I rarely embed videos here at LabKitty even though all the Blogging fer Dummies books recommend doing so because shiny object. I like to think my traffic is a more refined lot, one who prefers long ranting text bricks about whatever esoterica is driving me this morn. I also don't embed video because Google changes their embedding HTML like my sister changes boyfriends (known throughout the Circle Pines metro area as the "one hitter quitter"). If anything screams blogging fer dummies, it's a dead video link.
Today I make an exception. This because Rafael Yuste and Christophe Dupre (Columbia) recently published an extraordinary result and took the extra PR step to post about it on YouTube.
Here's the vid (fingers crossed):
What you are viewing is the entire nervous system of a live hydra at work. By "entire" I do mean entire. Yuste and Dupre conjured gene magic that makes the hydra transparent and its neurons glow (Corrigendum: Apparently hydra come transparent out of the box. However, the glowy thing is an upgrade). Each flash in the video is one neuron firing. The accompanying everything else is the hydra doing hydra stuff.
This is what we (well, I) in the brain biz call the Niels Bohr Brass Ring of Neuroscience. It comes from Bohr's famous carp Nur die Fülle führt zur Klarheit. Only wholeness leads to clarity. Plainly translated: The brain is so complicated that no matter how many individual or group neuron recordings you make, no matter how skilled your experiment or analysis, it will never tell us what is really Going On. Ergo, somebody best figure out a way to record the whole thing at once. (Bohr was carping about Quantum Mechanics not Neuroscience, but what is Quantum Mechanics if not applied Neuroscience?)
With the advent of glowy hydra, we finally have an experimental platform in which we can record an entire nervous system -- every last neuron -- at the individual cell level and all at once. Bold revelations about biological information processing can't help but follow.
In fact, a casual viewing of the video already tells us one thing above all else.
As far as processing power goes, neurons suck.
The brain has been called the most complex device in the universe. I suppose. It's output is certainly impressive. From Archimedes to LabKitty, Shakespeare to LabKitty, or Newton to LabKitty, all manner of wonder has sprung from the brain's internal fits and spurts. This comes at a price, however, and the price is an almost impenetrable design. Just the layout of the thing is hard to grok, as anyone who has ever stared down the barrel of a neuroanatomy practical can attest. Setting aside the layout, just the sheer number of neurons that goes into making a sonnet or calculus or a quality Zazzle item beggars belief.
The figure usually thrown around for humans is 100 billion neurons. Such numbers should be taken as a ballpark. Counting neurons is a notoriously difficult task, and as the numbers get bigger, the ballpark gets fuzzier. Yet, my rant for today concerns what happens when we tally small brains, not big ones.
Someone compiled a neurometry list on Wikipedia, ordered by increasing brain size and including links to primary sources, which should get them some kind of Truth-in-Wiki award assuming Wikipedia doesn't hound them from the site for making everyone else look bad. (Confidential to Wikipedia: Aristotle was not Belgian, the central message of Buddhism is not "every man for himself," and the London Underground is not a political movement.)
I have studied this list many times, because that's the sort of thing I do. Humans are down near the bottom, with their 100 billion neurons or whatever. Elephants top primates, but they also top primates in sensory surface and motoneurons to process, so I am not much alarmed by Proboscidean ascendence. Indeed, all of the large mammals are billionaires (except for Donald Trump, amirite? Topical humor!). Dogs and cats and humpty-back camels. Lions and tigers and bears. Laboratory critters like rats and mice are in the dozens-to-hundreds million range. Birds and fish and frogs also, with things that fly generally having larger brains than things that swim or crawl.
It's only when we fall off the vertebrate cladistic branch that the numbers cease being astronomical. Cockroaches have about a million neurons. Ants and bees and flies are in the hundreds of thousands. Ocean weirdness like jellyfish and sea squirts are in the plain thousands. An odd duck here, so to speak, is the nematode -- a roundworm barely visible to the naked eye and without much of an intellectual life to speak of -- with 302. Exactly (a story for another time perhaps).
Personally, I can't scroll too far down the list before the numbers break my brain. I can't fathom 100 billion neurons or dozen billion or even one billion. I simply nod politely and drink. On the other hand, I kind of have a feel for the numbers closer to the top. I know what a million feels like. A million is a megabyte. A million is the population of a big city. A million is a square of numbers 1000 on a side. A largish jpeg on any screen with decent resolution.
That's a million.
And that's where my trouble begins. It shouldn't take a million neurons to operate a cockroach. To be clear: I'm not disputing the count; it's that I find the count philosophically offensive. It assaults my sense of good design. My engineering esprit de corps. My reverence for the CPU cycle.
I've had ample opportunity to observe the cockroach in its natural habitat (aka "off-campus housing"). I assure you these are not deep thinkers. There is no cockroach semiconductor industry. No cockroach poetry slam. You can't train them to fetch (we tried). They can't play baseball. They don't wear sweaters. They're not good dancers and they don't play drums. The little bastards don't do much more than scurry and eat.
One million neurons.
And not only the cockroach, but all creatures great and small. An ant shouldn't require ~100,000 neurons, nor a bee. And 302 to run a nematode? What the heck does a nematode even do besides get studied by developmental biologists? I'm thinking a dozen neurons should do the trick. Tops.
Which brings us to the hydra. I can't make sense of the hydra entry on the list (the scientific name and reference is hydra, but the common name and picture is a jellyfish) but my sleuthing elsewhere turned up a claim that its thinky bits comprise several thousand neurons arranged in a nerve "net" (the hydra nervous system occurs as a distributed network rather than being organized into discrete lumps, which is another reason the glowy hydra idea worked).
What does that buy us?
A hydra is about a quarter inch long. It lives in water. It spends its days stuck to something at one end, and waving tentacles about at the other to bring food specks into its gullet. When it wants to move, it does a somersault. When it wants to make babies, it buds.
People, this is not a lifestyle that requires several thousand neurons to implement. I've written Javascript apps that do more interesting things.
Let's put this in perspective. Consider 1000 classic spiking neurons. Ignore subthreshold processing. Ignore the time and length constant integration of synaptic inputs. Ignore intrinsic and network spike patterning. Simply 1000 yes/no, go/no-go, spike/no-spike processing elements. Such a system generates 2^1000 unique states. A terabyte is 2^40. If we simply wanted to list the states, assuming one kilobyte per state, it would require ~2^970 of those fancy SSDs you have on your Amazon wish list. That's a flash drive the size of a Dyson sphere.
The hydra has at least twice as many neurons as that.
For all of Shakespeare's fawning praise of man's noble reason and infinite faculties, it sure takes quite the neuron horde to accomplish anything useful. Yes, there are notable exceptions -- the goldfish Mauthner cell implements a tail-flip reflex all by its lonesome -- but Yuste and Dupre's hydras clearly (wordplay!) illustrate that, on the whole, neurons are an embarrassment. A Brobdingnagian waste of processing. Brains are what you'd get if Rube Goldberg was the head of NASA. They make the Trabant look like Formula-1 engineering. They make the bubble sort, COBOL, and Windows Vista seem sleek and efficient. No wonder half the people you meet are below average intelligence.
I suppose it was possible to be offended long before Yuste and Dupre happened along. After all, neuron counting goes back at least as far as Golgi. But, as is often the case, a new result has made it impossible to ignore an inconvenient truth. There is a cusp. A transition. A Eureka moment. That is what all good science does.
Neurons suck. Not something you'll read in a textbook, grant proposal, or presidential BRAIN initiative, but true nonetheless.
Neuroscience might want to start looking into that.
Today I make an exception. This because Rafael Yuste and Christophe Dupre (Columbia) recently published an extraordinary result and took the extra PR step to post about it on YouTube.
Here's the vid (fingers crossed):
What you are viewing is the entire nervous system of a live hydra at work. By "entire" I do mean entire. Yuste and Dupre conjured gene magic that makes the hydra transparent and its neurons glow (Corrigendum: Apparently hydra come transparent out of the box. However, the glowy thing is an upgrade). Each flash in the video is one neuron firing. The accompanying everything else is the hydra doing hydra stuff.
This is what we (well, I) in the brain biz call the Niels Bohr Brass Ring of Neuroscience. It comes from Bohr's famous carp Nur die Fülle führt zur Klarheit. Only wholeness leads to clarity. Plainly translated: The brain is so complicated that no matter how many individual or group neuron recordings you make, no matter how skilled your experiment or analysis, it will never tell us what is really Going On. Ergo, somebody best figure out a way to record the whole thing at once. (Bohr was carping about Quantum Mechanics not Neuroscience, but what is Quantum Mechanics if not applied Neuroscience?)
With the advent of glowy hydra, we finally have an experimental platform in which we can record an entire nervous system -- every last neuron -- at the individual cell level and all at once. Bold revelations about biological information processing can't help but follow.
In fact, a casual viewing of the video already tells us one thing above all else.
As far as processing power goes, neurons suck.
The brain has been called the most complex device in the universe. I suppose. It's output is certainly impressive. From Archimedes to LabKitty, Shakespeare to LabKitty, or Newton to LabKitty, all manner of wonder has sprung from the brain's internal fits and spurts. This comes at a price, however, and the price is an almost impenetrable design. Just the layout of the thing is hard to grok, as anyone who has ever stared down the barrel of a neuroanatomy practical can attest. Setting aside the layout, just the sheer number of neurons that goes into making a sonnet or calculus or a quality Zazzle item beggars belief.
The figure usually thrown around for humans is 100 billion neurons. Such numbers should be taken as a ballpark. Counting neurons is a notoriously difficult task, and as the numbers get bigger, the ballpark gets fuzzier. Yet, my rant for today concerns what happens when we tally small brains, not big ones.
Someone compiled a neurometry list on Wikipedia, ordered by increasing brain size and including links to primary sources, which should get them some kind of Truth-in-Wiki award assuming Wikipedia doesn't hound them from the site for making everyone else look bad. (Confidential to Wikipedia: Aristotle was not Belgian, the central message of Buddhism is not "every man for himself," and the London Underground is not a political movement.)
I have studied this list many times, because that's the sort of thing I do. Humans are down near the bottom, with their 100 billion neurons or whatever. Elephants top primates, but they also top primates in sensory surface and motoneurons to process, so I am not much alarmed by Proboscidean ascendence. Indeed, all of the large mammals are billionaires (except for Donald Trump, amirite? Topical humor!). Dogs and cats and humpty-back camels. Lions and tigers and bears. Laboratory critters like rats and mice are in the dozens-to-hundreds million range. Birds and fish and frogs also, with things that fly generally having larger brains than things that swim or crawl.
It's only when we fall off the vertebrate cladistic branch that the numbers cease being astronomical. Cockroaches have about a million neurons. Ants and bees and flies are in the hundreds of thousands. Ocean weirdness like jellyfish and sea squirts are in the plain thousands. An odd duck here, so to speak, is the nematode -- a roundworm barely visible to the naked eye and without much of an intellectual life to speak of -- with 302. Exactly (a story for another time perhaps).
Personally, I can't scroll too far down the list before the numbers break my brain. I can't fathom 100 billion neurons or dozen billion or even one billion. I simply nod politely and drink. On the other hand, I kind of have a feel for the numbers closer to the top. I know what a million feels like. A million is a megabyte. A million is the population of a big city. A million is a square of numbers 1000 on a side. A largish jpeg on any screen with decent resolution.
That's a million.
And that's where my trouble begins. It shouldn't take a million neurons to operate a cockroach. To be clear: I'm not disputing the count; it's that I find the count philosophically offensive. It assaults my sense of good design. My engineering esprit de corps. My reverence for the CPU cycle.
I've had ample opportunity to observe the cockroach in its natural habitat (aka "off-campus housing"). I assure you these are not deep thinkers. There is no cockroach semiconductor industry. No cockroach poetry slam. You can't train them to fetch (we tried). They can't play baseball. They don't wear sweaters. They're not good dancers and they don't play drums. The little bastards don't do much more than scurry and eat.
One million neurons.
And not only the cockroach, but all creatures great and small. An ant shouldn't require ~100,000 neurons, nor a bee. And 302 to run a nematode? What the heck does a nematode even do besides get studied by developmental biologists? I'm thinking a dozen neurons should do the trick. Tops.
Which brings us to the hydra. I can't make sense of the hydra entry on the list (the scientific name and reference is hydra, but the common name and picture is a jellyfish) but my sleuthing elsewhere turned up a claim that its thinky bits comprise several thousand neurons arranged in a nerve "net" (the hydra nervous system occurs as a distributed network rather than being organized into discrete lumps, which is another reason the glowy hydra idea worked).
What does that buy us?
A hydra is about a quarter inch long. It lives in water. It spends its days stuck to something at one end, and waving tentacles about at the other to bring food specks into its gullet. When it wants to move, it does a somersault. When it wants to make babies, it buds.
People, this is not a lifestyle that requires several thousand neurons to implement. I've written Javascript apps that do more interesting things.
Let's put this in perspective. Consider 1000 classic spiking neurons. Ignore subthreshold processing. Ignore the time and length constant integration of synaptic inputs. Ignore intrinsic and network spike patterning. Simply 1000 yes/no, go/no-go, spike/no-spike processing elements. Such a system generates 2^1000 unique states. A terabyte is 2^40. If we simply wanted to list the states, assuming one kilobyte per state, it would require ~2^970 of those fancy SSDs you have on your Amazon wish list. That's a flash drive the size of a Dyson sphere.
The hydra has at least twice as many neurons as that.
For all of Shakespeare's fawning praise of man's noble reason and infinite faculties, it sure takes quite the neuron horde to accomplish anything useful. Yes, there are notable exceptions -- the goldfish Mauthner cell implements a tail-flip reflex all by its lonesome -- but Yuste and Dupre's hydras clearly (wordplay!) illustrate that, on the whole, neurons are an embarrassment. A Brobdingnagian waste of processing. Brains are what you'd get if Rube Goldberg was the head of NASA. They make the Trabant look like Formula-1 engineering. They make the bubble sort, COBOL, and Windows Vista seem sleek and efficient. No wonder half the people you meet are below average intelligence.
I suppose it was possible to be offended long before Yuste and Dupre happened along. After all, neuron counting goes back at least as far as Golgi. But, as is often the case, a new result has made it impossible to ignore an inconvenient truth. There is a cusp. A transition. A Eureka moment. That is what all good science does.
Neurons suck. Not something you'll read in a textbook, grant proposal, or presidential BRAIN initiative, but true nonetheless.
Neuroscience might want to start looking into that.
No comments:
Post a Comment