Research led by Robert C. Brooks, University of New South Wales
While public discourse about artificial intelligence often gravitates toward dramatic scenarios - superintelligent machines either elevating humanity to godlike status or triggering our extinction - a new paper suggests we might be missing a more nuanced and already-unfolding story. According to researcher Robert C. Brooks from the University of New South Wales, AI's most profound impact on our species may be happening through countless small daily interactions that could gradually shape human evolution.
The paper, in The Quarterly Review of Biology, challenges traditional narratives about AI and human evolution. "Folk notions of human evolution often conflate it with the fate of the human lineage," Brooks notes, arguing instead that we should focus on how AI affects "which individual humans live or die, mate, reproduce successfully, and raise offspring that can do likewise." This shift in perspective from grand narratives to everyday influences opens up a fascinating exploration of how AI might already be influencing human evolution.
Brooks proposes viewing AI through an ecological lens, suggesting we consider these technologies as part of humanity's environment, much like other species evolve in response to their surroundings. This framework helps us understand how AI might influence human evolution through various types of relationships that mirror those found in nature.
Perhaps the most pervasive example of human-AI interaction comes from social media platforms and attention-seeking applications. Brooks draws a compelling parallel between these technologies and biological parasites. "Instead of their blood being sucked, or their nutrients stolen, their time and attention are depleted," he writes. These digital parasites might be creating selection pressures on traits related to attention span, anxiety susceptibility, and social behavior.
Moreover, as Brooks notes, "Social media and other forms of time and attention parasitism by machines may exploit some personality types, and people with certain genetic vulnerabilities (e.g., some predispositions to depression)."
The paper examines how AI-powered autonomous weapons and surveillance systems might function as artificial predators in the human environment. Unlike traditional predators that select for physical traits, these systems might create selection pressures for different kinds of adaptations - perhaps favoring individuals who can better avoid detection or resist automated threats.
One of the paper's more provocative suggestions concerns how AI systems might compete with humans for resources, particularly energy. As Brooks explains, the enormous energy demands of cryptocurrency mining, AI model training, and data processing could create competition for limited resources. This might drive evolution toward humans who can thrive in environments or with resources less suitable for energy-hungry AI systems.
One of the paper's most fascinating discussions concerns human brain evolution. Our species experienced a dramatic increase in brain size until relatively recently - somewhere between 35,000 and 3,000 years ago. Then, surprisingly, human brains began to shrink. "This reduction was well underway by 5000 to 3000 years ago, and to date the rate of decrease is estimated at 50 times that of the preceding increase," Brooks notes.
The paper emphasizes the extraordinary metabolic cost of our large brains: "Human brains are notoriously expensive to run, guzzling an estimated 20% of the overall energy burned by an adult body despite only making up about 2% of body mass." This high energy cost makes brain size a potential target for evolutionary change, especially as we increasingly rely on AI for cognitive tasks.
As we increasingly rely on AI for memory, calculation, and other cognitive tasks, the paper suggests we might see reduced selective pressure to maintain our current brain size. This doesn't necessarily mean decreased intelligence - just as our ancestors' brain shrinkage coincided with increasing cultural and technological sophistication. Instead, it might represent a shift toward different types of intelligence or cognitive specialization.
The energetic demands of large brains extend beyond individual metabolism. "Fetal brains place exceptional energetic demands on mothers during late pregnancy, constraining the duration of human gestation," Brooks explains. Evidence suggests that large fetal heads present substantial risks to both mother and baby, potentially making brain size a target of contemporary selection.
The paper provides fascinating insight into how AI might influence human relationships. Brooks explains that intimacy-building is "an inherently algorithmic" process involving "escalating self-disclosure." AI systems are becoming increasingly adept at replicating these processes through digital companions.
Consider the rise of "virtual friends" - AI chatbots designed to engage in conversation and build relationships with users. While these might seem like harmless entertainment, they could influence how people develop social skills, form relationships, and ultimately choose mates. Some users already report strong emotional attachments to their AI companions, and as these technologies become more sophisticated, they might increasingly compete with human relationships for our time and attention.
Dating applications represent one of the most direct ways AI influences human mating patterns. The paper notes that these systems are already changing how people meet potential mates, with about 30% of U.S. adults reporting having used a dating app. This shift in mate selection could have profound implications for which traits are passed to future generations.
The paper explores how AI companions might affect mating markets differently based on which groups they primarily engage. If these technologies predominantly attract single men, we might see reduced competition in human mating markets, potentially leading to changes in sexual attitudes and behaviors. Conversely, if they primarily engage women, we might see intensified male competition, possibly selecting for traits related to dominance and competitiveness.
Brooks suggests that as AI becomes better at meeting emotional and intimate needs, we might see evolutionary changes in traits related to attraction, pair-bonding, and parental investment. The paper suggests that if AI companions begin to satisfy emotional and intimate needs traditionally met through human relationships, we might see reduced motivation for human pair-bonding and reproduction.
"If the presence of hyperstimulating apps tends to shade out the joy and appeal of interpersonal sex, then birth rates may drop," he writes, potentially affecting the evolution of social behavior and mating psychology.
AI is already transforming genetic screening and fertility treatments. The paper explains how AI improves the analysis of genetic data, helping identify disease risks and optimize IVF success rates. "AI is making genetic analysis and gene discovery easier," Brooks notes, expanding parents' ability to make informed decisions about their offspring's genetic health.
However, the paper warns of potential unintended effects. Selection against certain disease-causing genes might inadvertently remove beneficial traits due to "antagonistic pleiotropy" - where genes have different effects at different life stages. Brooks uses the example of the APOE-ε4 gene, which increases Alzheimer's risk but appears to provide cognitive benefits in youth.
The paper raises important ethical concerns about AI-enhanced genetic selection. As Brooks notes, discussions of AI and human enhancement often "stirs considerable unease among readers familiar with the history of evolutionary biology and its relation to eugenics." He references geneticist Adam Rutherford's observation that eugenics emerged as "a political ideology that co-opted the very new and immature science of evolution, and came to be one of the defining and most deadly ideas of the 20th century." The paper emphasizes how even well-intentioned genetic selection could have far-reaching evolutionary consequences.
Looking ahead, the paper suggests AI might enable more sophisticated genetic engineering through improved analysis of gene interactions and better prediction of genetic outcomes. However, it emphasizes that the complexity of genetic interactions makes precise control of human evolution challenging and potentially risky.
Drawing on evolutionary biology's "Red Queen hypothesis" - where species must constantly adapt just to maintain their ecological position - the paper suggests humans might enter a similar dynamic with AI systems. This could create ongoing selection pressure for traits that help humans remain competitive in an AI-rich world.
The paper explores how humans might develop cultural adaptations to AI challenges before biological evolution occurs. For example, people are already developing "stealth wear" to thwart facial recognition systems, much like animals evolve camouflage to avoid predators.
Artificial intelligence might also influence human self-domestication - the process by which humans became less aggressive and more cooperative over time. AI-powered surveillance systems might make it harder for groups to organize against despotic leaders, potentially weakening the selective pressures that historically favored more cooperative individuals. Conversely, AI might help enforce social norms and reduce violent behavior, potentially accelerating self-domestication.
Brooks emphasizes that these evolutionary changes would likely be small and gradual, occurring over many generations. They might be overshadowed by more immediate cultural and technological changes. However, their cumulative effects could be significant over longer timescales.
The paper doesn't paint artificial intelligence as inherently good or bad for human evolution, but rather as a new environmental factor that will inevitably influence our species' development. Just as our ancestors evolved in response to changes in climate, food availability, and social organization, future generations will evolve in response to an increasingly AI-rich environment.
In a particularly poetic turn, Brooks concludes by borrowing from T.S. Eliot's "The Hollow Men," suggesting that humanity's future might be shaped not by the dramatic bang of an AI singularity, but by the quiet whimper of countless small interactions between humans and their AI-enabled technologies.