Intelligence: An abstract approach

Scientists are often depicted as researchers with magnifying glasses. A prism would be a much more relevant tool, in my opinion. As a complex system, the brain can be described on many different levels: electron movement, chemical reactions, protein functionality, and cell activity.


Definitions and motivation

The abstraction layer is the layer of a system that provides a simplified view of the underlying complexity. Understanding abstraction is as the core of the hacker mindset. Software reverse engineering is a subdiscipline of software engineering. It strives to provide support for the comprehension of software systems by creating suitable representations of the system in another form or higher level of abstraction. When studying a complex multi-layer object such as the brain it is crucial not to confuse different layers of description. If we use too high a level, we might be left with a set of undefined concepts like feelings, that are simple to reason about but not quantify. Using too low a level, like the activity of a each and every single cell will introduce complexity that we might be able to measure but not reason about.

The brain’s abstraction layers

Only the brain has its own research discipline. This is due to its ability to process information, identify patterns and causal relations, guide complex behavior, produce thinking, creativity and consciousness. It’s possible to break the brain down into layers: physics, chemistry, biology, and psychology. What would be the right abstraction layer to focus on? Something is missing! A hint would be that computation is at the heart of the abilities mentioned above.

A note on computation

Without spoiling the next post on computation, a brief introduction is neccessary:

Computation is the transformation of sequences of symbols according to precise rules. - Konrad Hinsen, What is computation?

  • It might seem suprising that this defintion has nothing to do with the specific of a computer, and that it is that general. The strength of computer-science is the ability to study and reason about such general framework.
  • It is also clear and agreed upon that the brain is involved in computation, like any machine it encodes, stores and processes information.

The biological lens and computation

The majority of neuroscientists today approach their research from a biological perspective only or might only acknowledge the visible aspects of the brain. They seek to understand “how neural activity leads to behavior?”, or “what is the neural basis of consciousness?” by trying to skip the computational layer. I find these questions puzzling, since they are similar to asking “how subatomic particles led to the French Revolution” or “how each bit in RAM relates to software GUI”. It would make more sense to split each question in two: how biology is able to implement a computational machine, and than later how this machine is able to guide behavior or give rise to consciousness.

Could we understand biological computations without biology?
It might seem that biology is crucial for understanding how information processing in the brain works. An analogy that is used often is of a cloud, whose properties are derived from the electrical/chemical properties of each water drop. By combining these drops, some new properties emerge. A single drop of water must be studied closely in order to understand what emerges and how it occurs.

In contrast, this analogy is similar to a magnifying glass which only explores what is visible and is incompatible with what we know about computation. Computer science teaches us that algorithms are completely separate from implementations. In the next post we will further discuss computation…


Dougles Hofstadter (Godel, Escher, Bach author) makes a similar argument against the biological monopoly over brain studies:

“Analogously, a brain is a thinking machine, and if we’re interested in understanding what thinking is, we don’t want to focus on the trees (or their leaves!) at the expense of the forest. The big picture will become clear only when we focus on the brain’s large-scale architecture, rather than doing ever more fine-grained analyses of its building blocks.” (“I am a Strange Loop”, Douglas Hofstader)

“I HAVE often been asked, when people hear that my research amounts to a quest after the hidden machinery of human thought, “Oh, so that means that you study the brain?” One part of me wants to reply, “No, no — I think about thinking. I think about how concepts and words are related, what ‘thinking in French’ is, what underlies slips of the tongue and other types of errors, how one event effortlessly reminds us of another, how we recognize written letters and words, how we understand sloppily spoken, slurred, slangy speech, how we toss off untold numbers of utterly bland-seeming yet never-beforemade analogies and occasionally come up with sparklingly original ones, how each of our concepts grows in subtlety and fluidity over our lifetime, and so forth. I don’t think in the least about the brain — I leave the wet, messy, tangled web of the brain to the neurophysiologists.” Another part of me, however, wants to reply, “Of course I think about the human brain. By definition, I think about the brain, since the human brain is precisely the machinery that carries out human thinking.” This amusing contradiction has forced me to ask myself, “What do I mean, and what do other people mean, by ‘brain research’?”, and this leads naturally to the question, “What are the structures in the brain that someone could in principle study?” Most neuroscientists, if they were asked such a question, would make a list that would include (at least some of ) the following items (listed roughly in order of physical size): amino acids neurotransmitters DNA and RNA synapses dendrites neurons Hebbian neural assemblies columns in the visual cortex area 19 of the visual cortex the entire visual cortex the left hemisphere Although these are all legitimate and important objects of neurological study, to me this list betrays a limited point of view. Saying that studying the brain is limited to the study of physical entities such as these would be like saying that literary criticism must focus on paper and bookbinding, ink and its chemistry, page sizes and margin widths, typefaces and paragraph lengths, and so forth. But what about the high abstractions that are the heart of literature — plot and character, style and point of view, irony and humor, allusion and metaphor, empathy and distance, and so on? Where did these crucial essences disappear in the list of topics for literary critics? My point is simple: abstractions are central, whether in the study of literature or in the study of the brain. Accordingly, I herewith propose a list of abstractions that “researchers of the brain” should be just as concerned with: the concept “dog” the associative link between the concepts “dog” and “bark” object files (as proposed by Anne Treisman) frames (as proposed by Marvin Minsky) memory organization packets (as proposed by Roger Schank) long-term memory and short-term memory episodic memory and melodic memory analogical bridges (as proposed by my own research group) mental spaces (as proposed by Gilles Fauconnier) memes (as proposed by Richard Dawkins) the ego, id, and superego (as proposed by Sigmund Freud) the grammar of one’s native language sense of humor “I”. I could extend this list arbitrarily. It is merely suggestive, intended to convey my thesis that the term “brain structure” should include items of this general sort. It goes without saying that some of the above-listed theoretical notions are unlikely to have lasting validity, while others may be increasingly confirmed by various types of research. Just as the notion of “gene” as an invisible entity that enabled the passing-on of traits from parents to progeny was proposed and studied scientifically long before any physical object could be identified as an actual carrier of such traits, and just as the notion of “atoms” as the building blocks of all physical objects was proposed and studied scientifically long before individual atoms were isolated and internally probed, so any of the notions listed above might legitimately be considered as invisible structures for brain researchers to try to pinpoint physically in the human brain. Although I’m convinced that finding the exact physical incarnation of any such structure in “the human brain” (is there only one?) would be an amazing stride forward, I nonetheless don’t see why physical mapping should constitute the be-all and end-all of neurological inquiry. Why couldn’t the establishment of various sorts of precise relationships among the above-listed kinds of entities, prior to (or after) physical identification, be just as validly considered brain research? This is how scientific research on genes and atoms went on for many decades before genes and atoms were confirmed as physical objects and their inner structure was probed.” (“I am a Strange Loop”, Douglas Hofstader)

Written on September 5, 2022