AI is an amazing tool. A few tips based on my experience in applying it to product development:
Ideation: Give it a problem and ask for innovative physical products to address the problem. It will give you some good avenues to ponder/pursue. I tried once to get it to actually “invent” a product, but in general I don’t think it’s ready for prime time yet. In my experience, the closer you try to get to finite solutions w/AI, the less efficient it becomes.
Prior Art Search: ChatGPT is great for organic prior art searches, not so much for patent searches (though this may be more of a user shortcoming). It can provide keywords for searches and can provide awesome boolean search strings for searching the USPTO site, I just haven’t been able to get it to whittle down the searches to a reasonable number of pertinent results. I find it interesting that I can do this manually but have not yet figured out the prompts for AI to complete the task well. Tip of the day: try https://projectpq.ai, an open source AI patent search engine. Note that it searches google patents, not USPTO, and cannot be considered as a comprehensive search tool. I use the results from pqai to help draft search inquiries for classification searches on the USPTO site. Another GREAT thing chatGPT can do is analyze claims of patents and tell you what is actually being claimed in laymen’s terms. This alone has saved me hours of work.
Prototyping: It’s written some amazingly elegant code for Arduino on the first pass (amazing how well it can write code while often failing at the most basic math). It can be used to provide some innovative paths to pursue when trying to achieve a specific function in a prototype. It’s can also quickly provide specific technical/operational information needed to produce prototypes, as well as information about materials, manufacturability, and DFM, providing the education to make your prototypes closer to the end product. I’ve found chatGPT-4’s graphics capabilities to be frustrating and completely useless for my purposes. I’ve been told that Midjourney is much better but I haven’t tried it.
Patent Writing: I have written a GPT that does an impressive job of writing the front end of PPA’s (Abstract, Field, Background, Summary). I like to include a couple of broad claims in my PPA’s, and I’m tweaking a GPT now to draft these. I’m also working on a GPT to identify possible embodiments of the invention, which is an important part of a PPA.
In short, the only task I’ve found in the invention process that AI can complete on its own is coding. While it can open some mental doors and make many other tasks more efficient, it still requires intervention/knowledgable oversight. Hope this information is helpful to someone.
Ideation: Give it a problem and ask for innovative physical products to address the problem. It will give you some good avenues to ponder/pursue. I tried once to get it to actually “invent” a product, but in general I don’t think it’s ready for prime time yet. In my experience, the closer you try to get to finite solutions w/AI, the less efficient it becomes.
Prior Art Search: ChatGPT is great for organic prior art searches, not so much for patent searches (though this may be more of a user shortcoming). It can provide keywords for searches and can provide awesome boolean search strings for searching the USPTO site, I just haven’t been able to get it to whittle down the searches to a reasonable number of pertinent results. I find it interesting that I can do this manually but have not yet figured out the prompts for AI to complete the task well. Tip of the day: try https://projectpq.ai, an open source AI patent search engine. Note that it searches google patents, not USPTO, and cannot be considered as a comprehensive search tool. I use the results from pqai to help draft search inquiries for classification searches on the USPTO site. Another GREAT thing chatGPT can do is analyze claims of patents and tell you what is actually being claimed in laymen’s terms. This alone has saved me hours of work.
Prototyping: It’s written some amazingly elegant code for Arduino on the first pass (amazing how well it can write code while often failing at the most basic math). It can be used to provide some innovative paths to pursue when trying to achieve a specific function in a prototype. It’s can also quickly provide specific technical/operational information needed to produce prototypes, as well as information about materials, manufacturability, and DFM, providing the education to make your prototypes closer to the end product. I’ve found chatGPT-4’s graphics capabilities to be frustrating and completely useless for my purposes. I’ve been told that Midjourney is much better but I haven’t tried it.
Patent Writing: I have written a GPT that does an impressive job of writing the front end of PPA’s (Abstract, Field, Background, Summary). I like to include a couple of broad claims in my PPA’s, and I’m tweaking a GPT now to draft these. I’m also working on a GPT to identify possible embodiments of the invention, which is an important part of a PPA.
In short, the only task I’ve found in the invention process that AI can complete on its own is coding. While it can open some mental doors and make many other tasks more efficient, it still requires intervention/knowledgable oversight. Hope this information is helpful to someone.
Dislike ads? Remove them and support the forum:
Subscribe to Fastlane Insiders.