The unbearable slowness of being …

The unbearable slowness of being: Why do we live at 10 bits/s?

This article is about the neural conundrum behind the slowness of human behavior. The information throughput of a human being is about 10 bits/s. In comparison, our sensory systems gather data at ∼10⁹ bits/s. The stark contrast between these numbers remains unexplained and touches on fundamental aspects of brain function: what neural substrate sets this speed limit on the pace of our existence? Why does the brain need billions of neurons to process 10 bits/s? Why can we only think about one thing at a time? The brain seems to operate in two distinct modes: the “outer” brain handles fast high-dimensional sensory and motor signals, whereas the “inner” brain processes the reduced few bits needed to control behavior.
Plausible explanations exist for the large neuron numbers in the outer brain, but not for the inner brain, and we propose new research directions to remedy this.

This articke is discussed on “Human Thought Lags Behind Sensory Speed“

“Every moment, we are extracting just 10 bits from the trillion that our senses are taking in and using those 10 to perceive the world around us and make decisions.
This raises a paradox: What is the brain doing to filter all of this information?”

“Human thinking can be seen as a form of navigation through a space of abstract concepts,”

---

‘‘OUTER BRAIN’’ VS. ‘‘INNER BRAIN’’

The discrepancy between peripheral processing and central cognition suggests that the brain operates in two distinct modes:
the outer brain is closely connected to the external world through sensory inputs and motor outputs. This is a realm of high dimensionality: many millions of sensory receptors and muscle fibers and extremely high information rates.
The inner brain, on the other hand, operates on a dramatically reduced data stream, filtered to the essential few bits that matter for behavior at any one moment. The challenge for the inner brain is to combine the animal’s goals with current inputs from the world and previous memories to make decisions and trigger new actions. The information rates are very low, but the processing must remain flexible because context and goals can shift at a moment’s notice.
A number of interesting research questions emerge regarding the relationship between the inner and outer brain.
First, how can the inner and the outer brain communicate? The slow inner brain listening to the sensory onslaught from the outer brain seems similar to ‘‘drinking from the Hoover Dam.’’ The rate of water flow through the Hoover Dam is ∼10⁸ times the rate of human drinking, the same ratio as between information rates in the outer vs. the inner brain. Presumably, the matching occurs in multiple steps, along some graded continuum of information rate. For example, at the output of the retina, the image
information has already been reduced by a factor of 10 or more,
leaving only image features that are useful in downstream processing. Given the large sifting number, there are many log units of visual compression that remain to be understood.
Second, what are the principles of neural function on the two sides of the interface? Taking the water flow analogy further: the design of the Hoover Dam turbines relies on rather different engineering principles from the design of a beer bottle. No one would consider making them of the same materials, given the enormous difference in performance. Yet, the brain seems to rely on the same materials throughout: neurons, synapses, and glia. The cerebral cortex is touted as the substrate for both rapid and parallel sensory processing (e.g., visual cortex) and for slow and serial cognition (e.g., the prefrontal cortex). Very similar neocortical circuits seem to be engaged in both modes. Are we missing some principles of brain design that categorically differentiate the functions of the inner and outer brain?

Comparing research reports from the two sides of the inner/outer brain divide can be a challenge, because the practitioners tend to operate on entirely different assumptions. In the sensory regions, tradition holds that ‘‘every spike is sacred’’ because each neuron has its own receptive field, and single spikes are known to convey a few bits of information each about the stimulus.
By contrast, researchers working closer to behavioral output, say in the prefrontal cortex or motor cortex, are often happy to boil the activity of millions of neurons down to just two or three ‘‘latent dimensions.’’ The typical finding identifies a lowdimensional manifold on which the neuronal population vector travels, following simple dynamics.

We see here that the paradoxical contrast between inner and outer brain information rates translates to an equally stark contrast regarding the presumed dimensionality of neural activity.
Are we to believe that the billion neurons in the primary visual cortex elaborate many channels of visual processing in 10,000 parallel modules, whereas the billion neurons in the prefrontal cortex just deal with a handful of slow variables, such as rules and values, associated with the single task currently at hand? It seems difficult to accept that two regions of the neocortex with overtly similar cell types and architecture (ignoring the minor differences) are organized to operate in such radically different ways.
Alternatively, the empirically observed difference in dimensionality may be an artifact of experimental design. Most of the studies of inner brain phenomena involve experimental tasks of very low complexity, such as a mouse making repeated binary choices between the same two stimuli or a monkey moving a lever along a few possible directions. Obviously, the resulting neural representations will be no more complex than what the
animal is doing, especially if one averages over many trials.
Under those conditions, one finds by necessity that some lowdimensional manifold accounts for much of neuronal dynamics.
However, real-life behavior is not that simple. For example, the Speed Card player, shuffling through the card deck, switches between different neural computations every few tenths of a second: every eye saccade brings a new card image in focus and triggers a bout of image recognition. These episodes are interleaved with storage in short-term memory and, again, with an update of the inner narrative or memory palace that can hold the new card, followed by another saccade. In conversations, we switch quickly between listening and talking, sometimes interleaved with a moment of thinking. When driving, we check the windshield, dashboard, rear mirror, and side mirrors and process the results in entirely different modes, such as estimating distance from the road edge vs. reading highway signs. The act of making tea requires 45 different brief subtasks.
Thus, we really perform thousands of different ‘‘microtasks’’ on any given day, switching between them as fast as we can saccade. Each of the microtasks depends on real-time feedback, requiring the sifting of sensory inputs and expansion into motor outputs within sub-second time frames. The flexible configuration and control of all these data streams seems essential to our cognitive functions. Perhaps the associated routing machinery accounts for the billions of neurons in the inner brain?
On this background, it would be informative to study the inner brain under such naturalistic conditions where it controls a rapid sequence of distinct microtasks. Without experimental designs tuned to the specialty of the inner brain, we might fail altogether to discover its essential mechanism.

In summary, we have a sense that major discoveries for a global understanding of brain function are waiting to be made by exploring this enormous contrast between the inner and outer brain.
We need to reconcile the ‘‘high-dimensional microcircuit’’ view of the outer brain with its ultimately low-rate information products. Vice versa, one may need to adopt a more high-dimensional view of the inner brain to account for the computations that happen there to organize behavior.