Watch a flock of starlings at dusk and you’ll see something that some believe looks similar to thought. Thousands of birds swoop, fold, and turn as one, as if a single mind were steering them. But there is no leader. As Gaurav Suri and James McClelland write in The Emergent Mind, “there is no overseeing intelligence. Each bird is simply interacting with a small number of neighboring birds.”
That image is near the heart of their book, and it is the second thread of our upcoming COSM panel, which I described earlier. Their “central proposal,” they write, “is that thinking emerges from components that themselves cannot think.” No single neuron understands anything, any more than a single starling choreographs the flock. The mind, on this view, is what billions of units do together.
What Emergence Explains
Emergence is a powerful idea, and the authors offer a second example of it: “No single water molecule has the property of wetness.” Wetness appears only when many molecules interact.
It helps to notice that scientists use “emergence” in two senses. In weak emergence, the whole is surprising but fully explained by its parts. Once you understand how molecules attract one another, wetness holds no further mystery. In strong emergence, something genuinely new appears, with powers of its own that act back on the parts.
Which kind is the mind? The answer matters. If thought is emergent only in the weak sense, then a complete account of neurons would, in principle, be a complete account of thinking. If it is emergent in the strong sense, the mind may be more than its machinery, even if it arises from it. The book’s examples lean toward the first sense. Whether the mind fits that model as neatly as a flock or a puddle is one of the questions our panel will take up.
Where Knowledge Lives
The authors make a second claim with striking implications. What we as humans know, they argue, is not stored like files in a cabinet. Instead, “what is left behind after the experience is not the pattern of activation itself but rather changes in the strengths of the connections” between neurons. Memory and knowledge live in how strongly neurons influence one another.
They extend this to knowledge we’re born with. A newborn, they note, seems to “know” how to cry differently depending on what’s wrong. They put “know” in quotation marks, and that choice invites reflection. A baby’s built-in response and a scientist’s grasp of a theory are both called knowledge. Are they the same kind of thing at different scales, or different kinds of thing altogether?
Rules Without Rules
A related idea is just as provocative. We often seem to follow rules, in grammar, in logic, in everyday reasoning. The authors argue that the rules are, in a sense, an appearance: “our networks often behave as if they are following rules, just as our minds often operate as if they are implementing hard-coded rules, but they do so without actually using these rules.”
If that’s right, it reshapes how we think about reasoning itself. When you conclude that if A implies B and A is true, then B must be true, are you following a rule, or is your brain settling into a pattern that happens to match it? And if the latter, what makes one inference correct and another a mistake? These are old questions in philosophy, and the authors’ framework gives them new urgency.
An Interesting Vocabulary
One more feature of the book is worth noticing. In explaining how evolution shapes the brain, the authors write that it “prepares specific pathways,” “endows us with learning mechanisms,” and provides connections “designed to be used in specific contexts.” Elsewhere, they describe the artificial networks behind today’s AI as networks “that did not rely on the intelligence of a designer,” since they learn their connections from experience.
Biologists often use purposeful-sounding language as convenient shorthand, and the authors may intend nothing more. Still, the vocabulary raises a fair question. An AI system learns its connections on its own, but people designed its architecture, its learning rule, its training data, and its goal. Perhaps the right way to put it is that the knowledge isn’t designed, but the learner is. If so, what plays that role for the human brain? That question doesn’t belong to any one side of the debate, and I expect both of our panelists to have something to say about it.
When Gaurav Suri and Jay Richards take the stage, I’ll ask them directly: Is the mind emergent in the weak sense or the strong one? Is knowing something just a pattern of connections? And is the language of design in the book only a figure of speech? Their answers will help clarify what we mean when we say a mind emerges.
Join us at COSM in Scottsdale, AZ, November 18–20; our panel is Thursday, November 19, at 11 am. Next in this series: the hardest question of all: consciousness.