Description

This report was inspired by two major developments in the global artificial intelligence ecosystem – one in generative artificial intelligence research, and another in the domain of international artificial intelligence safety.

The first development, was Anthropic’s release of an ambitious research paper in mid-2024.

Anthropic’s paper, “Scaling Monosemanticity: Extracting Interpretable Features from Claude 3,” addresses the topic of complexity of large language models (LLMs) and how the extraction of monosemantic neurons takes place. These neurons are actually designed in a way to respond to single, interpretable features, as opposed to polysemantic neurons that react to multiple, often unrelated, features.

The second development, was the International Scientific Report on the Safety of Advanced AI (interim report) released by the Government of United Kingdom and other stakeholders who were part of the AI Seoul Summit, May 2024, as a follow-up to the Bletchley Summit (on AI Safety) held in 2023.

This report focuses on these two developments, and examines Anthropic’s work on monosemanticity, with technical, economic & legal-ethical perspectives. The report also focuses on the evolution of neurosymbolic artificial intelligence and proposes some pre-regulatory ethical considerations that may be possible to be developed around this emerging class of AI technology.

About the Authors

Abhivardhan

Abhivardhan is the Managing Partner of Indic Pacific Legal Research & the Chairperson & Managing Trustee of the Indian Society of Artificial Intelligence and Law.

Samyak Deshpande

Samyak is a former Research Intern at the Indian Society of Artificial Intelligence and Law and pursuing undergraduate studies in law at the Maharashtra National Law University, Mumbai. He holds editorial positions in various publications in matters of technology policy.

Sanvi Zadoo

Sanvi is a former Research Intern at the Indian Society of Artificial Intelligence and Law and pursuing undergraduate studies in law at the Hidayatullah National Law University, Naya Raipur.

Alisha Garg

Alisha is a former Research Intern at the Indian Society of Artificial Intelligence and Law and pursuing the three-year LLB Programme at the Symbiosis Law School, Pune.