we're as smart as we're going to get - limits of human intelligence

How smart can humans get?  How much can our brains expand processing speed and computational power?  Can they get larger?

The animal kingdom has startling anomalies in "intelligence".  In an earlier blog, we saw that wasps recognize friends or enemies, and respond accordingly, being more aggressive to unfamiliar wasps.  They even use a specialized neural network for faces than they do for other objects.

Honeybees, with only milligrams of brain, navigate complex landscapes and mazes as well as "higher" mammals.   Elephants have a brain 5,000,000X as large as a honeybee's, but with that, the signal transit time across their brain is 100X longer.  Long transit times mean that neural reflexes are too slow to be effective, so elephants must dedicate enormous amounts of processing power to actually plan each step, ponderously.  Much brain capacity that could have been dedicated to "intelligence" is diverted to the motor and sensory cortex.   (You might be thinking "but insects aren't animals"...but they are.)

Gerhard Roth

These examples point to possible limitations on just how smart, fast, efficient and large the human brain can get.

Looking at just making the brain larger, scientists spent 100+ years mapping the masses of all animals and seeing how brain weight and "intelligence" changed as their mass increased.

Brain weight vs body weight

The results of these studies was presented in the frequently cited "Brain Size in Vertebrates" by van Donger in The Central Nervous System of Vertebrates,Vol 3, 1998, and is shown to the left.  A more recent work is "Evolution of the Brain and Intelligence" by Roth, G, and Dicke, U in Trends in Cognitive Neuroscience, 9, 5, May 2005.

This (log/log) plot, the slope of which demonstrates that brain size grows as the 3/4 power of body mass, shows that most fall close to the line.  Above the line, a species is "smarter than they should be", and if below the line, not so much.  Humans are "best" in this regard, although some small rodents and dolphins are close.

 Adding brain mass was found to increase intelligence only if increased brain mass wasn't diverted to sensory or motor capacities, as for elephants or cows.  Humans and elephants have virtually the same number of cortical neurons, but there is clearly a great difference in "intelligence".  A mouse is as smart as a cow but it has only 1% of the brain size.

Bigger "should" be better; more neurons means more neural pathways and more information being communicated per second.  However...energy consumption becomes a problem.  Our brains, which are 2% of our body weight consume 20% of our calories at rest, 80 % of which are consumed by cortical communication.  This energy consumption generates heat, which must be dissipated through cooling flows of blood.  Some evolutionary theories contend that larger, higher capacity, human brains could not develop until there were reliable, energy-rich food supplies available after the ice ages passed and agriculture developed.

In an effort like that on brain vs body weight, scientists analyzed neural structures.  Larger brains had larger neural cells which gave more interconnections.  However, with the increased size, when the cells are placed together, the distance between cells has to increase leading to longer connections to be bridged.

Jan Karbowski, a computational neuroscientist in "Thermodynamic Constraints on Neural Dimensions, Firing Rates, Brain Temperature and Size", J. Comput Neuroscience, 2009, 27, 415-436. says that "neurons do get larger as brain size increases, but not quite quickly enough to stay equally interconnected, which leads to longer conduction delays."   Larger size goes not into computational capacity, but into wiring, i.e. scaling up can't go on for long.

Olaf Sporns

  

Studies by Olaf Sporns, ("Mapping the Structual Core of Human Cerebral Cortex", Sporns, et. al. , PLos Biology, 6, 7, July 2008) have shown that to optimize this trade off, the inevitable result is a mix of short and long-range interconnections called "small world" networks.  In this model, small processing networks are interconnected through a few hubs by "superhighways", like modern cities.  As the brain grows, it gets increasingly specialized and regionalized, but interconnected.   This configuration minimizes the number of long, slow interconnections.

Primates, however, evolved a "way around" these limitations.  When their brains increase in size, their neurons do NOT increase in size, so high density packing arrangements can be used without requiring disproportionate amounts of wiring or a slowing in communication.  This turns out to be a HUGE advantage.  In studies on how directly brain regions talk to each other, fMRI and EEG measurements showed that those folk with shorter paths between brain regions had higher IQ and the best working memory.

So what's to limit our evolving smaller neurons and axons and packing them more densely into a "super brain".  Well, as the neurons get smaller, they reach the limit of the ability of cell membranes to handle the flow of chemicals involved, and they become unstable.  This results in switching back and forth randomly with increasing frequency and with much increased "circuit" noise.  It turns out that the size we have now is about an optimal balance between speed and acceptable reliability.

There is also the limitation on the size of the pelvises that evolved to support passage of a larger brain during birth.  Our bodies evolved over 5,000,000 years, and it is not conceivable that they will suddenly be able to successfully deliver babies pelvically with significantly larger brains.   Only if we went to non-pelvic births would that be possible, and there are complications with multiple births by that route.

So good new, bad news, really good news.  The good news is that humans have developed an almost optimal combination of brain mass to body mass, and abstract an optimal "intelligence" from that mass, by optimally down-sizing neurons and creating the best-possible cortical communications network.  The energy consumption and heat dissipation are also near their maximum trade off.

The bad news is that we are unlikely to be able to evolve a significantly larger intelligence within these bodies.  The really good news is that we have offloaded many routine tasks (memory, calculations, data retrieval, sorting, etc.) to a huge, shared, network/"cloud", which does those tasks better than we do.  That has freed up our intelligence for what it does uniquely well.  The internet has come along at just the right time.

Scientific American July 2001 has some useful diagrams on some of this.