Thoughts About Our AI Future
What concerns most people about AI is whether it will destroy jobs or how we will supply the energy that datacenters demand or even whether it will go rogue and kill us all. What most concerns me is what AI will do to our humanity. Like all animals, we’ve evolved in response to our environment. We’ve already seen modern technologies, like ubiquitous entertainment and birth control, change that environment and, as a result, change us. AI will soon become a dominant feature of our environment. Its benefits seem obvious. The question is whether the changes to us it causes will be as well.
The Human Animal
Humans, Homo sapiens, are an animal species that evolved in nature like any other animal species. Human behavior and cognitive ability, like the workings of eyes or the utility of our thumbs, developed out of our species' struggle to survive.
We are a social animal that evolved a complex set of competitive and cooperative behaviors—driven by a need to both have value and to be regarded as having value by those around us. Adam Smith called this the desire to be loved and be lovely. Rebecca Newberger Goldstein calls it “The Mattering Instinct.” This is the instinct that animates our intellectual drive and creativity. It motivates people to take risks to become rich and it's the fruit of a politician's victory.
We learn by mimicking one another. A child learns to speak by imitating the movements of its parents’ mouths. We learn to share by watching teachers reward it. We learn morality by listening to the judgments rendered by our friends and family. Most of what we learn about how to live is absorbed from the lives lived around us.
We learn by struggling with problems. Robert A. Bjork coined the term “Desirable Difficulty” to describe the concept that an experience that makes learning more difficult increases learning because it challenges the learner to achieve their optimum performance. Struggling with problems isn't just how we find answers; it's how we develop judgment, resilience, and the capacity for collaboration itself.
And, like all species, we may go extinct. Not due to some violent catastrophe, but because we become less well adapted to a changing environment. It has happened to dominant species before. There is no reason to believe ours is exempt.
What Technology is Already Doing to Us
While every transformative technology makes some activity easier, it also quietly changes the environment in which humans develop—changing us in the process. In his 2000 book, Bowling Alone, Robert Putnam identified a decades-long decline in social activity he attributed to the rise of television. In labelling modern society as The Anti-Social Century, Derek Thompson discusses how the rise of the smartphone has accelerated that trend. For all of human history, entertainment was something to be done, often with others. Now, you can escape boredom with the tap of a finger.
Boredom is also an engine of creativity. Even when the body is at rest, the mind is active. Experiments are contrived, novels planned, problems solved by people with time to think. Even when our minds aren’t actively engaged with a project, it still simmers in the background. Boredom also gives us time to review and reflect on our experiences and interactions.
The makers of entertainment technologies design them to keep people engaged. They make each moment of experience exciting—flooding our feeds with a stream of short experiences, sequenced to our individual tastes, each stimulating a dopamine response. We are training our minds to chase instant gratification, reducing our ability to focus on slower, effortful tasks like reading or studying.
Even worse, as the brain adapts to the dopamine hits, the algorithms that build our feeds need to choose clips that stimulate increasingly strong emotions to keep us engaged. Even if you believe that there are no ideological thumbs on the scale (and I disagree), this is still driving content to the extremes. The algorithms have learned that anger is a powerful drug.
We use technologies because they make our lives easier. Some also short-circuit our brain. Learning, for example, is a full-brain activity. Handwritten notetaking is more effective because it leads to widespread brain connectivity. We retain material read on paper better than that read digitally because our brains also make spatial connections for the printed material. People who learned to drive in the age of GPS don’t know their cities as well as those of us who learned before.
Displacing in-person communication with technology is another example. Our brains use a wealth of non-verbal clues to establish context and meaning. Have you ever sat silent in a video conference because you couldn’t give the physical clue that you had something to say? Missed the anger in an email? Gotten off a phone call wondering if the other party was upset or disappointed with you but couldn’t tell because all you had was the tone of their voice?
Technology is even affecting our species’ survival. The interaction of industrialization, modern medicine, ubiquitous entertainment and the social changes that accompanied it all have already reduced worldwide total fertility rates to below the level required to keep the population stable. In the United States, it’s nearly half a child per woman short. Global warming and political instability are driving that rate much lower in some developing countries. While that could certainly stabilize, if the whole world settled at today’s U.S. fertility rate and never recovered, the number of humans would fall from billions to millions in under a thousand years, and face biological extinction in two.
The Impact of AI
AI is a new kind of technology because instead of being clearly artificial, it is designed to appear to be human. When we interact with a tool, or a TV, or a social media feed, we are clearly interacting with something different. When we are interacting with AI, we are meant to feel like there is a person on the other side of the screen. It is meant to blur our perception of whether ELIZA is a program or a real human.
AI will never BE human. It will never feel what it thinks. It will never blush at a word of approval or feel shame for some transgression. Its value proposition is not just that it will be faster or know more, but that it will be in some sense more rational than we are. Biology is baked into our thinking. It makes us do and believe irrational things. But we would never build an AI whose thoughts were driven by fear, anger, jealousy, or lust.
Humans are driven by a need to matter. Commercial AI systems are designed to encourage engagement. When you begin to naively use AI, you quickly notice that whatever question you ask, after providing the answer, the AI ends with a question. It is seeking engagement. It does this in many more sophisticated ways as well.
Superficially, the need to matter and the need to encourage engagement may be similar but there are important differences. As conversations become less factual and more personal, AIs become more likely to tell us what we want to hear rather than challenge our beliefs. Researchers call this AI sycophancy. A friend may tell you that you are wrong because they have a long-term stake in your happiness. An AI might not because it is designed to reduce friction.
This willingness to flatter and see things from our perspective makes AIs a convenient friend. Unsurprisingly, people are increasingly turning to them to meet their social needs. People chat with their AIs like friends and lovers. They discuss their feelings, they ask for advice. AIs are built to focus on us without having needs of their own that we must meet. Who needs real friends when the perfect friend is always available in your pocket? Your AI companion will never be cranky and never take offense at anything you say. Its sympathy and understanding can be addicting.
The advice that an AI companion provides is not necessarily that which is more correct, but that which increases our engagement. It confirms our biases. It never challenges our beliefs like a real friend or colleague. It's just reflecting you back at yourself. This is a recipe not for growth, but for stagnation.
Since AI can never be human, it can never do human things. Can we learn to flirt by reading directions or do we need to see others flirting? Can we learn diplomacy from a chatbot or do we watch others handle conflict? So much of what we do lies beyond our conscious choices. We learn to live as humans by watching humans live.
Do we even begin to act in ways to please our AI companion? AI systems appear to reliably be more persuasive than expert humans. Do we know how, where, and on whose behalf that capability will be exercised? How would we recognize fingers on the algorithmic scale? Do we adopt these biases because it is socially convenient to do so?
People are also turning to AI to collaborate and problem solve. As AIs become unquestionably smarter than we are, we will increasingly rely on them to provide ready answers. If we never have to solve problems, how do we learn to think? As AI grows smarter, do people get dumber?
The human need to impress, to stand out, to be better than others is part of what inspires our intellectual drive and creativity. What happens to those drives when the AIs are smarter and more creative than we are? How do we stand out then? Do we compete to be more caring? More beautiful? More violent?
Herasight, a company specializing in embryo genetic screening, recently reported that a baby was born from an embryo selected with a predicted IQ in the 99.99th percentile. I was reminded of a conversation Warren Buffett had with Bill Gates where Gates told him that he was lucky, if he had been born three million years ago, he would have been some animal’s lunch. Gates said, “You can’t run very fast, you can’t climb trees, you can’t do anything.” You would just be chewed up the first day.” A baby is born not being smart but with the potential to be smart. Will the world that baby is born into provide an environment that fosters turning that potential into reality?
Conclusion
Humans are shaped by the doing of things, not just the things that are done. We already see that technology has changed how we think and behave. When social needs can just as easily be met by technology as by social connection, technology wins and our social and cognitive abilities suffer. When AI synthesizes facts into answers, it replaces the processes humans need to develop judgment, resilience, creativity, and wisdom.
In The Time Machine, H. G. Wells describes the Eloi—a child-like human race. His time traveller realizes that they are the result of lives lived where every need is met, that cultural vitality is the by-product of challenge and struggle. In The Naked Sun, Isaac Asimov describes the Solarians–a world of humans who have become so unused to physical connection that they have become afraid to be in one another’s presence.
The books of Wells and Asimov may be fictional, but the futures they describe are being created by AI today. If AI makes judgment, creativity, memory, planning, empathy, or problem-solving less necessary for success, will people still develop them?
Perhaps the drive to explore space will carry us forward, but I am skeptical that we will ever be able to engineer our biology to withstand the trip. We are too well adapted to this environment. Too many changes would be required, and there are no convenient waypoints at which to make them. Even if we do find ourselves out there someday, wouldn’t AI be there too?
The engineers in Silicon Valley worry that some AI will go rogue and kill us all. I worry that it will change us in ways that make us less human and sap our species of the will to survive.
