Toward a History of Intelligence
Intelligence is not a possession
AI engineers and researchers are placing the definition of intelligence on the table in a way that goes beyond specialized philosophical speculation. The speed and scale of what they are building operationalizes their definitions in such a way that they have real consequences for how we live.
Because the engineering efforts drive the definitions, intelligence will reflect what we are able to build. While what we are able to build is and will continue to be powerful, remarkable, and potentially era defining, leaving definitions of intelligence to engineering concerns comes with a number of blindspots.
First, it will tend to treat intelligence as a possession rather than a distributed capacity. Engineers build things, and they build things that can be owned as intellectual property (IP). This IP has measurable value and can be bought and sold just like any other commodity. Insofar as we are calling these possessions ‘intelligence’, we are treating intelligence as a thing that can be possessed, commodified, and run on Turing-complete machines, which themselves appear in the world as purchasable commodities.
Second, treated as a possession, ontological questions will dominate the public discussion. ‘What is intelligence?’ and its surrogates—what is consciousness, what is life—already dominate our public discourse about AI. How we answer the ‘what is’ questions will be shaped by the desire to build it as the possession of a machine.
This leads us to the third risk of treating intelligence as an engineering objective. We distract ourselves from attending to what these machines do in order to debate whether they are intelligent, conscious, or a form of life. These are interesting questions, but they feel more like the Scholasticism of the Middle Ages—questions of essences—rather than the Enlightenment’s focus on the practical effects of our knowledge.
Intelligence is time before it is a thing
What if intelligence isn’t reducible to a possession? What if it is more like weather—an emergent event distributed across a broad range of motions rather than a thing reducible to a set of programmable features?
Time as Practice offers as a conjecture a broad definition of intelligence: the adaptive and expansive capacity to make the future less like fate and more like an open field of possibilities. This definition pushes us to attend to effects, not essences. As such, it sees intelligence, like all phenomena in the Universe, as temporal before it coalesces into a definable thing.
All things in the Universe must come into being as assemblies of what the Universe has made possible. But not all things are possible at all times. It takes time for the Universe to produce the Earth, and for life to emerge on the Earth, and for bacterial life to become the Cambrian Explosion, and so on. Getting to Homo sapiens doesn’t happen until roughly 200,000 years ago of a 4.8 billion year old Earth. While all of these things are made possible by the Universe and theoretically exist universally, they are not all possible all at once. Some events have to proceed others. Lineages need to form and combine, and as they do so, they make other possibilities less likely and therefore less possible.
Nor does this make any assembly inevitable. Assemblies are contingency all the way down and all the way back up. But as contingency becomes reproducible assemblies, those assemblies become recursive processes—what Sara Walker calls lineages. As lineages continue to reproduce themselves, they make some events more possible than others—contingency remains, but it is reduced as some lineages become more influential in their surroundings.
At some point, the debris that eventually formed the Earth gained a center of gravity more powerful than other objects in its surroundings. That gravity turned motions into an active assembling action that became the lineage that is the Earth. As a result, some events are simply left behind as practical possibilities. Another Earth in the vicinity became highly unlikely.
Will Neanderthals reappear on the Earth? Probably not, but they were far more numerous than Homo sapiens for many thousands of years. Our lineage has displaced them in such a way that, while some of their genetic code exists in most of us (2-4% depending upon your own ancestry), they are highly unlikely to reappear in any significant way.
Time without a Subject
To leave behind a definition of intelligence that implies possession is to leave behind the language of subjects and objects. If intelligence is not a possessed object, there is no subject that can possess it.
Therefore I’ve left the subject of my definition intentionally vague: the adaptive and expansive capacity to make the future less like fate and more like an open field of possibilities. It is a capacity without a subject. Nietzsche once said that we should separate lightning from its flash—i.e., the lighting is not the subject that is responsible for making itself flash. Lightning is the flash and the flash is the lightning. No subject is responsible for the objective output.
Holding open the question of who or what exercises intelligence seems to me to be, at the moment, a more promising orientation for inquiry. It allows me to focus on the effects rather than the essential features. Nor am I tempted to go down the more traditional analytical path that insists on clear definitions structured as ‘What is _____?’ Similar to studying the weather, we should care mostly about effects of what we tend to call intelligence and how to get out ahead of them.
Rather than pursue the rabbit warrens of ‘what is intelligence?’, we should attend to the effects of what we are building in the name of intelligence. How does it work? What should we be worried about? How should we mitigate its effects? How can we predict those effects sooner rather than later?
A Field of Inquiry
To repeat the definition: intelligence is the adaptive and expansive capacity to make the future less like fate and more like an open field of possibilities.
I do not offer this as a hard definition—it remains a conjecture—but as a field of inquiry capable of addressing issues that could be gathered under the category ‘intelligence’, such as:
Can we craft Histories of Intelligence that would look at how life gains increasing control over the way nature’s rhythms unfold? David Deutsch’s chapter on ‘The Jump to Universality’ in The Beginning of Infinity offers what I believe to be a good model of these histories. Each jump represents an expansion of intelligence’s capacity to shape the future by increasing possibilities. The movement from pictographic writing (like cuneiform) to alphabet-based writing is one such jump. From thousands of symbols to a handful of highly abstract letters allows for a much easier to learn system while making any word possible, whether existing or not yet invented.
These Histories of Intelligence allow us to re-read the historical record as the wandering tracks of its expansion, not just representations of the human condition or isolated expressions of their particularly place and time.
If recursive predictive modeling dominates current engineering concerns, how do we broaden this perspective to account for other aspects of intelligence that don’t readily fit this definition—discernment, generosity, wisdom, humility? These, I argue, belong within a field of inquiry that focuses on how necessity turns into possibilities. Recursive predictive modeling will likely stretch to include each of these eventually, but we should start with them inside the definition of intelligence, not outside.
These Histories of Intelligence should avoid becoming mere polemics rather than accompaniments to the engineering endeavor? Accompaniment itself is a form of collective intelligence born of our spiritual traditions. They have as much to each us about intelligence as back-propagation as a learning method.
Such Histories of Intelligence would re-read the cave paintings of Northern Spain and France that were created as long ago as 30,000 to 40,ooo years. These paintings pre-date the Neolithic Age, which some have argued we left behind in the twentieth century. These are all part of the history of intelligence in so far as they are episodes in the history of intelligence learning to use representations to achieve some level of imaginary control over predators.
Babylonian astronomers watched the night sky for omens sent by their gods. This was a direct confrontation with fate—predicting omens ahead of time could lead to outmaneuvering the gods. As they documented their observations, very long term patterns became predictable. Humanity learned to compute faster than nature could run the programs.
To win the Longitude Prize, John Harrison invented a clock that no longer required navigators to measure the movement and position of the stars to determine where they were on the Earth. Clock time jumps to universality—it can travel anywhere on Earth and remain the same. Time is untethered from the heavens and made purely and completely mechanical—all in the service of expanding colonialism and overcoming the fate of hostile and disorienting oceans.
To build a supercollider is not to oppose nature or transcend it—it is nature's own complexity arriving at a new capability. When we recreate the energies of the Big Bang inside a tunnel beneath Geneva, intelligence is reopening a timeline that seemed foreclosed—those first fractions of a second at the beginning of everything. The Universe turns out to have created the possibility of simulating its own reconstruction in a lab.
All of these episodes—currently parceled out to various disciplinary swim lanes—can be seen as important moments in the adaptive and expansive capacity to intervene into the laws of necessity to turn fate into possibilities.
Venus and Mars
Such an inquiry would be attentive to the good and evil contained in intelligence itself. As Pope Leo XIV has just written:
Technology has the power to heal, connect, educate and protect our common home; but it can also divide, exclude and generate new forms of injustice. In the abstract, technology in and of itself is not a solution to humanity’s problems, just as it is not inherently evil. In practice, however, technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it. (Magnificat Humanitas, paragraph 9)
To return to the weather: is the wind good or bad? A gentle breeze on a warm day feels great. A microburst that topples a park full of trees and damages property is bad. We call both wind, but that tells us very little about the phenomenon. ‘What is wind?’ doesn’t seem to matter. What matters is what it does. The same with intelligence. We must be attentive to its motions and its effects.
Postscript
To set out to write a history of intelligence is impossible if we think of history as a single threaded story that lines up all the events, breaks, and continuities that sum up a concept. Time doesn't work this way. It is layered, and any thing insofar as it is a thing has a history that are the traces of its multi-layered assembly within the Universe. This is especially true of intelligence, which is now called upon to tell its own history. How does intelligence tell its history? It can only do so through episodes that themselves layer time.
My definition of intelligence—the adaptive and expansive capacity to find contingency in necessity and turn it to other ends—opens a field of inquiry composed of these episodes. In the closing pages of their wonderful history of classification systems, Sorting things Out: Classification and Its Consequences, Geoffrey C. Bowker and Susan Leigh Star write of the relationship between science and philosophy with reference to Michel Serres, a thinker central to Time as Practice:
In this book we have attempted to develop tools for maintaining these open spaces. Michel Serres has best expressed the fundamental ethical and political importance of this task. He has argued that the sciences are very good at what they do: the task of the philosopher is to keep open and explore the spaces that would otherwise be left dark and unvisited because of their very success, since new forms of knowledge might arise out of these spaces. (321)
Bowker and Star have captured in these sentences the case I’m making for a History of Intelligence: to take an existing category (intelligence) that is successful in some areas of knowledge (science, technology, business) and seek out its blind spots so as to make other forms of knowledge possible—specifically forms of knowledge that re-orient us to the Enlightenment’s wager on an infinitely open and better future.
Today his is an ethical and political task precisely because of the power and influence of artificial intelligence. This power is engineered to operate at great speed and scale, which makes the philosophical task as described by Serres all the more urgent.