Tech entrepreneur Max Bennett is too young to have grown up on The Jetsons, a 1960s futuristic cartoon series based on the more popular stone-age Flintstones (both, in turn, cartoon adaptations of the 1950s The Honeymooners). But Jetsons reruns, which featured smart watches, cell phones, and video calls, made an impression on the budding tech entrepreneur.
One gadget in particular caught Bennett’s attention with its twenty-first century absence: Rosie, an autonomous robot human enough to help with babysitting, homework, and cooking. Moreover, Rosie possessed charm, emotional intelligence, and a sharp sense of humor. Why, Bennett wondered, do even today’s most impressive AI tools fall short of Rosie’s human-like chops, stumbling over seemingly easy intellectual tasks or loading a dishwasher?
Bennett reasoned that just as early aeronautical inventors sought inspiration from birds’ flight mechanisms, surely AI scientists can gain insight by studying the human brain, the result of which is A Brief History of Intelligence.
Spoiler alert: Bennett fails. Just as the flight characteristics of modern flying machines have little in common with those of birds, so too will the mechanisms of successful artificial intelligence not resemble those of our three pounds of gelatinous brain tissue, with its nearly hundred billion neurons and hundred trillion connections, and whose underlying electronics in no way resemble those of silicon-based semiconductors.
If that were not enough, Bennett has little formal training in the biological or neurosciences. Why, then read his book? Because the author is a world-class autodidact whose abilities have attracted the notice of luminaries such as renown neuroscientist Joseph LeDoux and the late Nobel-winning psychologist Daniel Kahneman. Because his intellectual syntheses have been published in high-level peer-reviewed venues, and because the book sits at, or at least near, the apex of the how-we-got-here genre of human evolution narratives, on a par with Joseph Henrich’s The Secret of Our Success. His expositions of basic neuroanatomy and neurophysiology were crisp and on-target enough to induce PTSD flashbacks from this reviewer’s first year in medical school.
A Brief History of Intelligence, then, delivers exactly what its title advertises—a wide-angle perspective on how organisms have evolved to think, starting with the first self-replicating molecules that arose out of the early planet’s primordial soup four billion years ago. Bennett delineates five breakthroughs that got us to where we are, starting almost from the moment the Earth first started to cool:
1) Steering: The first animals were radially symmetric—think jellyfish, or, for that matter, an early human embryo, organisms that navigate only very slowly, if at all. Approximately six hundred million years ago, the first simple bilaterally symmetric animals (“bilaterians”—think a microscopic worm) evolved, under the direction of only a few sensors and neurons, the ability to navigate quickly towards food and away from predators.
2) Reinforcement: Shortly thereafter the first vertebrates began to learn from experience, what AI researchers call “model-free reinforcement learning,” the ability to modify future behavior from trial and error.
3) Simulation: Around 100 million years ago, the first mammals evolved the ability to map the external world to plan future actions.
4) Mentalizing: Approximately 20 million years ago, our primate ancestors acquired the ability to analyze their own mental states, which enabled them to anticipate their own future needs and, more importantly, the needs of their peers: A chimpanzee may not be hungry in the moment, but he will understand that tomorrow he will be hungry and should avail himself today of the bananas in a nearby tree. Mentalizing also allows primates to understand the needs of others and to imitate, and thus convey, cultural information, prime among which are hunting and toolmaking. Critically, mentalizing breaks the link between physical superiority and reproductive success: The mentalizing ability of a politically gifted male chimpanzee (or human) enables him to dominate bigger and stronger male competitors for mates.
5) Language: While many animals can communicate information through chirps and shrieks—“predator near!” “food here!”—such vocalizations are almost always hardwired and stereotyped. Approximately five million years ago, our hominid ancestors developed both the cognitive ability and anatomical hardware (primarily the descent of the larynx down the airway) to formulate and convey complex abstract information in a way unique to the human species.
Exactly what use is all this neurophysiology, neuropsychology, and evolutionary history to financial practitioners? Or, for that matter, what is their relevance to artificial intelligence?
Start with how electronic and biological systems differ in the conveyance of information. Analog electronic systems depend on the flow of electrons through efficient metallic conductors, like copper, while digital ones periodically interrupt these flows through less conductive materials—semiconductors.
Neurons, in contrast, convey information through their firing rate, from a resting rate of roughly once per second up to a maximum of about three hundred times per second. Now, consider the human visual system, which can perceive minor changes in light intensity from dim moonlight all the way to bright sunshine, a luminosity range of 11 orders of magnitude. How, then, do we handle light intensity over 11 orders of magnitude with neuronal firing-rate hardware that can only handle 2-3 orders of magnitude?
With adaptation, a neuronal property present in even the most primitive bilaterians, and is presumably hundreds of millions of years old. When our visual systems are exposed to bright light, the neuronal firing rate in the visual cortex gradually falls off as we adapt to it; conversely, the baseline firing rate rises after prolonged exposure to dim light. As we proceed from bright sunlight to a dark room, the neuronal firing rate manages to stay within a range much smaller than that of the change in light stimulus.
It’s a short conceptual throw from neuronal adaptation to psychological adaptation, which pervades our financial lives. The thrill of owning that first beamer fades within months; upgrade your air travel from economy to business class, and after a few years you yearn to fly private—the dreaded “hedonic treadmill.”
Albert Einstein supposedly said that compound interest is the most powerful force in the universe. (He didn’t.) Thanks to neuronal adaptation, however, amnesia is the most powerful force in the financial universe. During the late 1990s, after nearly a decade of stock market froth, participants came to see double digit returns as normal. Contrariwise, for a generation after the 1929 stock market crash, “prudent” investors avoided stocks. More recently, after years of negative real interest rates, bond market participants came to see them as normal, until much to their chagrin in 2022, they weren’t. Among the many accomplishments of Amos Tversky and Daniel Kahneman was the realization that this amnesia/recency illusion was a subset of a much more general psychological malfunction, the “availability heuristic,” which likely also depends on neuronal adaptation.
The book’s penultimate chapter provides the reader with a concise and readable summary of the altruism/cheating nexus that pervades human behavior in general, and the financial markets in particular. Human beings cannot run fast, fly, and don’t have deadly teeth or great strength; our late Pleistocene ancestors would have become the lunch of physically more capable predators but for one characteristic: our ability to cooperate through language. (Humans are also possessed of two extraordinary physical abilities: most famously, manipulation with opposable thumbs and, less obviously, a throwing ability unmatched in the animal kingdom.)
Our language ability enables altruistic behavior which is beneficial at the tribal level: the sharing of food and sacrifice in battle. Altruism, though, competes against cheating, which confers a survival advantage to the individual cheater. Evolutionary biologist Robin Dunbar points out that humans engage in one form of verbal communication above all others: gossip, which tamps down cheating by establishing an individual’s faithfulness to tribal norms. Without putting too fine a point on it, individuals do not go into finance for the same reasons that they become marines, Jesuit priests, or elementary school teachers: In large-scale financial markets, gossip—that is reputation—is no longer adequate to maintain honesty.
The book’s last chapter entertains with the comical failures of GPT-3, such as the completion of the sentence “He threw the baseball 100 feet above my head, I reached my hand up to catch it, jumped . .” with “and caught it. It was a lot of fun!” Similarly, GPT-3 made the same mistake as humans do with Shane Frederick’s famous conundrum of a baseball and bat costing $1.10 together, with a $1.00 difference between the two. Most people answer incorrectly that they cost, respectively, $0.10 and $1.10, when the correct answer is $0.05 and $1.05.
Programmers trained GPT-4 to break down the problem into its algebraic parts, but no sooner was this deficiency fixed than others popped up in a never-ending whack-a-mole process. Bennett cites Stephen Pinker’s famous hypothetical exchange as one that AI may never be able to correctly parse:
Bob: I’m leaving you. Alice: Who is she?
A Brief History of Intelligence supplies readers at all levels of scientific knowledge with valuable insights, provides the lay person with an accessible explanation of brain function, and even schools the clinical neurologist with a detailed exposition of the relationship between small variations in otherwise nearly perfectly uniform microscopic cortical architecture and the ability to mentalize; the bibliography and reference notes were voluminous to require their own separate online repository.
This reviewer found the book to be nearly flawless, from its prose quality to its conceptual exposition, the perfect Christmas or birthday gift for both the educated general reader and the financial, bioscience, or IT specialist.
A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out some of our webcasts.
More Fixed Income Topics >
Tech entrepreneur Max Bennett is too young to have grown up on The Jetsons, a 1960s futuristic cartoon series based on the more popular stone-age Flintstones (both, in turn, cartoon adaptations of the 1950s The Honeymooners). But Jetsons reruns, which featured smart watches, cell phones, and video calls, made an impression on the budding tech entrepreneur.