Functional AI Application Scoring (9-Point Profile)

Functional AI Classification: An Interactive Infographic

The Functional Turn in AI

Author - Dr. Sharad Maheshwari, imagingsimplified@gmail.com

A 2-Part System: Scoring AI with Core Categories & Context Modifiers

What is the Approach-Based AI Framework?

The ABAF is a 2-part classification system from the "Functional Turn in AI" paper. It moves beyond the simple "ML vs. DL" debate by describing AI in two layers: first, 5 Core Functional Categories (how it operates), and second, 4 Context Modifiers (what it requires and where it lives). This 9-point profile gives a rich, accurate description of modern AI systems.

Part 1: The 5 Core Functional Categories

1. Rule-Guided (RG)

Operates on explicit, human-defined rules, logic, and knowledge bases (e.g., traditional expert systems, symbolic AI).

2. Representation-Driven (RD)

Learns patterns and creates internal models from data (e.g., standard machine learning, deep learning, computer vision).

3. Hybrid Reasoning (HR)

Blends rule-based symbolic logic with data-driven models to achieve a result (e.g., neuro-symbolic AI, RAG).

4. Resource-Adaptive (RA)

Adjusts its operation based on available resources like compute, power, or bandwidth (e.g., on-device AI, federated learning).

5. Autonomous Learning (AL)

Modifies its own objectives, strategies, and learning processes over time without direct human input (e.g., advanced reinforcement learning).

Part 2: The 4 Context Modifiers

Next, the framework adds context by asking *what it requires* and *where it lives*. These four modifiers provide the crucial second layer of description.

Data Dependency

Describes how much and what kind of data the AI needs to function *at runtime* (e.g., none, a single input, or a constant live stream).

Learning Mode

Defines how the model was "taught" or how it improves (e.g., fixed/no learning, batch-trained, or continuous/reinforcement learning).

Knowledge Source

Identifies where the AI's "knowledge" comes from (e.g., purely human-defined rules, purely data-driven, or a mix of both).

Deployment Environment

Specifies where the AI runs (e.g., highly constrained on-device, a local server, or a massive, distributed cloud).

Live AI Application Scoring (9-Point Profile)

Select an application to see its complete 9-point functional profile. A score of 5 represents a high dependency on that approach, while 1 is very low.

Comments