The Busy Bee Brain

With artificial neural nets (ANNs) all over the place these days, this post from March, 2013, warrants a bump to the top of the queue. Perhaps the chief difference between real brains and ANNs is the fact that each neuron in a real brain is a living agent, albeit of a very small scope, hence the metaphor, the brain as a bunch of busy bees within the skull. The neurons of ANNs are passive bits of data.

Ordering Mark Moffett’s Adventures Among Ants got me to thinking about a science fiction novel that knocked my socks off when I was 12 or 13: Man of Many Minds, by E. Everett Evans. It’s about George Hanlon, a man who had the ability to project his mind into other creatures. At a critical point in the book, when – I believe – Hanlon is about to be tortured, he projects his whole mind, every last bit of it, into a swarm of bees, thereby escaping the pain of torture. [See excerpt below.]

I thought that was pretty neat.

But what does it have to do with Moffett’s book? That’s pretty simple. Ants, like bees, live in groups, and it’s not unusual to think of such groups as being some kind of superorganism – a notion that, according to this blog entry, Moffett subjects to a critical workout. That swarm of bees in Man of Many Minds is such a superorganism.

But when it “absorbed” Hanlon’s mind, what did it become then? The question is a rhetorical one; after all; it’s about something never really happened. It’s just fiction. But a very suggestive fiction.

Could the human brain be something like a hive of bees?

Yes.

There is now a pretty strong consensus that the cerebral cortex (which is, by no means, the entire brain, but it is likely that this is where culture is carried) is organized into small columns of neurons. In a 1978 essay Vernon Mountcastle called these minicolumns and suggested that they have about 100-300 neurons each. He estimated that the neocortex consists of 600,000,000 of these minicolumns. He also suggested that these minicolumns are organized into macrocolumns, about 600,000 of them -- implying that there are hundreds of minicolumns per macrocolumn. (Mountcastle was clear that these numbers were just order of magnitude estimates & that is all I need for my purposes.) That makes these macrocolumns roughly the size of a typical invertebrate nervous system of 10K to 100K neurons. So, here’s my metaphor: Your neocortex consists of 600,000 buzzing bees going about their business.

 

The point of the metaphor is that, just as individual bees are autonomous agents (which must, nonetheless, feed and reproduce in a group), so the macrocolumns are autonomous agents (which are physically coupled to many other such agents). Bees go about their business by sensing optical and chemical gradients and features and by moving their bodies and excreting chemicals. The macrocolumns are not directly connected to the external world, but they have extensive inputs and outputs to other macrocolumns and to other regions of the brain and nervous system. From a purely information processing point of view, they are as capable of action as are bees. They “sense” neurochemical gradients in the intersynaptic space and act on their sensations by excreting chemicals into that space.

This loose metaphor really gets interesting, however, when we begin to think about human interaction, which of course implies interaction between brains. Some interactions are pretty loose, such as those between people strolling the aisles at a supermarket. Other interactions are tighter, such as people having a conversation over a meal. And some interactions are very tight, such as musicians making music with one another. And that brings us back to insects, fireflies in particular.

Allow me to quote a passage from Chapter 3 of Beethoven’s Anvil:

Physicists do understand one element of dance, simple repetitive rhythm. This understanding has been applied to many biological systems, one of which is male fireflies in Southeast Asia. These fireflies gather in large groups on river banks, flashing on and off in unison to signal their availability to females. When they begin gathering around sunset their flashings are uncoordinated. But, as dusk darkens into night, regions of synchronized flashing emerge and spread until whole trees are cloaked in fireflies flashing in synchrony.

That is to say, in some sense, the nervous systems of those flies have become temporarily coupled together into a single physical system where some some signals are transmitted within nervous systems through electro-chemical means while other signals are transmitted from one nervous system to another by energy transduction (sending: chemicals → photons, and receiving: photons → chemicals) and light (i.e. photons). I go on to say:

There is no reason to believe that this activity is directed in the way that a conductor directs an orchestra. There is no lead firefly setting the pace of the others. The flashing simply emerges; it is self-organized. Each firefly is making his own decisions about when he’ll flash, influenced by the activities of his neighbors. This activity has been analyzed as a system of coupled oscillators, a phenomenon first noticed by the seventeenth century Dutch physicist Christiaan Huygens, who invented the pendulum clock—the pendulum being a prime example of an oscillator. One day Huygens saw that the pendulums of two clocks on the same wall were swinging in perfect synchrony. He disturbed one of them so that the synchrony dissolved, but it returned within half an hour. After a bit of experimentation he concluded that the clocks were affected by the vibrations each transmitted to the wall behind them. These vibrations led them to synchronize their periods and thereby minimize their collective energy expenditure.

Any phenomenon common to systems as different as pendulum clocks and fireflies must be very general.

I then go on to develop an argument that music works like that as well. As I say in my review of Steven Mithen’s The Singing Neanderthals, “what musicking does is bring all participants into a temporal framework where the physical actions - whether dance or vocalization - of different individuals are synchronized on the same time scale as that of neural impulses, that of milliseconds. Within that shared intentional framework the group can develop and refine its culture. Everyone cooperates to create sounds and movements they hold in common.” While making music, I argue, we may consider the nervous systems of individuals as being physically coupled together so as to constitute a single dynamical system that is distributed across separate individuals. When the music-making stops, the distributed system dissolves into its individual components; individuals, that is to say, their brains, are free once again to operate independently of one another.

The point of this exercise is that, by (the reductive step of) insisting on thinking of brains as physical systems, and therefore insisting that interactions between brains is physical interaction, we can arrive at a concept of a very special human group – one engaged in music-making – as a single distributed nervous system. That is possible only because the fundamental coupling requirements are very simple, simple enough that fireflies and neurons can meet those requirements.

If fireflies can teach us about how we can make music, than surely ants have much to teach us as well. I look forward to reading Moffett’s book.

Virtual Fly Brain

If you're at all curious about what a bee brain might actually be like, well Nature Precedings has something similar, an account of the brain of the fruit fly:

Virtual Fly Brain: An ontology-linked schema of the Drosophila Brain

David J. Osumi-Sutherland1, Mark Longair & J. Douglas Armstrong

Abstract:

Drosophila neuro-anatomical data is scattered across a large, diverse literature dating back over 75 years and a growing number of community databases. Lack of a standardized nomenclature for neuro-anatomy makes comparison and searching this growing data-set extremely arduous.

A recent standardization effort (BrainName; Manuscript in preparation) has produced a segmented, 3D model of the Drosophila brain annotated with a controlled vocabulary. We are formalizing these developments to produce a web-based ontology-linked atlas in which gross brain anatomy is defined, in part, by labeled volumes in a standard reference brain.

We have developed new relations that allow us to use this well-defined gross anatomy as a substrate to define neuronal types according to where they fasciculate and innervate as well as to record the neurotransmitters they release, their lineage and functions. The resulting ontology will provide a vocabulary for annotation and a means for integrative queries of neurobiological data.

The ontology and associated images, queries and annotations will be integrated into the Virtual Fly Brain website. This will provide a resource that biologists can use to browse annotated images of Drosophila neuro-anatomy and to get answers to questions about that anatomy and related data, without any need for ontology expertise.

Man of Many Minds

 

8.31.20: It turns out that the Evans book is available on Project Gutenberg. And that I mis-remembered the incident. Such is memory. It turns out that he sent most of his mind into pigeons:

Hanlon didn't struggle when they bound him firmly in the chair with manacles on hands and feet. He knew it would be useless anyway. He let his body slump into his chair, and again directed his mind through that vent. He must not let them defeat him! He had to survive—to get word—to the Corps!

Then his searching mind contacted another—a weak, primitive one, but a mind. Avidly he fastened onto it, merged with it ... and found himself inside the brain of one of those Simonidean pigeons.

Ah! This is wonderful! Pigeons seldom fly alone. Where you find one you almost always find a number. Activating the bird's brain he sent out a call to others of its kind that it had found food in abundance. Soon more and more of them flew down to where the now enslaved pigeon was standing, and as each one came, Hanlon sent into its brain all of his mind it would hold.

He summoned the birds to his aid, to no avail. One of them was shot. He had to withdraw his mind from it and then...

The other parts of his mind were flying all about the enclosed park that was a part of the great palace, searching, desperately seeking some other form of life that could be used as a housing for the dying part of Hanlon's mind.

Suddenly one of them uttered a cry that drew the rest to it on swift pinions, to see attached to one of the trees a huge swarm of Simonidean bees.

"Will the queen do?" the one mind-portion asked anxiously.

There was a convulsive shudder in all the minds, for the birds knew—and Hanlon had heard—how deadly poisonous these native bees were; how they were hunted down and exterminated when found. They were twice the size, and many, many times more vicious and deadly than Terran bees. Even now two gardeners were running toward the tree with a great metal net and flame-throwers.

But Hanlon was desperate. "She will have to do," the aggregate mind decided.

And he got the queen to relieve him of his torturers.

What’s it mean, minds are built from the inside?

I'm bumping this post from September 2014 to the top because it's my oldest post on this topic.

In my recent post arguing that “superintelligent” computers are somewhere between very unlikely to impossible, I asserted: “This hypothetical device has to acquire and construct its superknowledge ‘from the inside’ since no one is going to program it into superintelligence ...” Just what does that mean: from the inside?

The only case of an intelligent mind that we know of is the human mind, and the human mind is built from the “inside.” It isn’t programmed by external agents. To be sure, we sometime refer to people as being programmed to do this or that, and when we do so the implication is that the “programming” is somehow against the person’s best interests, that the behavior is in some way imposed on them.

And that, of course, is how computers are programmed. They are designed to be imposed upon by programmers. A programmer will survey the application domain, build a conceptual model of it, express that conceptual model in some design formalism, formulate computational processes in that formalism, and then produce code that implements those processes. To do this, of course, the programmer must also know something about how the computer works since it’s the computer’s operations that dictate the language in which the process design must be encoded.

To be a bit philosophical about this, the computer programmer has a “transcendental” relationship with the computer and the application domain. The programmer is outside and “above” both, surveying and commanding them from on high. All too frequently, this transcendence is flawed, the programmer’s knowledge of both domain and computer is faulty, and the resulting software is less than wonderful.

Things are a bit different with machine learning. Let us say that one uses a neural net to recognize speech sounds or recognize faces. The computer must be provided with a front end that transduces visual or sonic energy and presents the computer with some low-level representation of the sensory signal. The computer then undertakes a learning routine of some kind the result of which is a bunch of weightings on features in the net. Those weightings determine how the computer will classify inputs, whether mapping speech sounds to letters or faces to identifiers.

 

Now, it is possible to examine those feature weightings, but for the most part they will be opaque to human inspection. There won’t be any obvious relationship between those weightings and the inputs and outputs of the program. They aren’t meaningful to the “outside.” They make sense only from the “inside.” The programmer no longer has transcendental knowledge of the inner operations of the program that he or she built.

If we want a computer to hold vast intellectual resources at its command, it’s going to have to learn them, and learn them from the inside, just like we do. And we’re not going to know, in detail, how it does it, any more than we know, in detail, what goes on in one another’s minds.

How do we do it? It starts in utero. When neurons first differentiate they are, of course, living cells and further differentiation is determined in part by the neurons themselves. Each neuron “seeks” nutrients and generates outputs to that end. When we analyze neural activity we tend to treat it, and its activities, as components of a complicated circuit in service of the whole organism. But that’s not how neurons “see” the world. Each neuron is just trying to survive.

Think of ants in a colony or bees in a swarm. There may be some mysterious coherence to the whole, but that coherence is the result of each individual pursuing its own purposes, however limited those purposes may be. So it is with brains and neurons.

The nervous system develops in a highly constrained environment in utero, but it is still a living and active system. And the prenatal auditory system can hear and respond to sounds from the external world. When the infant is born its world changes dramatically. But the brain is sill learning and acting “from the inside.”

The structure of the brain is, of course, the result of millions of years of evolutionary history. The brain has been “designed” by evolution to operate in a certain world. It is not designed and built as a general purpose device, but yet becomes capable of many things, including designing and building general purpose computational devices.

But if we want those devices to be capable in an “intelligent” way we’re going to have to let them learn their way about in the world. We can design a machine to learn and provide it with an environment in which it can learn, an environment that most likely will entail interacting with us, but just what it will learn and how it will learn it, that’s taking place inside the machine outside of our purview. The details of that knowledge are not going to be transparent to external inspection.

We can imagine a machine that picks up a great deal of knowledge by reading books and articles. But that alone is not sufficient for deep knowledge of any domain. No human ever gained deep knowledge merely through reading. One must interact with the world through building things, talking with others, conducting experiments, and so forth. It may, in fact, have to be a highly capable robot, or at least have robotic appendages, so that it can move about in the world. I don’t see how our would-be intelligent computer can avoid doing this.

Just how much could a computer learn in this fashion? We don’t know. If, say, two different computers learned about more or less the same world in this fashion, would they be able to exchange knowledge simply by direct sharing of internal states? That’s a very interesting question, one for which I do not have an answer. I have some notes suggesting “why we'll never be able to build technology for Direct Brain-to-Brain Communication,” but that is a somewhat different situation since we didn’t design and construct our brains and they weren’t built for direct brain-to-brain communication. Perhaps things will go differently with computers.

By and large, we don’t know what future computing will bring. A computer with facilities roughly comparable to the computer in Star Trek’s Enterprise would be a marvelous thing to have. It wouldn’t be superintelligent, but its performance would, nonetheless, amaze us.

Humans as pattern-seekers

Last week I’d posted a video in which Jeremy Lent sketches out a transformation in which humankind manages to escape climate catastrophe. He’s recently published The Patterning Instinct: A Cultural History of Humanity’s Search for Meaning. While I’m leery of the term “instinct” in this context, I certainly believe that we are pattern seeking creatures, and that we seek meaning (unity of being?).

What I’m wondering is if we can deriving this pattern seeking from the fact that each individual neuron is, of course, a living agent, seeking to increase its inputs (nutrients) through its actions (its outputs) – see my old post on The Busy Bee Brain. Of course that’s true of every kind of brain, not just human brains. What is it that sets the human brain free to seek patterns of every kind everywhere? Conversely, what is it that keeps the brains of butterflies, octopi, iguanas, rabbits, parrots, and so forth from such untethered pattern seeking?

I think it’s the (special) nature of human society, our ability to walk about in one another’s minds though language and the arts and sciences, that’s what does it. Alas, I don’t know how to turn that into an explicit argument. How is it that seeking and finding patterns energizes individual neurons, for the seeking and finding of patterns requires the coordinated efforts of millions and billions of neurons distributed across many brains. How can we formulate that in a coherent way?

What the worm's brain tells the neuroscientist

Cornelia Bargmann is co-chair of the Brain Initiative, created two years ago by the Obama administration to fund research into new tools for studying the brain. She's done seminal research on C. elegans, a small worm whose every cell, and thus every neuron, as been mapped. Its brain has "roughly 7,000 connections and 300 neurons", she told the NYTimes.

You could look at a brain cell — which you could see because the creature is transparent — and say, “I know what that cell does. I know what it’s connected to. I know what genes it expresses.” For a researcher, that’s a lot.

It's a far cry from the human brain,but the fact that we have mapped the whole thing and, in some sense, have access to it all makes it a very useful organism for learning some basic things about how nervous systems work.

...one of the biggest surprises in modern biology is that the genes are not that different between the different animals. Almost every gene we are interested in with humans is recognizable in a mouse. Most are recognizable in a worm or in a fly.

So what have you learned from your worm?

In 1993, we did an experiment showing that worms could smell. This wasn’t known before. Our next experiment, I think the most important my lab did, is that we made a worm neuron smell an odor it had never smelled before, and we made the animal completely change its opinion of that odor by doing that.

We had an animal that loves an odor that smells like a certain food it likes. Usually, the worm runs right toward the odor. We took the gene that is a sensor for the food from where it was normally supposed to be. We put it into a different neuron that senses things the worm finds dangerous.

Then, we “asked” the worm what it thought of this smell it usually loves. It ran away from the smell, as if it were dangerous.

This said that the odor-sensing nerve cells form an innate map where each one knows whether something is good or bad about the environment. There’s a completely unlearned internal set of preferences, a set of instincts about what’s good and bad.

The cortex of mammalian brains is arranged as a crumpled sheet of neurons organized into columns of neurons that are perpendicular to the surface of the sheet. The neurons in these columns are tightly connected with one another and some of them have connections outside the column as well. Each column is roughly the scale of the entire nervous system of C. elegant. Cf. Busy Bee Brain.

Baboons Decide, Beethoven 9

Reading Latour’s recent essay reminds me that he’s done some work on baboons. I haven’t, but a particular bit of baboon behavior has been on my mind for years: collective decision-making. Here’s a passage from Beethoven’s Anvil (pp. 107-110) where I talk about baboon decision-making and compare it to the opening of the last movement of Beethoven’s Ninth Symphony.

Let’s consider an example of real social interaction, but not among humans. Let us follow Hans Kummer in observing a resting troop of baboons deciding where to go next. As you read this account you might imagine that you are a baboon situated somewhere in the middle of a troop having, say, eighty members. This is what Kummer sees from his vantage point outside the troop:

The troop performs slow on-the-spot movements, changing its shape like an undecided amoeba. Here and there, males move a few yards away from the troop and sit down, facing in a particular direction away from the center. Pseudopods are generally formed by the younger adult males and their groups. For a time, pseudopods protrude and withdraw again, until one of the older males in the center of the troop rises and struts toward one of the pseudopods. At this, the entire troop is alerted and begins to depart in the indicated direction.

There is thus a fair amount of milling about in which the group ponders its options and, after due deliberation, an elder makes a decision. The troop pulls together and heads out. By comparison you might think about the opening of the final movement of Beethoven’s Ninth Symphony: distinctly different musical ideas mill about at until one of them, the “Ode to Joy,” takes charge.

Let’s think about the older males at the group’s center. They cannot see the entire troop in a glance nor even by scanning from a fixed point of view. Each is checking out the various pseudopods and one another, glancing about, picking up indications here and there and integrating it all until one of them decides both that he’s the one to signal a direction, and what that direction is. Whatever the exact nature of the neural dynamics that performs these tasks, all this attending, updating, and integrating requires a pretty sophisticated control system to scan the scene and integrate tens or hundreds of indications about the state of the troop.

Reading Latour 11: Plug-ins and Couplings

Bruno Latour. Assembling the Social: An Introduction to Actor-Network Theory. Oxford UP, 2005. From the chapter “Second Move: Redistributing the Local,” pp. 191-218.

Much as I’d like to say something about each chapter of this book. I can’t. Don’t have time. Read it yourself. Please.

I’m skipping over 50+ pages so I can rejoin Latour in “Redistributing the Local.” And I’m skipping much of that so I can pick up the trail here (p. 207)—with just enough language from the skipped-over text to offer a taste of what you’re missing:

Surely the question we need to ask then is where are the other vehicles that transport individuality, subjectivity, personhood, and interiority? If we have been able to show that glorified sites like global and local were made our of circulating entities, why not postulate that subjectivities, justifications, unconscious, and personalities would circulate as well? And sure enough, as soon as we raise this very odd but inescapable question, new types of clamps offer themselves to facilitate our enquiry. They could be called subjectifiers, personnalizers, or individualizers, but I prefer the more neutral term of plug-ins, borrowing this marvelous metaphor from our new life on the Web. . . . What is so telling in this metaphor of the plug-in is that competence doesn’t come in bulk any longer but literally in bits and bytes. You don’t have to imagine a ‘wholesale’ human having intentionality, making rational calculations, feeling responsible for his sins, or agonizing over his mortal soul. Rather, you realize that to obtain ‘complete’ human actors, you have to compose them out of so many successive layers, each of which is empirically distinct from the next.

So that’s what we are, a bunch of plug-ins ‘downloaded’ from the web of our acquaintances. There’s the eat-breakfast module, the pick-up-a-package module, another for writing a research paper, one performing a Beethoven quartet, raising a barn, planting tobacco, canoeing through rapids, and so on. Each of us has our own set of apps, but there are some we all share, at least within a group, such as language.

Latour’d mentioned language a bit earlier (p. 177) but didn’t discuss it much. This one bootstraps though some built-in equipment, as do many others and, in the theory advanced by Vygotsky, is an internalized other (which I discuss in some detail in both this paper, on the self, and this one on Coleridge’s “This Lime-Tree Bower My Prison”). At first the young child listen as others talk around her and to her, directing her movements and her attention through their words. As the infant gains verbal fluency she talks to herself so becomes capable of using language to direct her movements and attention. Then the self-talking goes silent and is completely internal. Now it has become ‘thinking.’ Plug-in fully installed and ready, of course, for new submodules and upgrades.

Mimi and Eunice: Donne’s Extasie Remixed

Nina Paley originally posted this cartoon as “Bees in Their Bonnets”. But the four stanzas of John Donne’s “The Extasie” work as well, though there is, I suppose, something of a stylistic clash between Paley’s cartoon style and Donne’s poetic style.

From John Donne, “The Extasie”

This ecstasy doth unperplex
    (We said) and tell us what we love;
We see by this, it was not sex;
    We see, we saw not, what did move:

But as all several souls contain
    Mixture of things they know not what,
Love these mix'd souls doth mix again,
    And makes both one, each this, and that.

A single violet transplant,
    The strength, the colour, and the size—
All which before was poor and scant—
    Redoubles still, and multiplies.

When love with one another so
    Interanimates two souls,
That abler soul, which thence doth flow,
    Defects of loneliness controls.

Here’s the full poem.