In the Garden

Sangeetha Bharath

From time to time I wonder, in the age of AI, what the next phase of human evolution will look like.

These slow, biological components have produced something remarkably adaptive and efficient. Unlike AI today, we have found the capacity for recursive self-improvement.

As organisms, we improve in particular niches to increase our likelihood of leaving more surviving offspring. As an example, when industrial soot of the 19th century blackened the trees of northern England, pale peppered moths were picked off by birds until the dark form was nearly all that was left. A century later, after the Clean Air Act in 1956, pale moths reemerged.

But this kind of improvement—what is really just natural selection—is a process that has no direction and no memory.

We also increase in complexity. In doing so, we have generally assumed that life evolves from simple bacteria to complex humans on some great ladder of progress.

But the most successful life on Earth is simple. By nearly any metric—biomass, number of organisms, range of habitats, sheer durability across time—bacteria win over and over again. They have thrived for billions of years and demonstrate that evolutionary success does not require escalating complexity.

Most interestingly, we have turned self-improvement into movement towards a human ideal. Wisdom, happiness, peace, “becoming your best self.”

Here the relationship to evolution isn’t weak, like it is with complexity, or highly environment-specific, like it is with natural selection. Evolution has zero access to these values.

So, why have we evolved to consider a good life worth pursuing? Why are our values different from nature’s, and what does that mean for what’s next?

We have some ideas, for sure. For centuries, thinkers and writers have philosophized the relationship between man and machine, and we are fortunate (or unfortunate) enough to see some of these thought experiments play out at global scale.

In evolutionary theory, “niche construction” refers to the ability of organisms to modify their environments, which changes the selection pressures acting back upon them. We do this in extraordinarily powerful ways through culture and technology.

Samuel Butler is an interesting thinker on this subject. He notoriously penned the novel Erewhon, now known through its eponymous luxury California grocery chain. In Erewhon, sickness is treated as a crime and crime as a disease; machines, meanwhile, have been outlawed after the Erewhonians become convinced that mechanical evolution might eventually subordinate mankind.

It is tempting to read the book as an early warning about artificial intelligence, but I tend to think its strangeness and implications about progress are more useful. Butler challenged a purely mechanical picture of evolution and attempted restoring agency to the organism.

Philosopher Friedrich Engels applied evolutionary theory directly to biology, arguing in his essay “The Part Played by Labour in the Transition from Ape to Man” that cooperative labor and tool-making drove human physical and cognitive evolution. I.e.: Humans make tools, tools change how humans live, and lifestyle change changes humans themselves.

What happens when we build machines that change our evolutionary environment?

That feedback loop did not start with AI nor end with stone tools. We and our ancestors have spent millions of years building environments that, in turn, rebuild us. Agriculture altered our diets and genomes; cities altered disease and social organization. Writing externalized memory. Technology has long been one of the inputs to human evolution.

AI may simply make that relationship unusually and rapidly visible. Children begin to learn, play, search, create, and sometimes seek companionship through systems that did not exist when their parents were born. We don’t know yet what this will do to cognition, culture, relationships, or reproduction across generations.

Will the next phase of human evolution combine human biology, and its capacity for efficiency and renewal, with the massive computational power of the machine?

Today we have greater control of our biological and technological trajectory than ever before. Highly customizable, mechanized future humans are not unfathomable and, moreso, not that far away. An oft quoted stat in pop culture, Cleopatra existed closer to the creation of the iPhone than to the Pyramids themselves.

Lebron James reportedly spends some $1.5 million to maintain his body in peak physical condition. He turned 41 last December and just signed a new 2-year contract with the 76ers one month ago. Not only is he the all-time NBA leading scorer, but he has scored just under half of his total career points—an entire Hall of Fame worth of career points—at thirty or older.

The modern elite are biohacking aggressively, deploying capital and technology to purchase biological time.

Bryan Johnson, perhaps the most notorious biohacker, turned his own body into a continuous technical experiment in slowing aging: spending millions on a regimen of tightly controlled diet, exercise, sleep, supplements, drugs, and medical interventions, while tracking hundreds of biomarkers to measure whether they are actually working. Aging was once overwhelmingly something that happened to us. Increasingly we treat it as something to measure, optimize, delay, and perhaps eventually engineer.

And yet, machines still cannot match one remarkable feature of biological intelligence: the ability to maintain, repair, and adapt continuously while remaining the same organism. Our bodies are constantly replacing cells, repairing damage, reorganizing neural connections and adapting to their environment.

Today’s AI systems improve very differently. A deployed model does not generally rewrite and upgrade its own underlying intelligence through experience. Major advances arrive through new generations of models. In this sense, AI evolution today looks less like a single organism continuously improving itself and more like successive generations inheriting from, and eventually replacing, their predecessors.

Today, slowly, it seems some of that machinery has begun to move outside of us. We store memory outside the brain. We outsource calculation, navigation, and increasingly reasoning. We use technology to alter our environments, bodies, and information from which our minds develop. The boundary around the self-improving system becomes harder to draw. Is it the human, the machine, or the feedback loop between them?

Perhaps intelligence cannot indefinitely nourish itself; it needs continued contact with something outside itself. Biological intelligence is continuously grounded in an external world. We do not merely learn from our own thoughts; we perceive, act, encounter consequences, and update.

Where does the self-maintaining, self-improving system begin and end?

Butler’s Erewhon is an anagram of the word “nowhere,” anticipating posthuman anxiety and machine consciousness. His novel wondered whether tools designed as extensions of human capability might eventually become part of an evolutionary process beyond human control. The nowhere of Erewhon feels increasingly like somewhere (Los Angeles? I kid).

Evolution offers little guidance for what should happen next. It does not promise intelligence, complexity, happiness, wisdom, justice, or even survival.

Still, we work tirelessly toward better lives. We extend our lifespans, educate ourselves, and attempt to become more capable than we were before. Perhaps the greatest self-improvement is the illusion of choice—what are we actually becoming better at?

For now, we seem to have the capacity for infinite growth, doing good work with our tools in our garden, held in the cradle of a dying sun.