Foundation Model Platforms and Bottom-of-the-Pyramid Innovation

Abstract

We analyze whether artificial intelligence (AI) foundation model platforms meet the twelve principles of bottom-of-the-pyramid innovation. Perhaps surprisingly, we find that they do satisfy several of the principles. On many of the ones they do not, active AI research is underway. Some platform functionality beyond the actual foundation model and education in the language of low-resourced users is still missing. This analysis is supported by a case study of ChatGPT to perform semantic mapping of sentences in a social good context. A call to action for grantmaking institutions is provided.