{"id":1,"url":"https://pm.philipcastiglione.com/papers/1.json","title":"The remarkable, yet not extraordinary, human brain as a scaled-up primate brain and its associated cost","read":true,"authors":"Suzana Herculano-Houzel","year":2012,"auto_summary":"The paper \"The remarkable, yet not extraordinary, human brain as a scaled-up primate brain and its associated cost\" by Suzana Herculano-Houzel challenges several widely-held beliefs about the human brain's uniqueness. Traditionally, the human brain has been considered extraordinary due to its supposed 100 billion neurons, high glia-to-neuron ratio, and significant energy consumption relative to its size. These beliefs have contributed to the notion that the human brain is an outlier in evolutionary terms, explaining our superior cognitive abilities compared to other animals with larger brains.\n\nHerculano-Houzel argues that these assumptions are not well-supported by empirical evidence. Using a method called the isotropic fractionator, she and her colleagues have provided more accurate counts of neurons and nonneuronal cells in various species' brains. Their findings suggest that the human brain, with approximately 86 billion neurons and a similar number of nonneuronal cells, is not an outlier but rather a scaled-up version of a primate brain. The human brain's cellular composition and metabolic cost align with what would be expected for a primate brain of its size, rather than being exceptional.\n\nThe paper explores the implications of these findings, particularly the relationship between brain size, neuron numbers, and cognitive abilities. It challenges the idea that encephalization quotient (brain size relative to body size) is the primary determinant of cognitive ability, suggesting instead that absolute numbers of neurons and their connectivity might be more critical. The study shows that the human brain's cognitive abilities may be remarkable simply due to its large number of neurons, a result of the particular scaling rules that apply to primates.\n\nHerculano-Houzel also examines the metabolic costs associated with brain size and neuron numbers, proposing that the human brain's high energy consumption is a direct consequence of its large number of neurons. This challenges previous theories that brain size is a limiting factor in evolution, suggesting instead that the number of neurons imposes a more significant metabolic constraint.\n\nThe paper concludes that while the human brain is remarkable, it is not extraordinary in the context of evolution. Its characteristics can be understood as part of the broader pattern of primate brain evolution, rather than as unique deviations. This perspective invites a reconsideration of how we view human cognitive abilities in the context of evolutionary biology.","notes":{"id":1,"name":"notes","body":"\u003ch1\u003eNotes\u003c/h1\u003e\u003cdiv\u003e\u003cbr\u003eHuman brain averages ~86bn neurons \u0026amp; approx the same nonneuronal brain cells.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eOur brains are not some crazy outlier on any core metric:\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cul\u003e\u003cli\u003eThere are animals with larger brains, though humans are large: elephant (2-3x), some cetaceans (4-6x).\u003c/li\u003e\u003cli\u003eThere are animals with higher encephalisation quotients (brain mass to body mass ratio), though humans is high.\u003c/li\u003e\u003cli\u003eThere are animals with larger absolute size of cerebral cortex.\u003c/li\u003e\u003cli\u003eThere are animals with greater gyrification (surface area - wrinkling).\u003c/li\u003e\u003cli\u003eThe human cerebral cortex is the largest in relative size amongst mammals, but not by that much. A horse is pretty close.\u003c/li\u003e\u003cli\u003eOn limited evidence (2012) connectivity (synaptic density) is pretty consistent amongst species\u003c/li\u003e\u003c/ul\u003e\u003cdiv\u003e\u003cbr\u003eThere can be significant differences between the brain “plan” of animals from different orders (eg. cow vs primate vs whale). Primate neuronal density is greater for the same brain volume.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003ePrimate “scaling rules” eg. neuronal density and distribution, are more efficient than for other mammals. For a rodent brain to gain 10x the neurons, it must gain 35x total volume, whereas a primate must only gain 10x-11x.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eNonneuronal brain cells have more or less the same scaling rules between species. Nonneuronal density does not differ systematically with structure size.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eCerebral cortex and cerebellum scaling actually follow some consistent rules; even though cerebral cortex volume scales faster, neuronal density decreases such that core ratios are maintained.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cstrong\u003e\u003cbr\u003eThe paper is fundamentally arguing that human brains are just a big, scaled up brain on the standard primate plan. Not outlier/unique in any other special respect.\u003cbr\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eNetwork structure:\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cul\u003e\u003cli\u003edense local connections\u003c/li\u003e\u003cli\u003esparse long range connections\u0026nbsp;\u003cul\u003e\u003cli\u003eonly a proportion of neurons connect through/into the white matter\u003c/li\u003e\u003c/ul\u003e\u003c/li\u003e\u003c/ul\u003e\u003cdiv\u003e\u003cbr\u003eOn the primate plan, brain-body scaling relationship is very dependent on species, but neuronal scaling rules are consistent between species.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003ch1\u003eQuestions\u003c/h1\u003e\u003cdiv\u003e\u003cbr\u003eWhy is gyrification good? Is it? Or is it a byproduct of something good.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eWhat differences are there between the brain of a child and that of an adult? Obvious: experiences. What else? What about on the metrics? Presumably encephalisation quotient in children is higher (”better”). What does this tell us? What changes and what doesn’t during maturation? What does this tell us?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eHow much variation is there on core metrics between different individual humans? What effect does that variation have?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eHow does this pan out (re: encephalisation quotient)? “you need X neurons dedicated to sensorimotor function, Y to perception, Z to … . Then, a higher amount of \u0026lt;cerebellum? cerebral cortex? …\u0026gt; provides for … emergent intelligence? Given scale?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eWhat are the different core characteristics that might lead to intelligence?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cul\u003e\u003cli\u003eNumber of neurons\u0026nbsp;\u003cul\u003e\u003cli\u003eby area\u003c/li\u003e\u003cli\u003eby type? What are these?\u003c/li\u003e\u003c/ul\u003e\u003c/li\u003e\u003cli\u003eNonneuronal cells\u0026nbsp;\u003cul\u003e\u003cli\u003ewhat do they do?\u003c/li\u003e\u003c/ul\u003e\u003c/li\u003e\u003cli\u003econnectivity\u003c/li\u003e\u003cli\u003elatency\u0026nbsp;\u003cul\u003e\u003cli\u003ecommunications methods? patterns?\u003c/li\u003e\u003c/ul\u003e\u003c/li\u003e\u003cli\u003eHow does storage work?\u003c/li\u003e\u003cli\u003e…\u003c/li\u003e\u003c/ul\u003e\u003cdiv\u003e\u003cbr\u003eWould network theory be useful for understanding brain communication patterns?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eSince human neuronal/nonneuronal (G(lial)/N(euron) ratio) characteristics are consistent with other species, can we rule this out as being an important part of difference in the plan?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eSynaptic homeostasis and elimination of excess synapses “(eg during sleep)” are postulated as both a) the basis of plasticity and also a necessary tradeoff to manage the cost of neuronal energetic expenditure.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cul\u003e\u003cli\u003eWhat is synaptic homeostasis?\u003c/li\u003e\u003cli\u003eElimination of excess eg during sleep is pruning synapses? How does that work?\u003c/li\u003e\u003c/ul\u003e\u003cdiv\u003e\u003cbr\u003eHow smart are corvids? Octopi? They’re little.\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eBrains are energetically expensive, which creates evolutionary friction. But energy is cheap for us now… this should be a large advantage for AI right?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cdiv\u003e\u003cbr\u003eData received by human perception is actually pretty rapid. And it takes years to get smart. But, since a) the data humans receive is not well structured for broad intelligence as a goal and b) our ability to store, curate and manage data scales so well, this should be a large advantage for AI right?\u003cbr\u003e\u003cbr\u003e\u003c/div\u003e\u003cul\u003e\u003cli\u003eWhat are relative bandwidths for brain senses (is this the right way to think about it?)\u003c/li\u003e\u003cli\u003eDigital equivalents?\u003c/li\u003e\u003c/ul\u003e\u003ch1\u003eTakeaways\u003c/h1\u003e\u003col\u003e\u003cli\u003eWith the right plan, intelligence emerges with scale (and experiences/training data)\u0026nbsp;\u003col\u003e\u003cli\u003eThe right plan makes a big difference! Whales have much larger brains and live much longer.\u0026nbsp;\u003col\u003e\u003cli\u003eAssuming LLMs + more scale → AGI/ASI misses this point\u003c/li\u003e\u003c/ol\u003e\u003c/li\u003e\u003cli\u003eThe amount of scale matters a lot\u003c/li\u003e\u003cli\u003eThe amount of training data/experience matters a lot\u003c/li\u003e\u003c/ol\u003e\u003c/li\u003e\u003cli\u003eYou should be able to see/measure/expect progress along the way. With the right conditions, a \u0026lt;monkey\u0026gt; demonstrates intelligence on some tasks. More so than a \u0026lt;less scaled monkey\u0026gt;.\u0026nbsp;\u003col\u003e\u003cli\u003ethe tests matter\u003c/li\u003e\u003c/ol\u003e\u003c/li\u003e\u003c/ol\u003e\u003cdiv\u003e\u003cbr\u003e\u003c/div\u003e","record_type":"Paper","record_id":1,"created_at":"2024-12-10T02:55:13.113Z","updated_at":"2024-12-10T03:52:46.480Z"},"created_at":"2024-12-10T02:54:11.031Z","updated_at":"2024-12-10T03:52:46.481Z"}