Many articles have been written about the interplay between AI and patents, for example –
- Are AI-based inventions patentable? The answer, generally speaking, is yes.
- If you have invented a new AI architecture, should you patent it? I tried to answer that question in an article that you can find here.
- If your AI program made an invention of its own, can you patent it and name your program as the inventor? Based on recent court cases in the US and other countries, the answer is a clear no. But if you make an invention with the help of AI, you can still patent it and list yourself as the inventor.
There is another question that has gotten less attention: When will you be able to get ChatGPT to write your patent application for you instead of having to work with a patent attorney? When that happens, will patent attorneys find themselves unemployed?
I say “when that happens,” because the day will come. With LLMs (large language models) becoming more powerful every month, and new LLM applications developing on all fronts, it is only a matter of time until most legal writing tasks are taken over by machines. Practical LLM tools already exist for jobs like writing contracts. It is just a matter of time until someone comes out with a product capable of doing most of what your patent attorney does for you now.
In fact, a number of companies already offer LLM-based patent writing software packages. I have tried a few of them and have not found one that actually made my job easier. There are particular challenges in writing patent applications that LLMs are not yet capable of handling well. One of them is preparing the patent drawings and integrating them with the text. Another is claim writing – An LLM can master the craft of writing claims in the proper style and syntax, probably better than most junior patent attorneys, but the art of drafting claims that bring out the subtle technological point of the invention remains a human endeavor.
When I write a patent application, I generally start by sketching out the drawings and writing what I consider to be the key claims. Once I have done so, I feel that I have a grasp of the invention and know what details need to be filled in. There is still a substantial amount of work left to do to complete the patent application, but it does not require much creativity. That is the part of the job that I would like to pass to an LLM. I am waiting for a product that will let me input some claims, scan in my drawings, and come back to me with the complete patent application.
After that, what is next? I can imagine that in the space of a few years, we will start to see interactive AI-based software that interviews the inventor, learns the invention, writes a patent application, and then files it online in the patent office. It may not be a patent application of the highest quality, but it is no secret that even today, most human-written patent applications are no better than mediocre. Patent firms like ours, who pride themselves on high-quality patent drafting, will not be happy, but my guess is that in-house patent departments will be thrilled to see their costs reduced and perhaps will call on outside counsel only for strategic, high-value patents.
What about the other side of the equation – the government patent offices? Patent examiners already use AI tools for patent searching, and the examination reports that they issue are generally formulaic documents made up mostly of cut-and-pasted text. AI should be able to do most of the examiners’ work, and will probably do it better than what most examiners are turning out now.
Patent practice has changed very little in the 28 years I have been doing it. Technological tools have made it easier for us to communicate and produce the documents that we file, but the jobs of creating patent applications and navigating them through the examination process have remained essentially the same. That may no longer be the case ten years from now. In the not-too-distant future, our clients may be using AI machines to write and file their own patent applications, which will then be examined by other AI machines in the patent office. Patent attorneys should be thinking about their place in that future.