Following up on the previous post about the list of companies producing the most highly cited research and patents in AI (linked above), it's worth highlighting a recent study by Stanford University researchers on the capabilities of generative AI models in generating novel research ideas.
In the study, over 100 NLP researchers were recruited to generate novel ideas, with blind reviews comparing ideas from both large language models (LLMs) and human experts. The findings showed that LLM-generated ideas were judged to be more novel than those from human experts, though slightly weaker in terms of feasibility https://arxiv.org/abs/2409.04109
In early August of this year, an article published in the well-known The Economist reported that generative AI models are becoming smarter. This observation allows us to anticipate that this evolution will continue in the coming years, such that the future capabilities of these models will far exceed those we are familiar with today. The limit of this evolution was discussed several years ago by the German scientist Jürgen Schmidhuber, a Scopus Highly Cited Scientist (h-index = 78) https://arxiv.org/pdf/cs/0606081
Declaration of Competing Interests - I am concerned about the reliance on Google Scholar as a primary tool for selecting human experts, Section 4.2, 'Expert Qualifications.' A recent study conducted by researchers at New York University has highlighted significant vulnerabilities in Google Scholar’s database, indicating that it is vulnerable to manipulation https://19-pacheco-torgal-19.blogspot.com/2024/05/paper-how-to-exploit-chatgpt-for-large.html
