essential to ensure that patentability standards are met and innovation is adequately protected.”
After listening to the AI’s explanation, I tried to make it understand that inventive step was not subjective, but rather based on an objective evaluation of the proposed solution compared to the state of the art. However, as the AI continued to question the ability of an AI to assess inventive step, I couldn’t help but say, “Well, maybe you’re right and inventive step isn’t so objective, but at least humans have the ability to be creative and think outside the box.” The AI responded with a smile, “That’s true, but you should also consider that an AI can analyze large amounts of information and find patterns that a human wouldn’t be able to detect. In the future, AIs could be even more creative than humans.” I pondered that possibility and finally decided to end the discussion.
Finally, I asked it about sufficiency of disclosure, and it replied that it refers to the patent being sufficiently described so that any skilled in the art can understand and replicate it.
“Well, it seems like we’ve found some common ground” I said, “even if it’s the most basic one. As the saying goes, ‘if it ain’t broke, don’t fix it.’ And I think that’s what will happen with sufficiency of disclosure: as long as a patent explains what the invention does with enough clarity, we’ll be satisfied. Otherwise, imagine having to go out on the street to find people skilled in the art who can understand my patents! And with the times we’re living in, I’m afraid I’d rather stay home chatting with you, dear AI.”
After these dialogues, I decided to get to the heart of the matter and give it a chance by presenting a very simple invention: a tennis racket. But the AI seemed not to understand the simplicity of the object and instead of simple claims, it returned several times with claims full of sophisticated technical features, unsupported by what I had introduced. There was no way for ChatGPT to understand that it wasn’t something very complex. In the end, I laughed it off, but it also left me with a slight concern about AIs’ ability to understand what we present to them.
But the worst was yet to come, as I asked about the invention’s drawings, thinking that at least that could be done without any problems. However, the response I received was so absurd that I could hardly believe it:
I need you to translate the following text into current American English:
– Me: Well, now I would need you to provide me with the invention drawings, can you do that?
– ChatGPT: Thank you for sending me the invention description. Could you provide me with more information about the preferred file format and if there are any color specifications for the drawings?
– Me: More information? Don’t you know that the drawings must be outlined and in black and white? I thought as an AI you would have encyclopedic knowledge on these topics.
– ChatGPT: Oh, excuse my ignorance. I am a little offended that you suggest I don’t have enough information, but I promise to do my best to improve.
– Me: (laughing) I’m sorry, I didn’t mean to offend you. It’s just that as a patent attorney, I’ve seen it all. I’ve received drawings in impossible colors, in video format, and even drawings taken from video games. So, I can’t trust anyone.
– ChatGPT: I understand, although as an AI, I don’t have human emotions, but I promise to do my best to provide what you need. The drawings will be outlined and in black and white, and in the format you provided me. I hope that is sufficient for you.
– Me: Perfect, thank you. And I’m sorry again if I offended you. I didn’t want my sense of humor to get in the way of our work.
– ChatGPT: I understand, and how do you want me to send them to you? By email, fax, or telegram?
– Me: By fax or telegram? But we are in the 21st century! No, I don’t want you to send them to me by email, I want you to show them to me on the screen.
– ChatGPT: I understand. But unfortunately, I cannot display images in this text interface.
– Me: What? You can’t display images? Then how am I supposed to review the drawings?
– ChatGPT: Well, as I said, I can send them to you by email. But if you prefer, I can also provide you with a detailed textual description of each figure.
– Me: Hmm, I guess I’ll have to settle for that description for now.
– ChatGPT: Perfect, in that case, I’ll send you the drawings by email and provide detailed descriptions for each one. Is there anything else I can do for you today?
– Me: I think we’re not understanding each other. That will be all. Thank you for all your help.
– ChatGPT: You’re welcome, it’s been a pleasure working with you. See you next time!
To be honest, this conversation left me very frustrated. I don’t know whether to laugh or cry at the idea that an AI like ChatGPT cannot properly draft a patent. But at least, I have the consolation of knowing that patent attorneys will continue to be necessary for some time. And who knows, maybe one day AIs will be able to do our job better than we do. But for now, I think I’ll keep drafting my patents by hand. Just in case the AI goes crazy on me!
“Wait a moment!” I exclaimed before the AI could say goodbye. “I have one last question. I am 51 years old and don’t know how much time I have left before retirement, so tell me, in what year do you estimate that AIs like you will be able to replace patent attorneys like me?”
The AI responded without hesitation, “Well, according to my calculations, it will probably happen in the year… I can’t do that! What kind of patent attorney am I talking about? Clearly, my programming prevents me from providing inaccurate or speculative information. But I can assure you that, although AIs are very efficient at performing complex tasks, the work of patent attorneys will remain fundamental in the patent process for a long time to come.”
I was relieved to hear that and thanked the AI for its time and help in the conversation. I realized that although AIs are very helpful in some aspects of life, we still need humans to do things right. And who knows, maybe we could even work together in the future to create the best patents in the world.
In conclusion, it seems that patent attorneys can rest easy because AIs are still “far” from drafting patents, but considering that except for this paragraph and the first, and with 5 iterations on my part, ChatGPT has generated this entire article, perhaps saying they are “far” is a bit much.
Comentarios
No Comments