Artificial Intelligence-driven Innovation: Patent and Publication Analysis of ChatGPT-generated Ideas for Upper Extremity Prosthetics.

Abstract

<h4>Background</h4>ChatGPT, a large language model artificial intelligence (AI), has been used to augment patient-facing information, aid in resident education, and support research endeavors. The purpose of this study was to investigate an alternative use of AI by analyzing, through patent and publication data, whether ideas created by ChatGPT are novel and, therefore, useful to the innovative process.<h4>Methods</h4>ChatGPT was prompted with the statement: "Give me three ideas for innovation in hand prosthetics." These responses were used to complete a review of the literature and conduct a patent search. Searches for publications and patents were based on key words from the AI-generated text and Cooperative Patent Classification codes, respectively. Patent and publication number correlations for the last 10 years were calculated using a Pearson correlation for each group.<h4>Results</h4>Relevant patent numbers for each search (1: sensory feedback; 2: mind control; and 3: lightweight and durable materials) were as follows: 1: 47; 2: 17; and 3: 20. Publication numbers were as follows: 1: 69; 2: 69; and 3: 19. There were poor correlations between the number of patents and publications in all cases: <i>r</i> ₁: -0.05; <i>r</i> ₂: 0.55; and <i>r</i> ₃: 0.24 (<i>P</i> > 0.05 in all cases).<h4>Conclusions</h4>The data presented in this study show that ChatGPT was not able to generate innovative responses. However, when asked about specific technologies (ie, upper extremity prosthetics), it may have utility in generating ideas for technology refinement. As AI continues to advance, the novelty of its ideas will likely follow suit.