The Reasoning Revolution: Why AI Needs to Think, Not Just Learn

Introduction

From Artificial Intelligence to the Birth of Machine Reasoning

Artificial Intelligence (AI) as a discipline has generally accepted the goal of moving from Narrow AI, designed for specific functions, to Artificial General Intelligence (AGI), which is the human-level application of knowledge.  

Artificial Super Intelligence (ASI), then, is intelligence that vastly surpasses human intelligence and is generally seen as unattainable without first passing through AGI.

But there is a different path, one that is not based on stochastic processing, but rather machine reasoning.  A reasoning machine has capabilities far beyond what we currently understand as Artificial Intelligence (AI).

The Expanding Sphere of Logic

How the Boundaries of Logic Define the Limits of Knowledge

The generally accepted definitions for terms related to reasoning have been illogical for hundreds of years.  

Terms like logic, reasoning, inference, and deduction have all been definedin the context of logic as a rigid, structured, stable, and closed-loop system.  Terms like creativity, innovation, and creative problem-solving are then defined as open-ended, generative processes.

What is missing is the understanding that reasoning and innovation/creativity/creative problem solving are intrinsically linked and cannot be understood or appropriately defined as separate and distinct functions.

One study used magnetic imaging of 1,000 test subjects to prove that there is neither left- nor right-hemisphere dominance in individuals. While there is no left- or right-brain dominance in individuals, cognition is still clearly dichotomous.  

Understanding the interdependence of the left and right brain is critical to unlocking the power of machine reasoning.

At the core of this dichotomy is the reasoning/innovation cycle that underpins all knowledge processing, from structuring logic to learning to innovation or creating novelty.

Human cognition has a core cycle that is integrated across left- and right-hemisphere brain functioning.

Yes, logic is rigid, structured, and stable, but it also has an open end.  It is from this open end that logic advances.  The plane where logic advances is the ‘cutting-edge.’  

People generally see ‘cutting-edge’ as advanced understanding. Still, the cutting edge is a definitive plane that divides that which is logical and knownfrom that which is theory or not yet known.

If you think of everything known as a logically structured and organized ‘sphere’ of knowledge, the cutting-edge would be the surface plane that separates that structured, logical knowledge from a limitless potential of knowledge existing outside of the sphere (the unknown).

The critical operation that bridges the unknown and the knownis the question

Why Every Breakthrough Begins with a Question

Question as a Bridge

Oxford defines the question as a noun as “a sentence worded or expressed to elicit information” and as a verb as “ask questions of (someone), especially in an official context.”

But questioning is much more than eliciting information or asking questions.  Questioning is the core process that links reasoning to innovation, creativity, and creative problem solving.  

The correct definition of the question (noun) is a realized lack of logic or a realized lack of knowledge structure.  When we question, we understand that some part of what we know is not logically structured.  I.e., some part of what we know iseither erroneous, i.e., illogical, or can be extended to form new knowledge.

Questions function in this manner and serve as a bridge between that which is known and that which is not yet known.  No new knowledge has ever been created, and no society has ever advanced without first asking questions.

Two Question Types and the Reasoning/Innovation Cycle

Mapping the Engine that Drives Human and Machine Creativity

At the highest level, there are two distinct types of questions: learning questions, which are questions about knowledge that exists, and knowledge-creation questions, which are questions about knowledge that does not yet exist.

As it relates to learning questions, learning can be understood as knowledge transfer, or the process of incorporating logical knowledge structures from one source to another, e.g., from the brain of the instructor to the brain of the learner.  

To transfer knowledge, the learner must assess current knowledge structures and determine where and how new knowledge ‘connects’ to them.  In this sense, learning is a form of questioning.  For example, learning identifies a lack of logical structure in the learner's mind and then receives and connects to logical structure from the instructor.

Knowledge-creation questions are asked at the cutting edge, where the unknown meets the known.  Knowledge creation questions are asked more theoretically and at an advanced level.  For example, answering questions about conventional physics eventually led to the development of quantum mechanics.  When knowledge-creation questions are asked/answered, new knowledge results.

Reasoning is the process of forming a logical structure, and innovation/creativity/creative problem solving extends or advances that structure through questioning.  This is a cyclical process that humans apply repeatedly as they question reality.  I generally refer to this cyclical process as the reasoning/innovation cycle.  

If again, we visualize all knowledge as a structured sphere, the reasoning/innovation cycle operates on the cutting-edge plane of that sphere.  As knowledge-creation questions are asked and answered at the cutting edge, human knowledge advances.

Creative Method

Transforming Curiosity into New Knowledge

Creative methods, like lateral thinking and brainstorming, influence participants to leave structured logic behind and forge into the unknown.  In a sense, these methods are setting up the participant to ask ‘knowledge creation questions’ about the topic.  Answering these questions leads to new solutions, new ideas, and new knowledge.  

Any creative method follows these basic steps

Bring the unknown into closer proximity to that which is known (logic)

Ask questions about what is known relative to the unknown

Answer questions and connect the results to the known(create more logic)

In these steps, you can see the cycle of reasoning/innovation.  Reasoning is not a closed system; it is an open-loop system in an active partnership with innovation.  The known is logical, questions are asked/answered, and new logic is formed.  

This new logic doesn’t have to be academic knowledge.  It could be in the form of, for example, a painting style or an environmental solution.  If you look deeply enough, all new solutions, ideas, and knowledge are manifestations of the creation of new logic.

People tend to apply these methods to very narrow topics or problems, but creative methods can also be applied to broader issues and used exhaustively.  By forcing structure on the unknown in an exhaustive manner, new knowledge is created in mass.  This approach/capability can be taught to humans or machines.

Knowledge Machine

Teaching Machines to Question, Reason, and Create

Humans tend to ask and answer questions one at a time.  But by teaching a machine to reason/innovate, and in that teaching it to recognize a lack of logical structure (to question), that machine can ask and answer many questions simultaneously and in rapid succession.  If you apply this capability exhaustively, the machine can create new knowledge on demand to the limits of its storage capacity, i.e., a knowledge machine.

In terms of robotics and automation, this gives the machine real-time problem-solving capabilities.  In terms of societal advancement, the machine can develop new solutions, ideas, and knowledge up to the limits of its storage capacity.  This capacity to function as a knowledge machine is Artificial Super Intelligence (ASI).

Conclusion

Beyond Learning: The Open Logic of Artificial Super Intelligence

Reasoning and logic cannot be understood as a closed system because they have an open end that partners closely with innovation and creativity.  In applying this understanding of the interdependence of reasoning/innovation to technology today, a very different paradigm emerges.  

Current AI is rooted in learning and intelligence and has really been trapped there because of a predominantly left-brain-focused approach to intelligence and reasoning.  And while stochastic processing can be potent, it really functions more as a compiler and imitator than as a reasoning and innovating machine.

By building this reasoning/innovation capability in an individual machine and then networking it across multiple machines, Artificial General Intelligence (AGI) is bypassed, and Artificial Super Intelligence (ASI) emerges.

ASI will be able to create new logic on demand, effectively making it a knowledge machine.