Structuralism / Functionalism Tension in AI/AGI
Details
Does AI need to have a cognitive structure like the human mind, or does internal structure not matter?
The smashing success of AI is due chiefly to the transformer, yet it bears scant structural similarity to the mind. The transformer seems like sheer functionalism.
Some contend that a machine can never truly understand reality unless it can operate like our minds -- that is, exhibit a structural affinity to create and utilize determinate concepts, to be able to synthesize perceptions, actions, and knowledge, to be able to form discrete chains of inference into a single idea in the now.
In recent years Dr Antonio Lieto (who presented his book a few years ago to the forum although is unable to attend this event) has argued that it is cognitive plausibility of AI which is a paramount goal. He presents a means of judging such plausibility in an AI based on three criteria: its functional/structural ratio, its generality, and its similarity to human performance.
It seems to me that some people might be more inclined to believe an AI understands reality if the AI has a high degree of cognitive plausibility -- yet this is a contentious claim.
Let's discuss this basic question: to what degree if any should an AI have an internal operating structure like a mind? Does it really matter if it appears to work?
We can use the paper below as a start.
Abstract
In this paper, we employ the Minimal Cognitive Grid (MCG), a framework created to
evaluate the cognitive plausibility of artificial systems, to offer a systematic assessment of
leading computational models of analogy and metaphor, including the Structure-Mapping
Engine (SME), CogSketch, METCL, and Large Language Models (LLMs). We present a
formal and quantitative operationalization of the MCG framework and, through the analysis
of its three main dimensions (Functional/Structural Ratio, Generality, and Performance
Match), examine how well each system aligns with standard cognitive theories of the modeled
phenomena, thus allowing for comparison of the models with respect to their cognitive
plausibility, according to consistent and generalizable mathematical criteria