Minimal Sufficiency: A Principle ‘Similar’ to End-to-End – Communications of the ACM

Minimal Sufficiency: A Principle Similar to End-to-End

In a recent BLOG@CACM post, authors Mallik Tatipamula and Vinton G. Cerf call for the development of shared tools and interoperable infrastructure to enable the evolution of AI/ML, similar to the growth of the Internet.

They propose creating an AI/ML stack with a layer of standard services and applying a principle similar to the end-to-end principle that guided the Internet's development.

The author, who has spent 25 years studying the end-to-end principle, suggests that any AI/ML common services stack must go beyond networking to include resources like process management and storage, such as Unix/POSIX.

The goal is to enable the evolution of AI/ML to mimic the growth of the Internet.

The author's work aims to extend the success of the end-to-end principle to a broader class of services, including AI/ML.

Author's summary: Exploring principles for AI/ML evolution.

more

Communications of the ACM Communications of the ACM — 2025-10-31

More News