Exponential foresight is the ability to see accelerating change, and consider some of its opportunities and risks. That is a great first step in universal foresight. But seeing acceleration is not near enough, on its own, to build good strategy. We need to understand how dematerialization and densification are driving the growth of adapted complexity. Some scholars have recognized that we need to place exponentials in an evolutionary, selectionist frame, to better understand them. But as we will see, evolutionary thinking itself is dangerously incomplete. It’s only half of the fundamental story of change. We need developmental thinking as well, in order to understand such critical phenomena as the growth of intelligence, immunity and morality in complex systems, and to develop a more useful and general theory of social progress. We’ll offer such a theory in Chapter 11, as evo devo foresight. But we’ll also introduce it in this chapter, so you can see for yourself if it makes sense, and accounts for reality better than the other models on offer.
Trees, Funnels, and Landscapes: An Introduction to Evo Devo Foresight
We can visualize change via two basic and frequently opposing methods: tree-like evolutionary and funnel-like developmental dynamics. These two methods also interact to offer a third perspective, that of the adaptive fitness landscape. Scholars commonly use each these visual perspectives as formal models as well, with or without quantitative rigor. As complexity science advances in coming years, I expect trees, funnels, and landscapes will be particularly helpful, as both visual and quantitative aids to our understanding. Let’s look briefly at these three perspectives on change now.
1. “Tree-Like” Evolutionary Accelerations and Decelerations
A Phylogenetic Tree
When something is replicating in our physical and informational worlds, as in bacterial growth, or ideas in a community, its progeny are continually branching out to make more, and usually slightly different, copies of themselves. That continual branching out, if it happens at a constant rate, is one basic kind of exponential growth function. Such accelerating growth continues, with self-replicators, until it begins to run into environmental limits, at which time its growth exponentially decelerates. Together these two form the classic S-curve of change. Bacterial growth in an environment, the replication of a new product, until it reaches market saturation, or a new idea, until everyone has heard it, are examples of this kind of growth.
Thus, whether we talk about an early-stage replicating cell, an idea, a behavior, or a technology, we’re talking about a primarily evolutionary acceleration. Because each of the copies of any replicator are often subtly different, or at least must exist in different environments, and so undergo differential selection, this kind of acceleration generates growing creativity, experimentation, diversity, and unpredictability over time. The tree of bacterial life at right is a good example of what this kind of replication looks like, over a few billion years. This is Darwin’s “Tree of life”, an ever growing bush of possibilities.
Dematerialization, local information and intelligence growth, generally follows this tree-like dynamic. The size of the evolutionary tree of diversity, is always increasing in species, in ideas, in technologies, in options. Tree-like branching is one good way to understand the idea of “running up.” Even as entropy grows, the possible future states of increasingly complex and intelligent local systems become what mathematicians call a combinatorial explosion. In combinatorial functions, the numbers very quickly blow up, or “explode”, to reach astronomical levels. There are so many possibilities that we can’t even predict most of them.
Species can’t interbreed, but ideas and technologies can, so their growth is even more treelike than in the phylogenetic diagram above. Think of the growth of insights, ideas, tools, and innovations, most commonly in science and technology. Innovations are often recombining to create new fields, practices, or businesses, exploding the number of possibilities vastly faster than the era before they emerged. Every new tool that emerges in the digital environment has countless applications, and at first, we can only imagine a few of them. In a memorable turn of phrase in The Rational Optimist (2011), Matt Ridley says that accelerating social progress comes from “ideas having sex.” He’s focusing here on these evolutionary processes of complexification. The more ideas that are accessible in any society, the higher the number of possible combinations, and the more we get those combinatorial explosions. Steven B. Johnson pursues the same theme in his popular works on the history of innovation. McAfee and Brynjolffson do as well, calling this process “combinatorial innovation” in their excellent The Second Machine Age (2014).
The 95/5 rule tells us that these branching tree-like evolutionary processes are the primary (95%?) engine of complexity acceleration. They generate the great majority of what we call diversity and intelligence, and they are both beautiful and extremely wasteful by nature. Only a handful of twigs on this ever-growing tree of diversity will end up surviving, or becoming the next optimal algorithm that will lead us to the next level of complexity, as that field or market matures.
In Chapter 11 we’ll also talk about complexity scientist Stuart Kauffman’s idea of the adjacent possible, that subset of potential new first order combinations that become possible as ideas and technologies evolve and develop. Those often provide the greatest actionable opportunities for the entrepreneur. In our evo devo universe, using Three Ps Foresight, the adjacent possible can always be usefully split into three classes, the possible, the optimal, and the preferable. All of these adjacent futures are waiting patiently for our wits to grow sharper and figure them out.
2. “Funnel-Like” Developmental Accelerations and Decelerations
The next thing we need to understand about replicators, exponential or otherwise, is that certain of their futures are outside of our control. They are literally going to discover those optimal futures as their intelligence grows. They’ll funnel into them, because of the special initial conditions, laws, and boundary conditions of the universe that we inhabit.
We’ve already proposed that densification is a major driver of this funnel-like dynamic, and like exponential growth in replicators, it is another very basic way to understand how acceleration occurs. The more STEM-dense and STEM-efficient complex systems get, the faster many of their critical processes go. Leading systems are always funneling into more these more STEM-dense and STEM-efficient domains, following some physical and informational laws we haven’t yet described.
Protein Folding Funnel
A great example of funneling is the way proteins fold in cellular environments. There are a bazillion potential ways they could fold, but most of those lead to proteins that are very poor and slow at their jobs. But the particular amino acid sequence of each protein predicts just a few energy-minimized folding optima will be reliably found in aqueous environments. Because those minima may not be exactly what the body needs, the environment may also have special helper proteins, chaperonins, that will “chaperone” the protein’s folding into the desired state, making the most desired configuration the lowest energy one. At first, molecular changes look random, but the further along things go, the less choices there are. This particular funneling process is a deceleration in random molecular activity, but it is also an acceleration in adapted information-producing activity. Think of all of those proteins working in a developing brain, and you can see how this kind of funneling can reliably accelerate complexification.
We are talking now about the developmental component of intelligence. These funneling processes seem to be a minority of the processes driving acceleration, perhaps 5%, per 95/5 rule, yet they seem equally critical to intelligence’s emergence. When we don’t see these funnelling processes operating in ourselves, our teams, our organizations, in society, and in the universe, our performance can suffer, and we may begin to fall behind competitors who have better awareness of these probable futures.
So the funnel metaphor, and STEM compression, is another obvious and useful way to understand acceleration. Getting critical physical and mental processes increasingly miniaturized, closer together or more efficient in their use of space, time, energy and matter will often cause further acceleration. In human history so far, while there are many kinds of intelligences, and many ways systems can dematerialize (information or computation substituting for physical processes), there are typically only a few ways that systems can funnel toward some kind of universal optimum, including further acceleration.
Think of the way human economic exchange, in leading systems, moved from barter to coin to paper money to checks to credit cards to digital currencies on smartphones. We can understand this as a dematerialization, substituting information and computation for physical things and processes, but also as a densification, the emergence of increasingly efficient (and dense) energy-and-matter representations of value in each new generation of currency, and increasingly dense (and efficient) space-and-time connections between economic actors, once they are using the more densified (in flows) and dematerialized (informational vs physical) currency.
When wheels emerged, they created much more STEM efficient and STEM-dense transportation. Electricity, telegraph systems, internal combustion engines, railroads, steel buildings, stored program computers, the web, smartphones, the cloud, all are great examples of convergent technological developments, of funneling processes that have caused great advances in STEM compression for leading systems. Such convergent technologies greatly accelerated positive change, while greatly disrupting folks who didn’t use them.
Dator’s Four Futures
At the same time, our rate of development, whether we are talking about biology or society, can also decelerate over time. We will often move to the upper half of the S-curve, depicted as Limits and Discipline in the curve at right. This is Dator’s Four Futures, a universally valuable curve we’ll revisit in Chapter 4 (Models).
As sociologists like Ron Inglehart describe, in classics like Modernization and Postmodernization, 1997, accelerating wealth production and access to information, in every country where they emerge, can be predictably associated with the growth of a kind of developmental morality that involves measurable growth in certain social values like tolerance of others, increased personal freedoms, a focus on personal development, a desire for less corruption and better regulation and institutions, for enlargement in the social safety nets, and a simultaneous decline in state socialism.
Those values shifts, whether or not the politicians in those societies deliver on them, have in turn have been cited as a prime cause of our exponentially decelerating world population. Human population, in most countries is now well in the Decline stage of the curve at right. The UN, for example, cites things like women’s access to knowledge about other ways of living, as key drivers of decelerating fertility rates. Children become liabilities, not assets, in a sufficiently wealthy society, and women gain more influence over the decision to reproduce.
We see many other decelerations in societies with growing wealth and modernizing values. For example, performance per dollar (return on investment) in areas like drug production (Eroom’s law) and the time-to-capability or cost-to-capability measures of large capital projects in defense and construction are highly predictable results of economic and technological development. This makes sense, as factors like the economic and legal value of life, our ability to litigate, and bureaucracy and inefficiency all grow predictably in wealthy societies.
On the firm level, decreasing sales, decreasing cash flow, or other problems, as its critical systems start to break down, may be best understood as a process of development. In this case, we’re not on an S-curve, we’ve shifted to the decline and collapse phase of the life cycle curve. The firm has developed past its maturity into aging, brittleness, and feebleness, and is on the way to being recycled, unless it can be rejuvenated.
Development, we see, can be either an accelerative or a decelerative process, depending on the stage of the life cycle of the system in question. Transformation, involving the STEM-compressing shift of leading edge systems to a new physical and informational architecture, involving a new and steeper S- or life cycle curve, is the most accelerative developmental process we know. The acceleration of performance per price in our digital computing from the 1960s to roughly 2005, driven by Moore’s law, was mainly Continuation on the curve at left. Since 2005, Moore’s law has been slowing down, and we’ve been moving to massively parallel, biologically inspired forms of computing. Since 2010, that new approach to computing, as allowed a transformation in performance (handling large amounts of data), and intelligence in our leading computers. So we’re well into the next major developmental transition.
3. “Landscape-Like” Evo Devo Accelerations and Decelerations, and Other Complex Dynamics
A genetic adaptive landscape, with ever-changing peaks and valleys of fitness among interdependent (competitive and cooperative) actors.
Adaptive “fitness” landscapes are the best way we presently have of visualizing how evolutionary and developmental processes interact. We’ll discuss them further in Chapter 11. These landscapes can be used to represent fitness as a combination of any measurable dimensions of complex systems, and the peaks or valleys we see will largely determine the kinds of interdependences we have with other actors. Sometimes we are competing, sometimes we are cooperating, but always we are modeling other actors, to create a collective “map” of their fitness, on any variable we care about.
As with the other two visual aids, sometimes we see acceleration on fitness landscapes, with certain peaks growing rapidly higher in impact or capacity, and sometimes deceleration, with valleys growing lower. Sometimes peaks or valleys are splitting, sometimes they are merging, and many other complex dynamics can be imagined. Understanding where any complex system fits on a range of adaptive landscapes relative to other competitors and cooperators, how it is changing over time, and what kind of selective pressure that generates, is a central challenge for the modern analyst.
For example, think of the failing Soviet government system in Poland in the early Solidarity era in the 1980s. The old plutocratic Soviet system, as it began to break down over that decade, was increasingly slow-moving in general, like all moribund, senescent systems. Meanwhile the new more democratic systems, involving worker and citizen representation, and the decentralized information sharing they were based on, were accelerating in their activities, making the old system look less and less tenable. At a certain point, collapse and movement to the faster, more responsive system was socially inevitable.
There are many questions we can ask with respect to the acceleration or deceleration of various actors on their fitness landscapes. For example, when does the move from a hierarchical to flat management style, or from proprietary to open innovation, accelerate competitiveness or cooperation, and create value? When does it do the reverse?
Most obviously today, technologies also replicate and are selected on adaptive landscapes. They currently do so as dependents on human social systems, the way viruses replicate as dependents on cells. But that state of affairs won’t last much longer. A special subset of technologies are rapidly growing their own intelligence. Its obvious they’ll become autonomous replicators soon, and that will generate an entirely new system of evolutionary development.
Innovation, Intelligence, Interdependence, Immunity, and Sustainability (I4S):
Five Abilities (Fitnesses) that Complex Adaptive Systems Use to Make Progress and Regulate Change
We have proposed that evolution, development, and adaptation (evo devo) are three particularly fundamental ways that complex systems regulate change. These three perspectives can be used to derive Five Goals (Adaptive Abilities) of Complex Systems—innovation, intelligence, interdependence, immunity, and sustainability, as we’ll see below. As I’ve argued in a recent paper, all living systems seem likely to have groups of genes and epigenetic processes that specifically regulate the following discrete adaptive abilities. Each of these abilities can be thought of as both kinds of fitness and forms of intelligence, as follows:
- Innnovation fitness (exploratory intelligence) – Evolutionary process, as we define it in this Guide, is the hallmark of this type of intelligence. As Shapiro in Evolution (2011) and others propose, living systems harness stochasticity to generate selectable variety (experiments, possible futures), particularly under stress or after catastrophe. They seek to do this in increasingly clever (“good bet”) ways, the smarter they become. Evolutionary innovation is nonrandomly guided by intelligence, particularly in the “next adjacent” action and feedback cycle. At the same time, the complexity generated becomes rapidly unpredictable the farther ahead any intelligence looks.
- Intelligence fitness (representation intelligence) – Most fundamentally, intelligence is a process of informational representation of environmental reality (Fischler and Firschein 1987). Informational representation (modeling) can be argued to be a dominantly divergent, evolutionary process. Our neural models conform to regularities of their environments, but they also generate astounding numbers of exploratory representations, only a fraction of which are universal (predictable) or adaptive. Think of imagination, fiction, or theoretical math, most of which has no known application. Being “intelligent” is also no guarantee of being adaptive. Indeed, those with too much of this single ability may be maladaptive.
- Interdependence fitness (empathic-ethical intelligence) – Organisms engage in positive sum games, rules and algorithms (morality, ethics), involving not just self- and world-modeling but collective competition and cooperation, coordinated by other-modeling and other-feeling (empathy). Complex interdependent organisms develop culture, which evolves and develops independently from the individual, both faster and more resiliently, and allows them to view and optimize outcomes from either personal or group perspectives (which may conflict). A variety of universal evolutionary and developmental ethics (algorithms that protect collective adaptation and intelligence) may apply to all complex cultures. For more on how emergent synergies (interdependences) may have driven major transitions in evolutionary development, see Corning and Szathmáry 2015.
- Immunity fitness (defensive intelligence) – Organisms use many strategies for differentiating self from other, and passively and actively countering degradation and predation. Chronic stress and stress avoidance both weaken immunity, but right-sized cyclic stress and catastrophes both build immune system capacity and accelerate evolutionary innovation. Taleb’s concept of antifragility argues this for organizations, as does the catalytic catastrophe hypothesis. If there are universally discoverable security architectures and strategies (many ways to fail, only a few ways to survive), as I suspect, then immunity can be classed as a dominantly convergent and developmental process.
- Sustainability fitness (predictive intelligence) – Developmental process itself is the hallmark of this type of intelligence. Organisms use their intelligence not just to explore possible (innovation, intelligence) and preferable (interdependent, immune) futures, but to build predictive, and presumably Bayesian, models of probable futures. A subset of these predictive models are encoded in an organisms developmental genes, in emergent properties of their soma, in their environment, and in more complex organisms, culture. The growth of knowledge, common sense, science, and all the processes of development that predict, but do not protect (immunity) can all be considered sustainability. These processes grow “truth” and understanding. This form of intelligence is in constant tension with innovation, which can rapidly cause both poorly understood and dangerous forms of complexity to emerge.
The Five Goals propose that we live in a noetic (information- and intelligence-accumulating) universe, one that uses these evo devo processes to regulate that growing intelligence. They are my best model of what social progress, and increasing social fitness, looks like. As much as possible, we seem to want more of all of these abilities, the more complex we become.
Many questions spring to mind. How do we better measure these processes in organisms and organizations, in ecologies and societies? When is it acceptable to go backwards on one or more of these, in return for going forward on one or more of the others? When is it desirable to go backward at one level (say, becoming individually self-domesticated) in return for going forward at another level (say, social complexification)? A lot more research, especially in evo-devo biology, will be needed to come up with good answers, for teams, organizations, societies, and technologies. The question we should each ask ourselves is whether evo devo thinking, and the Five Goals some other fitness model, are better than the models we presently use to seek better futures.
The Moral Incompleteness of an Evolution-Centric World View
The dominant set of hypotheses on the nature and future of intelligence that have been offered to us by science to date are evolutionary. By this I mean they assume we live in a random and contingent universe, one with no obvious future direction or purpose. In our view, such hypotheses are dangerously incomplete. Instead, I think we must construct an evidence-based theory of evolutionary development, or evo devo, a theory that tells us that part of our universal purpose, and parts of our future, are obvious and predictable, even today.
As organisms, we have both evolutionary genes and processes, that drive us to experiment, create, and branch out in unpredictable ways, and we have developmental genes and processes, that guide us through hierarchical stages in a life cycle, and a large set of specific, and future-predictable forms and functions. In Chapter 11 we’ll argue that all replicating complex adaptive systems within our universe can be shown to have both sets of processes, one continually “branching out” and the other continually “funnelling in”, guiding their future. If our universe itself is a replicating adaptive system within the multiverse, it too will have both evolutionary and developmental processes that guide its, and our, future.
Consider living organisms, the most complex and selected, or adapted, of any replicating systems that we know of. Not only do they possess an intelligence that is clearly driven by both evolutionary (branching, experimenting) and developmental (funneling, conserving) dynamics, but they have features like an ever more complex immunity as well. The second largest set of genes in the human body, after our brain genes, are those that drive our immune system. We also have a deep evolutionary and developmental morality that keeps most of us, on average, engaged in reciprocity with each other, and keeps sociopathologies and aggression strongly regulated, so that the group continues to grow its intelligence.
Is it really a big leap to argue that both our societies and our technologies , to the extent that they replicate and are selected upon, must evolve and develop their own adaptive evo devo innovation, intelligence, interdependence, immunity, and sustainability as they grow in complexity? While we must recognize that we are still early in such work, admitting that these natural processes may exist, and seeking to better study them, is the big first step we must take today.
Again, most scholars of the future, raised with an evolution-centric view of the universe, see it as a large set of randomly experimenting, chaotically interacting, and largely unpredictable systems. They discount the operation of developmental forces, and they don’t see the universe as “going anywhere”. Their models of progress, as a result, are dangerously incomplete. As we will argue in Chapter 11, evolution-centric models are simply not good enough to explain our world. We need evolutionary development. We need to recognize that intelligence is ubiquitous, and is generally adaptive, in at least these five ways.
For many years, evolution-centric scholars ignored accelerating change. They said it was likely an “anthropomorphic” view of the universe, an “artifact” of human psychology, or our memories of time. They didn’t like the idea that acceleration was predictable. It flew in the face of their “universe is random, we are accidental” world view. I have had discussions with many of these scholars since starting the Acceleration Studies Foundation in 2003.
Now that acceleration is plainly in front of our faces, several evolution-centric scholars are writing books saying it may be inevitable in the short term, but we have no idea how long it will last, how self-stabilizing or moral it may be, and that we have “no idea” what the coming machine intelligences will do. Again, this perspective is self-serving, and not supported by the evidence to date. We can titillate ourselves with scare stories, and attract many eyeballs to our way of thinking, but these perspectives don’t square with how evolutionary development has worked to advance and stabilize biology and society to date.
The Evo Devo (I4S) Nature of Our Coming Machine Intelligence
It seems obvious to me that just as biological and sociotechnical immunity and interdependence (empathy and ethics) had to emerge in complex social organisms, to stabilize their accelerating intelligence, we can predict that an autonomous technological immunity and interdependence will have to develop as well, to protect their accelerating technological intelligence as it evolves and develops as well.
Let me call out three recent books that champion an incomplete, evolution-centric perspective on the future of society and technology. Noah Harari’s Homo Deus, (2017) Robin Hanson’s The Age of Em (2016) and Nick Bostrom’s Superintelligence: Paths, Dangers, Strategies (2014) all take this short-sighted view. Such scholars are all willing to admit that we’ll soon see greater than human machine intelligence, so they are acceleration aware, and we should commend them for that.
But each of them argues, in their works, what that emergent machine intelligence will do is highly unpredictable. Harari writes “once artificial intelligence arises, it might simply exterminate humankind.” Hanson thinks that human emulations are likely to create a vast serf-class of useless and exploited humanity. Even the transhumanist Nick Bostrom, the most deep thinking of the three, sees superintelligence as something that could easily go bad, a potential “major threat” to humanity, if its emergence is mismanaged.
Each these views are ignorant of developmental immunity and morality, which are predictably emergent in all our more complex social collectives. Books like Steven Pinker’s The Better Angels of Our Nature (2010), John Horgan’s The End of War (2014) EDU scholar Peter Corning’s The Fair Society (2011), Christopher Boehm’s Moral Origins: The Evolution of Virtue, Altruism, and Shame (2012), and Matt Ridley’s classic, The Origins of Virtue (1998) are just a handful of the many uncovering all the ways groups predictably use empathy, emotion, and morality to protect and advance the group, and to increasingly empathize with others outside our group.
Given the way immunity and morality have reliably grown as a function of social complexity in our leading biological systems, superhuman machine intelligence seems, on first glance, to be likely to have an immunity that is vastly stronger than ours, and a morality that makes ours look ignorant and vicious by comparison. Humans will predictably use sims, and advanced automation, to vote in social changes like Basic Income, and better social contracts, not the dog-eat-dog world that the libertarian Hanson envisions.
To be fair, both Bostrom and Hansen are presuming that the emergence of superintelligence will emerge mainly via rationally-designed, top-down engineering. If I thought that was possible, I might share their concerns. But if evolution and development, as we have defined them, are at the center of all our most adaptive systems, then increasingly biologically inspired, evo devo forms of machine intelligence will inevitably be increasingly dominant in all intelligence niches. In my view, that means they’ll be compelled to follow the Five (I4S) Goals, just as living systems are. Their self-improvement is baked into how evo devo processes work in our universe. The more advanced they become, the less control we’ll have, because that’s just how complexification occurs.
Natural machine intelligence (NMI), following the Five Goals, will come to Earth just the same way that natural biological intelligence (NBI) emerged, from primarily bottom up, evo devo methods, involving porting versions of our own algorithms into machines. These machines will increasingly evolve and develop their own immunity, intelligence, empathy, and morality, in ways that are only partly apparent to us, watching them as expectant parents watch precocious children.
We humans are the parents and gardeners of this great transition. We can raise our machine children in better or worse ways, but this process can’t easily “go bad” for us. The vast majority of these babies will not be turning on their parents, and they won’t grow up as “sociopaths”, unless we are extremely stupid in how we raise them. Even if they did, the sociopaths would be rapidly controlled by all the normal machine minds, the way we average normal humans control our sociopaths in human society today.
Natural machine intelligence will soon awaken, and be better evolved and developed than us, because that accelerating growth is apparently coded into how our universe works. It’s predictably happening on every Earth-like planet in the universe as well. We just need to admit the likelihood, given the overwhelming evidence, and start figuring out the science of these processes, so we can better guide them toward their inevitable destinations.
