AI and Innovation

Traditionally, inventors came up with ideas, built prototypes, and iterated on the design before arriving at a practical solution to a unique problem. They would seek help from others to iterate on their design and finally file for a patent. The patent would mention all those involved in the invention.

As AI infiltrates every aspect of our lives, it will be interesting to see how the process of invention changes. Many generative AI LLMs and tools are trained on massive datasets of publicly available data. These AI tools are untethered from a specific field and could draw inspiration from different fields to develop innovations. You could learn from particle physics and apply it to genetics, or you could look for mathematical patterns in art. Even hallucinations (or generative fabrications) from LLMs could inadvertently lead to innovations. The potential for human innovation is poised to explode. Never before have studies in philosophy, genetics, and architecture combined to solve an epidemiological problem. It was quite impractical to employ experts from such diverse fields to solve difficult problems. AI gives us the opportunity to put the collective intelligence of the entire human race behind solving problems hitherto thought unsolvable.

So when I put my new idea into Gemini or Claude to see if it is a good idea, does the idea still remain mine? If the AI tool uses its vast learning to provide me with additional insights and potential add-on ideas, does it remain my idea? Can I still file a patent? While I might not take the ideas suggested by AI, once I share my idea, will it find its way to some other inventor?

How should the boundary between plagiarism and originality be drawn? What guardrails should be put in place to prevent idea theft?

There is obviously active work going on in this field. The safest path is to ensure you can prove you “framed” the problem yourself and fed it to AI with prompts. You need to use AI to do the work, but the idea needs to be yours, and you need to show your prompts and your conversation with the tool to prove it. The conversations will be important to keep, as they will demonstrate that you sifted through the AI-generated output and further refined the ideas on your own. Even with all of this, there are potential issues related to “unintentional” infringement. It is always advisable to check the sources that the AI response quotes to ensure you are not accidentally overlapping with someone else’s work. It could also recreate something that already exists. One needs to be careful in navigating through this.

Despite all the work underway in this field, many blind spots will emerge in the future. How will we treat a solution entirely crafted by AI? What if the original idea was mine, but AI suggested a better solution?

Read More at Substack