AI Science and Engineering

Longbing CaoIEEE Intell. Syst 2022 · OpenReview

Introduction

In a recent editorial for IEEE Intelligent Systems, Editor-in-Chief Longbing Cao argues for the formalization of "AI Science and Engineering" (AISE) as a distinct scientific discipline. This proposal marks a strategic shift for the field, moving toward an "X-intelligence" paradigm that integrates human, artificial, natural, and organizational intelligences to address the multifaceted challenges of the modern world.

The Evolution of AI: From Algorithms to AISE

Artificial Intelligence has undergone significant evolution over the last 70 years. However, the field has historically lacked a unified framework capable of bridging the gap between theoretical research and robust engineering. Cao identifies a critical transition point around the year 2000, marking the shift from traditional AI to a new era of "X-AI."

This new era is defined by "X-complexities"—the intricate intersections of cyber-physical-social systems and metaverses. To navigate these complexities, the field must move beyond isolated algorithms and toward a comprehensive AISE body of knowledge that treats AI as a rigorous engineering discipline.

Conceptual Frameworks for a New Era

The editorial outlines several conceptual pillars intended to standardize and advance the discipline:

  • The AISE Framework: A proposed structure for education and professional development that shifts the focus from simple task-solving to the engineering of complex systems.
  • Hanoi Intelligence: Introduced by contributor Fei-Yue Wang, this concept connects Human, Artificial, Natural, and Organizational Intelligence. It is designed to solve problems within robotics and the metaverse by leveraging "parallel intelligence."
  • Historical and Formal Analysis: By reviewing 70 years of publications from major venues like IJCAI and AAAI, Cao categorizes AI history into two distinct ages. To support this, he proposes new mathematical formulas to define AI systems, goals, and processes, providing a standardized language for the field.

Trustworthiness and Real-World Application

In the AISE paradigm, "Trustworthy Machine Learning" is not an optional feature but a fundamental architectural requirement. This includes inherent security, fairness, privacy preservation, and integrity, particularly for AI systems deployed in cloud environments.

The maturity of this approach is demonstrated through practical applications highlighted in the journal, including:

  • Mechanism Design: Utilizing the Vickrey–Clarke–Groves (VCG) mechanism to solve complex economic and resource allocation problems.
  • Robotics and Localization: Advancements in 6-D robot localization within GPS-denied environments.
  • High-Performance Training: AI-enabled systems used by gold medalists in the 2022 Beijing Winter Olympics and the development of "robot chefs."

Future Directions

Despite recent progress, a significant gap remains between current AI capabilities and the nuanced complexities of biological and social systems. The future of AISE involves:

  1. Shrinking the Gap: Developing AI that more closely mirrors the sophistication of natural and social intelligence.
  2. Standardization: Continuing to build the AISE body of knowledge to support the next generation of professionals.
  3. Emerging Technology Integration: Aligning fundamental AI research with the demands of Web 3.0, 5G/6G, and Industry 4.0.

By formalizing AI Science and Engineering, the research community can better facilitate the transition into a digital economy driven by secure, fair, and integrated intelligence.

Summary updated

BibTeX
@inproceedings{cao2022ai,
  author = {Longbing Cao},
  title = {{AI Science and Engineering}},
  booktitle = {IEEE Intell. Syst},
  year = {2022},
  url = {https://openreview.net/forum?id=A7Y4LekLQa},
}
Figure