‘Gainability’ and the Assertion of Purpose
Intelligence, Creativity, and Moving Faster than Nature
What is intelligence?
A recent paper by Joseph Chen ambitiously seeks a universal definition of intelligence: ‘Intelligence is the ability to predict the future accurately and to benefit from that prediction.’
The second part of the definition is the most important. It’s one thing to be able to predict, but the ability to evaluate predictions and recursively improve them is the human story. I believe this adds a crucial dimension that has gone missing from previous definitions: the capacity to sort good outcomes from bad ones, and to learn from that experience.
Implied in Chen’s definition is not a sequential movement much as a recursive one. We don’t necessarily predict then decide if we’ve benefited from the prediction. Modern AI research and engineering has been running up against the limits of this sequential approach to logic and computation. Rather, what we find at the heart of this definition is a recursive learning model that requires Chen to lean on words like ‘motivation’ and ‘creativity’ as we’ll see. In this Wednesday post, I want to begin exploring more deeply and concretely how the technical and engineering focus of AI research on the question ‘What is intelligence?’ is opening up the discussion to more venerable traditions that have a long history in philosophy and theology that is obscured when we reduce the discussion to the instrumental concerns of technical progress.
Gainability and Orientation to Better Futures
The recursive movement of learning from the outcomes of predictions makes orientation toward better futures fundamental to how intelligence works—at least human intelligence. This new feature Chen calls ‘gainability’: ‘the capacity to achieve beneficial outcomes through predictive interactions with the environment’ (39). This capacity necessarily includes the ability to shape the environment to make benefits more and more likely: ‘We can even use gainability to explain successful adaptation. For example, someone who learns to make an igloo gains a significant survival advantage, which is both a highly gainable behavior and a successful adaptation’ (43).
Gainability can be understood as an expanding capacity (i.e., learning) to assert purpose. From this perspective, is an igloo really that much different from a road or a dam or any other infrastructure? Is a road anything other than a predictable and repeatable pattern through a landscape?
We are not talking simply about mental capacities. We build the environment around us to reduce prediction error. The assertion of purpose is the origin of infrastructure and is a fundamental outcome of our expanding gainability.
What is purpose?
Purpose is not a thing, not an essence, not a static feature. It is a functional capacity that allows a living thing to influence what happens next.
The Earth’s evolution has created a broad spectrum of ways in which purpose can be effected and expressed. No catalog exists, but let’s look at some examples.
A bacterium asserts purpose by navigating its environment seeking sugars for nourishment. It does not build bridges to cross rivers to gain access to better sources. It does not domesticate apple trees in orchards, but it does navigate its environment seeking a fulfillment of a functional purpose—stay alive, regulate its need for sugar, and propagate.
Moths avoid bats by purposefully randomizing their movements. They seem to be able to enhance the inherent randomness in the firing of neurons, and when this enhancement combines with the moth’s wing structure, chaotic movement in flight occurs. The bat, for its part, is trying to predict the moth’s movements. This prediction would take the form of what a hockey player would call ‘skating to where the puck is going to be’, but the moth’s erratic movements frustrate the bat’s purposeful calculations.
While all life arguably exhibits purpose, our specifically human capacity for purpose seems to be our ability to continually increase the speed and scale of our influence—our ‘gainability’ grows, expands, speeds up, and takes more into its scope.
We can envision multiple futures for ourselves—individually and collectively.
We can set goals and make plans.
We can execute the plan, monitor progress, and adjust as needed.
Purpose and Creativity
Chen argues that increases in gainability require increases in creativity—the ability to imagine more and more practical possibilities in any given situation. He calls these possibilities ‘affordable actions’ and they require the ability to ‘propose’ options, which requires the creative ability to imagine options:
Creativity can actually be explained by the ability to propose affordable actions. Given equal levels of prediction accuracy and motor abilities, those who can propose more affordable actions are more likely to achieve better gainability. Therefore, if we define creativity as the degree of variance in affordable actions generated, we can infer that higher creativity yields higher intelligence through gainability. (44)
The pairing of the words ‘propose’ and ‘creativity’ should not escape our notice. The English word ‘purpose’ derives from old French ‘propos’, which derives from the Latin ‘proponere’—literally ‘to put forth, to propose, to place forward’. The capacity to assert purpose grows with gains in creativity—the ability to propose novel futures rather than simply repeating the same old things.
Discernment, as I’ve argued in my most recent Substack post, must be present and cultivated if purpose is not to become single-minded mania. Here I can equally argue that discernment is essential to creativity. If we are to keep our capacity for asserting purpose from devolving into automatic repetitions, discernment—as the capacity to suspend the automated sequence—is required for creativity to envision and propose more options.
Moving forward from discernment, motivation will become crucial to theorize, and Chen eventually will get there. But for the moment, let’s continue with the scalability of our capacity to assert purpose.
In terms of scale, our capacity to assert purpose is far larger than that of bacteria, moths and bats. We are more creative, which means we can envision more possibilities over widely varying time scales. But is our assertion of purpose different in kind or degree? I argue that if we are to assume that everything in the universe is more or less connected, then ultimately there is no difference in kind, only differences in degree.
With respect to living things’ capacity for purpose, all differences are of degree. Purpose, in other words, is not some-thing that we have; it is a function that evolves creatively, as Henri Bergson once formulated it.
The evolution of our purpose and our creativity—which we’ve now said go hand in hand—has everything to do with an expanding capacity to envision ever-longer time horizons.
Homo propositum
Can other animals envision a future and orchestrate actions accordingly? Certainly not at the scale that we do. No other species can send ships around the world stringing tens of thousands of miles of copper cable in order to synchronize clocks and make accurate maps. No other species funds laboratories to study their illnesses to find cures and prevention. No other species has figured out how to decode its own genome and invent editing tools that make the boundary between natural selection and human purpose very fuzzy indeed.
None of this happens without the assertion of purpose being deeply embedded in the kind of creatures that we have become—ones that continually expand our purposeful capacities to get the world to do more of what we want.
Homo calculus
In order for purpose to expand in scope—to increase in gainability—we have to experience time differently than any other creature. We have to have a consciousness that looks out onto the world and sees multiple possible futures. This consciousness must be able to evaluate and select preferred futures, and it must be able to arrange causal sequences accordingly.
Steven Wolfram calls this ‘computational reducibility’—the ability to find pockets of influence within an otherwise unpredictable environment that he calls ‘computationally irreducible.’ This is an expansive capacity that grows by predicting and learning from the good or bad outcomes of its predictions. To do this, we must be a ‘computationally bounded observer’: a form of consciousness that can see through the unpredictability to discover and exploit ‘pockets of computational reducibility’. To grow our capacity to see through the fog and to discover these pockets is to ‘colonize the ruliad’—to expand our capacity to out-compute the natural world.
Wolfram’s ‘computationally bounded observer’ and Chen’s ‘gainability’ share a common idea: the evolved capacity for intelligence to find opportunities to make its predictive power more and more influential. Blaise Agüera y Arcas’s definition of intelligence as ‘the capacity to model, predict, and influence one’s future’ equally shares this idea.
It works by turning prediction into influence at greater scales—what I’m calling the assertion of purpose. As a species, we have discovered how to expand this capacity, which is effectively what we call ‘learning how to learn’.
Faster than Nature
To understand gainability and the assertion of purpose, we need to understand our computational powers with respect to speed. To be able to predict accurately is to be able to run computations ahead of nature’s schedule. The accuracy comes from better explanations of how nature’s processes work when moving at a pace unaided by human influence.
Let’s take Babylonian astronomy as an example. Looking to the night sky, the astronomers saw omens—signs from the gods foretelling future events on Earth. To see the events of the heavens as signs pointing to the future sets the stage for a massive growth in the human capacity to benefit from predictions. If an eclipse signals the potential death of a king, then steps could be taken to avoid the ordained future. By the eighth century BCE, as Moudhy Al Rashid and others have argued,
… scholars were regularly predicting eclipses. Given that an eclipse could mean death for the king, it is unsurprising that it was a priority to figure out a way to predict and plan for them. Scholars go from observing lunar and solar eclipses over and over (and over and over) again to being able to predict an entire eighteen-year eclipse cycle and, eventually, to generating mathematical formulae to express that cycle. (Between Two Rivers, 159)
These seemingly simple sentences trace a remarkable increase in human intelligence understood as Chen sees it—the capacity to benefit from predictions. Seeing celestial events as omens leads to noticing their repetition, which leads to seeing their movements as measurable patterns.
Put in Wolfram’s terms, the computationally irreducible is becoming increasingly reducible as observation becomes pattern recognition, which becomes sophisticated computational predictions that allow the Babylonians to outsmart their gods.
We find ourselves hovering around a key theme of Time as Practice—the human ability to compute faster than nature can run its processes. The gods’ omens will move at a regular and predictable pace. As soon as humans see omens becoming patterns, we are motivated to discover and run computations that move faster than these patterns.
Out-computing nature’s processes is our special brand of prediction. The better we get at it, the more influence we have gained over fate and necessity.
‘Nature’ begins its long, slow separation from ‘Culture’ in which the latter becomes the human endeavor to out compute the former so as to increase our influence over how the future unfolds.
The movement from omens to predictive calculations is an episode in this history.
Read more from the Wednesdays series.
This essay is also part my series on Rejuvenation and Orientation.
For more on cultivating out power of purpose, see ‘Cultivating Purpose and Discernment in a Computational World’.
Read my Substack posts on Purpose and Discernment and Faster than Nature for deeper dives into these themes.