Back to Study Material Artificial Intelligence

Strong AI and weak AI

The distinction between strong AI and weak AI describes not just what a system can do, but what kind of intelligence it possesses. Understanding this diffe...

Strong AI and weak AI

Two Visions of Artificial Intelligence

The distinction between strong AI and weak AI describes not just what a system can do, but what kind of intelligence it possesses. Understanding this difference is fundamental to both philosophy of mind and AI research strategy.

Weak AI, also called narrow AI or applied AI, refers to systems designed to perform specific tasks extremely well without possessing general understanding. A chess engine beats grandmasters but cannot hold a conversation. A spam filter classifies email but cannot recognize faces. Every AI product in commercial use today falls into this category. These systems optimize a narrow objective and have no awareness of anything outside their designated domain.

Strong AI, sometimes called artificial general intelligence (AGI), describes a hypothetical system that can understand, learn, and apply knowledge across any intellectual task a human can perform. A strong AI system would seamlessly switch from composing music, to debugging code, to diagnosing illness, to planning a trip — just as a human expert with broad knowledge could. It would also possess self-awareness and consciousness, according to the strictest definitions.

A concrete example illustrating the gap: GPT-class language models can write legal documents and explain quantum physics in plain language. Yet if you ask such a model to remember that you changed your name at the start of the conversation and it exceeds its context window, it forgets entirely — a trivial task for any human. This fragility shows it remains firmly in the weak AI category.

Some researchers add a middle category — artificial narrow general intelligence — for systems that are general within a broad but bounded domain, such as robotics systems that adapt to new objects without reprogramming.

Exam tip: Examiners often ask students to classify current AI systems. All production systems today are weak AI. Be prepared to explain why, using examples like chess engines or image classifiers.