AI and the Doctor Dolittle challenge

Area of Science:

• Artificial Intelligence
• Animal Communication
• Bioacoustics

Background:

• Human desire to communicate with animals is ancient.
• Advancements in AI, particularly Large Language Models (LLMs), suggest potential for interspecies communication.
• Scientific consortia and commercial entities are actively pursuing this goal.

Purpose of the Study:

• To frame the challenge of interspecies communication as the 'Doctor Dolittle challenge'.
• To identify and analyze the primary obstacles hindering successful animal communication using AI.
• To define criteria for achieving success in the 'Doctor Dolittle challenge' and encourage further research.

Main Methods:

• Conceptual analysis of AI capabilities in generating and interpreting animal communication.
• Identification of challenges related to context determination, unsupervised learning validation, and the scope of animal communication.
• Definition of success criteria for AI-driven interspecies communication.

Main Results:

• Three main obstacles identified: 1) Difficulty in determining context and human bias (umwelt) in AI-generated communication, 2) Need for experimental validation of AI-extracted context to avoid spurious correlations, 3) Limited scope of natural animal communication (e.g., alarm, courtship).
• Generative AI can produce novel animal communication samples, but contextual interpretation remains a significant hurdle.
• AI's tendency to find spurious correlations necessitates rigorous experimental validation.

Conclusions:

• Achieving human-animal communication via AI faces significant challenges, including contextual interpretation, validation, and the inherent limitations of animal communication systems.
• Despite obstacles, AI offers powerful tools for deciphering and potentially mimicking animal communication.
• Further scientific endeavor is crucial to meet the criteria for the 'Doctor Dolittle challenge'.