Researchers are working to develop guidelines for the language used to describe intelligent systems, aiming to clarify communication among experts from diverse fields. The project will employ a mixed-method approach, combining expert opinions with advanced natural language processing capabilities of large language models (LLMs) like GPT-4-Turbo.
This collaborative effort involves multidisciplinary stakeholders and is facilitated by Cortical Labs, a company at the forefront of AI research. By leveraging LLMs to analyze existing definitions and identify commonalities and discrepancies, the project aims to establish a baseline for discussions among experts.
The modified Delphi method will be used to reach consensus on key terms, ensuring equal opportunity for contribution and collaboration. This work has far-reaching implications, as it seeks to create a comprehensive nomenclature guide applicable to various fields, including AI, machine learning, and cognitive science. The outcome is expected to ease scientific communication, promote clarity, and facilitate future research in the development of generally intelligent systems.
The use of large language models (LLMs) as a starting point to analyze existing definitions and identify commonalities and discrepancies is particularly innovative. By leveraging the advanced natural language processing capabilities of LLMs, the researchers can efficiently process a vast array of academic papers, discussions, and existing nomenclature to establish a baseline for discussions among multidisciplinary stakeholders.
The modified Delphi method, which involves an initial round of open-ended questions followed by subsequent rounds of refinement until a suitable level of consensus is achieved, is a well-established approach for reaching consensus in complex and contentious areas. The inclusion of preselected terminologies, asynchronous online format, and anonymity will help to facilitate collaboration, reduce bias, and ensure equal opportunity for contribution.
The use of qualitative methods to refine answers into key categories and the repeated consultation with collaborators until consensus is reached are also sound approaches. The role of LLMs as an intermediary tool to translate complex concepts across different disciplines, rephrase and contextualize viewpoints, and generate summaries and comparisons of different perspectives will be particularly valuable in facilitating a productive and less ambiguous dialogue.
The provision for a weighted majority voting system in the event that consensus is not reached on specific terms is also a pragmatic approach. While it may result in some terms that do not have full concordance from all collaborators, it is hoped that with a good-faith approach and fair consensus-making methods, the resulting nomenclature guide will be more useful than the current state of language usage.
The potential impact of this work is significant, as it could yield a nomenclature guide that is applicable to a wide range of fields, including artificial intelligence, cognitive science, neuroscience, philosophy, psychology, sociology, and computer science. The eventual outcome of this work could be a useful field guide for researchers exploring the intersection of these areas who are engaged in the development of diverse generally intelligent systems.
I would encourage all interested collaborators to register their interest and participate in this collaborative endeavor. By working together, we can develop a comprehensive and widely accepted nomenclature guide that will facilitate scientific communication, promote clarity, and advance our understanding of intelligent systems.
External Link: Click Here For More
