AI Safety

The Framework’s Potential Relevance to AI

The BAL-looping framework proposes that human cognition depends not only on a core system for internal modeling and goal-directed control, but on the way that core is embedded within a larger functional architecture.

As the framework papers point out, the o-channel and i-channel of the PTD operate in tandem. Beyond this, the framework posits that the BAL learns to exploit the naturally occurring resonance in the i-channel by attending to it as it would to any interindividual communication. According to the framework, this attendance generates every mode of subjective experience.

This raises a broader question for artificial intelligence:

How would the capabilities and autonomy of an AI core change if that core were embedded in some sort of larger functional architecture? And “embedded in a functional architecture” here does not mean that an AI is powering some sort of agentic operation. It means that the AI’s output is produced by an AI core embedded within a larger architecture that leverages it.

The BAL-looping framework is relevant because it describes one such architecture that exists in biological cognition, with the BAL functioning as a core intelligence leveraged by the operation of looping. Although many such architectures could be possible, BAL-looping describes one that already exists and is known to work.

According to the framework, it is the BAL-looping operation that allows for recollection, imagination, and conscious planning within a single mind. Humankind would not possess nearly as much individual or collective intelligence without this operation. The difference between our nonlooped potential and what we have achieved by using this operation may be an indication of how greatly an AI core could likewise be amplified when embedded in an effective functional architecture..

Given this potential difference in capability, these considerations are clearly relevant to questions of AI safety.

Consulting

I offer independent consulting to AI laboratories, research groups, companies, and other organizations examining questions at the intersection of artificial intelligence and cognitive architecture.

Consulting may include analysis of specific research or design questions, written responses or memoranda, review of conceptual models or proposed architectures, participation in meetings or advisory committees, and ongoing consultation where BAL-Looping may provide a useful perspective.

Organizations with a specific question are welcome to make contact with a brief description of the problem they are examining, at jnormansp@gmail.com.