Minsky’s Society of Mind

Artificial intelligence pioneer Marvin Minsky calls the multiple semi-independent modules in the human mind, generated by physically locatable modules in the human brain, The Society of Mind (not coincidentally the name of his book). It is a model of human cognition constructed, step by step, from the nonconscious interactions of simple mindless elements he calls “agents” (Minsky, 1986).

Awareness and the Limits of Knowledge

“What does it mean to say you're aware of yourself?” Minsky asks. It would be impossible “for any one part of the brain to know what's happening in all the other parts of the brain because there's just too much. Each part of the brain has connections to other parts of the brain and can get some ideas, but there's no place that knows everything” (Minsky, 2007b).

“The Society of Mind,” according to Minsky, is the end product of a vast evolutionary history, beginning with just clumps of neurons. Because neurons evolved early and had to keep their physiological integrity, progress was made by neurons gathering together, which led to the first small brains, and when these small brains began to specialize as well as to associate, “mind” began to develop (Minsky, 2007b).

Minsky is as blunt as he is insightful. “While many neuroscientists focus on how brain cells [neurons] work, to me, that's pretty much like trying to understand a computer from how transistors work. The neurons and synapses are maybe six levels of organization below the thoughts that you're actually aware of, the important things that distinguish a human from a crayfish. These high-level descriptions are what counts, and each of them has to be understood by itself. Any particular thing that happens in Level 5 can be understood as a combination of maybe 20 or 50 things that happen in Level 4 and so forth. But you can't understand Level 5 even if you know everything about how neurons and synapses work. The difference between a human and a crayfish is that a human has these multiple levels of brain organization that the earlier animals did not have” (Minsky, 2007b).

Models, Machines, and Memory

Actually, Minsky says, “I'm interested in how this piece of machine, the brain, can do things like decide that what it’s doing isn't working. How does it develop new goals? How does it develop new methods for achieving its goals? And, most important, how does it make a model of itself as a being in a world and think high-level stuff about its own past and its future?”

It has been known for well over 100 years that the brain has many different parts. Minsky envisions something “like a great network of computers, each of which is specialized. It's not that it’s a society of little people, but rather a society of biological machines, say 400 or more of these, each with different top-level functions, including the capacity to imagine planning proposals and counterfactual histories.”

Minsky speculates that cortical columns of related neurons, which are intermediate in complexity, can store things for a certain period without any changes in probability or conductions. We evolved these structures, he says, “so we could have reliable short-term memories that represent knowledge in many different ways.” In context, Minsky advises studying “insulation theory.” He says, “Theorists called ‘connectionists’ say what's important about the brain is how things are connected to each other. You could argue that it’s even more important to know how things are insulated from each other—why you don't get a big traffic jam because there's too many connections” (Minsky, 2007b).