Jeff Lichtman
About
Neuroscientist, podcast guest (episode 298)
Cast within
No topic-region cast yet — this appears once Jeff Lichtman's compiled claims are aligned into a topic region's argument tree.
Claims by Jeff Lichtman (20 of 36)
Brain may be irreducibly complex with no simplification
Understanding means having a compressed shorthand version of complexity such that you no longer need the details; but the brain may be a case where there is no simplification—it is its own most concise description—because if a simpler version were possible, the brain would have evolved to be simpler.
Convolutional neural nets derive from Hubel and Wiesel
Modern convolutional neural networks are based on the original studies of Nobel laureates Hubel and Wiesel on how visual information of the world is broken down into very small features in the visual cortex and then reassembled to produce perception.
Neuron as a selfish single-celled organism wanting to survive
A neuron can be understood as a single-celled organism with its own will to survive, living in the 'weird pond' of your head, doing everything for its own survival; learned wiring diagrams emerge from neurons doing what keeps them alive, and humans are simply the embodiment of all their neurons—we eat chocolate chip cookies because our neurons want glucose.
Consciousness may be a slippery slope down to every cell
If consciousness is defined as an organism responding to its environment, then every cell—even an amoeba moving away from heat—is conscious; this makes the concept slippery and possibly a linguistic problem rather than a brain problem, much as wave-particle duality is a problem only for human language and not for the photon.
Neuroscientists are physicalists about the mind
There is no magic in the brain beyond enormous complexity—most working neuroscientists, including Lichtman, are physicalists who do not think understanding a human requires anything beyond the underlying physical stuff, though what 'understand' means remains genuinely unclear.
Brain is sensory-in, motor-out processing
The fundamental purpose of the nervous system is sensory-in, motor-out: it takes information from sense organs, processes it through interconnected neurons, and turns it into a motor reaction; humans add the ability to act on stored, learned information through rumination even absent external stimulation.
Neurons fire based on threshold relative to noise
A neuron has a threshold and fires only when input exceeds the background noise at a given moment—analogous to how a small noise gets attention at night but not during a loud day—and salient signals can be amplified and propagated to all of the neuron's target cells.
Brain size relative to body, not absolute, matters
Humans are smart not because we have the largest brains—elephants and whales have much bigger brains—but because we are highly encephalized (large brain relative to body size) and have a large amount of association cortex, more than any other animal.
Neurons encode signal strength via AM/FM modulation
Neurons differ in their firing behavior—some fire a single action potential, others fire bursts whose frequency encodes signal strength—so input amplitude (AM) is converted into output frequency (FM), and these responses are governed by specific membrane ion channels making the system extremely nonlinear and hard to model.
Learning changes both wiring and synaptic strength
Learned behaviors in mammals and humans come from experience rather than genetic programming, and this works through both changes in the wiring diagram (new connections) and changes in the sensitivity/strength of existing synapses, since evolution exploits any mechanism that is useful.
Cajal inferred brain circuits from sparse Golgi stain
Modern neuroscience began with Ramón y Cajal, who used Golgi's stain that randomly labeled only about 1% of nerve cells; the random sparseness allowed him to see the complete connectivity of individual cells and infer that the brain is made of directional circuits with input on dendrites and output via axons.
Real wiring is infinitely more complex than stick figures
The classic Cajalian stick-figure diagrams of neuron connectivity, based on sparse labeling, led people to assume that was the brain; but when you actually image everything, the wiring is infinitely more complicated than the stick figures suggested.
Nervous systems self-organize with built-in chance variability
In both worms and mammals, the nervous system self-organizes to work but leaves a certain amount to chance—mouse muscle innervation shows the same rule-based range of axon sizes across animals, but the specific location each axon goes to is variable—because anything that mattered for survival would have been highly constrained, so variability marks what doesn't matter.
Brainbow labels every cell a different color
The brainbow technique solves the limitation of the sparse Golgi stain by labeling every neuron a different color, so densely-packed cells can be individually traced; however, in the brain itself the wires are so densely packed that diffraction-limited fluorescence microscopy lacks the resolution to trace them, useful mainly in the peripheral nervous system.
Whole mouse brain would be an exabyte of data
A consortium of labs, with Google's help, is building tools for a proof-of-principle whole-mouse-brain connectome of about 1000 petabytes (a million terabytes, an exabyte); they could start within about five years and finish imaging within about another five years.
Multi-beam electron microscopes accelerate imaging
A very fast multi-beam electron microscope built by Carl Zeiss expressly for connectomics scans multiple sample regions simultaneously, dramatically speeding imaging; these multi-million-dollar (about $5 million) machines did not previously exist, and only about 10-15 exist in the world, with Lichtman's lab having the first.
My Notes
Loading notes...