[LINK] “The Limits of Intelligence”
Article in current Scientific American (first para and bullet points, rest is paywalled).
Podcast by the author (free).
The author, Douglas Fox, argues that there may be physical limits to how intelligent a brain made of neurons can become, limits that may not be very distant from where we are now.
He makes evolutionary arguments at a couple of points, suggesting that he is talking about how smart an organism could have evolved, rather than how smart we might make ourselves; he certainly isn’t talking about how smart a machine we might create out of different materials.
From the podcast (I don’t have access to the article):
Four routes to higher intelligence, which he argues won’t get us very far:
Increase the speed of axons. But that means making them fatter, which drives the neurons further apart, neutralising the gain.
Increase brain size. That needs more energy, and before long you get something unsustainable. You get longer pathways in a larger brain, which slows them down. The neurons will make more connections, which makes them bigger, so the number of neurons scales slower than the volume of the brain. And anyway, whales and elephants have bigger brains than us but don’t seem to be more intelligent, and cows have brains a hundred times the size of a mouse brain but aren’t a hundred times smarter. So brain size doesn’t seem to matter; at best the relationship with intelligence is unclear.
Packing more neurons into the existing volume by making them smaller. You run into signal to noise problems. The ion channels involved in generating action potentials are a certain size, and you must have fewer in a smaller neuron, hence more random variation. The result is neurons spontaneously firing.
Offload intelligence support. Books and the internet will remember things for you and help you tap the collective intelligence of your social network. Compare social insects doing things that they couldn’t do individually. But by alleviating the necessity of intelligence this may even have reduced the evolutionary pressure to get smarter.
He’s described simply as an “award-winning author”, but I don’t know if he has any scientific background, and there are too many people of the same name to Google him.
Even if John von Neumann is the smartest thing that will ever exist getting a few million of him would have a civilization-changing impact.
I can’t access the article and I don’t have time for a podcast, but the obvious question is “Why wouldn’t all these same reasons apply to a chimp?”
There’s pretty good evidence that there are alternate brain designs which can at least gain us 10 or 20 IQ points. There’s a gene for promoting nerve growth (can’t remember the name) which is associated with a higher IQ and a higher rate of cancer; in most people it’s been selected against because of the cancer risk, but that suggests a very different reason why our brains are limited to their current power than Fox brings up. Gaucher’s Disease is another deadly condition linked to genes which can increase IQ in people who don’t have the disease (and increase it even more in people who have the disease but live with it).
So I would agree with him that there are probably biological reasons our IQ isn’t higher, but I think they’re more likely to involve disease than neural architecture, and that we can probably evolve ways around those diseases pretty quickly on an evolutionary scale.
I interpret the brain size data differently. That a cow is not smarter than a mouse shows that most of brain size is used up controlling a larger body. That a whale has a large brain shows that large brains work. Animals with slightly larger brains than predicted by body size are smarter (apes, dolphins, ravens). This can probably be pushed further.
(I don’t understand why a large body should require extra neurons. One explanation is that resolution of touch is independent of size. Is this true? Why would a large animal need fine resolution? This would only explain a 2⁄3 exponent, not the observed 3⁄4. Neurons for controlling muscle probably scale at a full rate, but they needn’t clog up the brain.)
It looks like having more brainpower controlling muscles is all sorts of useful, for example determining how finely you can do fine motor control.
So let me rephrase my question in terms of output: do large animals have the same resolution of muscular control as small animals? Or do they have coarser control? If I’m 100x as tall as a mouse, do I have 100x as much control over the angle of my elbow? What good would that be? Can I pick up as small objects as a mouse? Why would I want to?
ETA: and even if I could, I think that would only explain a scaling exponent of 1⁄3, not the observed 3⁄4.
Of course you can pick up objects as small as a mouse. How do you operate your computer otherwise? :D
I often find it useful to finely control how much force I apply, e.g. threading a needle or walking barefoot on rocks. I don’t know that my control over how much force I apply has to be as fine as a mouse’s, but pretty near.
Along those lines, I think the bigger brain is needed to move bigger limbs around, more electric power to signal the muscles but I haven’t been able to find a citation.
I think that’s what Douglas_Knight was getting at with the “controlling muscle… needn’t clog up the brain” comment. If the trouble is that muscles need a amplified electrical signal, why not send data from a small efficient brain down small efficient nerves then amplify those signals right next to the muscle?
There’s probably some design constraint we don’t see, though. Whales have much bigger brains than humans, but those brains seem to be composed of fewer neurons and far far more glial cells...
Google to the rescue: http://www.douglasfox.org/about/
He did study biology, then became a science writer for magazines like Discover.