Abstract
In the late 1980s, the fields of Artificial Intelligence (AI) and Database (DB) research operated largely in isolation. In this seminal position paper, Michael Brodie argues that the next generation of computing—Future Intelligent Information Systems—requires a deep integration of these two disciplines. By combining the high-level reasoning capabilities of AI with the scale, persistence, and robustness of database management systems, the industry can move beyond simple data storage toward true interoperability among distributed, intelligent agents.
Introduction: The Integration Gap
Writing in 1988, Michael Brodie identified a critical bottleneck in the evolution of information systems. At the time, AI systems were capable of sophisticated reasoning but remained "brittle," often limited to small, specialized knowledge bases and lacking the infrastructure for real-world production. Conversely, Database systems were the masters of reliability and scale but were essentially "dumb," restricted to tabular data and unable to handle complex objects or autonomous decision-making.
Brodie observed a sobering reality: contemporary database technology only addressed approximately 0.5% of existing corporate and scientific data. To unlock the remaining 99.5%, a bridge was needed between the logic-heavy world of AI and the data-heavy world of Databases.
A Conceptual Framework for Integration
Rather than proposing a specific algorithm, Brodie outlines a visionary framework for how these two fields should converge. This framework is built upon three primary pillars:
1. The Object Paradigm
Brodie suggests that the "object" serves as the common denominator. By treating data and logic as unified objects, a shared language is created that allows AI, Databases, and Operating Systems to communicate seamlessly.
2. Three Levels of Integration
Integration cannot be superficial; it must occur across the entire system stack:
- The Knowledge Level: Focusing on the logic, meaning, and semantic intent of the data.
- The Symbol Level: Addressing the underlying data structures and representations.
- The Architectural Level: Defining how physical systems and hardware interact to support these processes.
3. From Interconnectivity to Interoperability
Brodie distinguishes between systems that simply "talk" to one another (interconnectivity) and systems that work together toward a common goal (interoperability). The future lies in intelligent cooperation where distributed agents understand each other’s objectives and constraints.
Key Principles for Intelligent Systems
To achieve this vision, Brodie introduces several foundational concepts that mirror the evolution of database theory but apply them to artificial intelligence:
- Knowledge Independence: Just as data independence allows data to exist apart from specific applications, Knowledge Independence allows rules and facts to be shared concurrently by multiple, unanticipated applications.
- Representation Independence: The underlying storage method (the representation) should be optimized for performance without requiring changes to the AI’s reasoning logic.
- Robustness and Scalability: For AI to be viable in enterprise environments, it must adopt the rigors of On-Line Transaction Processing (OLTP). This includes graceful failure handling and managing concurrent access to millions of facts, moving AI beyond "toy" knowledge bases.
The Agent-Based Future
The paper envisions a world populated by "heterogeneous, distributed agents." In this model, information is naturally decentralized, often incomplete, and occasionally inconsistent. Future systems must be composed of agents that can reason through these inconsistencies to provide reliable insights, effectively scaling AI to meet the demands of a massive, distributed information base.
Challenges and Future Directions
The primary obstacle to this integration is not merely technical, but cultural. Brodie notes a profound lack of understanding between the two research communities: AI researchers often overlook the complexities of secondary storage and persistence, while database researchers frequently ignore the nuances of task-specific logic and reasoning.
To move forward, Brodie suggests several research priorities:
- Intelligent Interfaces: Replacing static, unfriendly user interfaces with dynamic, AI-driven interactions.
- Functional Partitioning: Determining the optimal division of labor between a database and a reasoning engine.
- Incremental Augmentation: Developing architectures that allow knowledge bases to grow and evolve without being rebuilt from scratch for every new use case.
Conclusion
Brodie’s 1988 vision served as a roadmap for the decades of research that followed. By identifying the 0.5% metric, he highlighted the vast untapped potential of data that could only be reached through the marriage of AI and Database technologies. The transition from simple data repositories to intelligent, interoperable systems remains a cornerstone of modern information science.
Summary updated