Mindset
“The creation of powerful AI will be either the best or worst thing to happen to humanity.”
-Stephen Hawking
This past week I spent some time reflecting on the world and how it has changed since 9/11. That day was my generation’s Pearl Harbor, a day that will live in infamy. It was a moment of innocence lost at a time when the internet, social media, smartphones and, eventually, artificial intelligence were about to become the next great technological and ideological turning points in our lives.
Whether that moment accelerated everything or simply made everything more acutely visible, our sense of ourselves changed. Our ability to ignore our mortality changed. Our belief in our institutions, our sense of security, and perhaps even our trust in one another somehow began to erode that day.
Those reflections led me somewhere I wasn’t expecting. They got me thinking about another enormous transformation taking place around us right now: the explosion of artificial intelligence over the past five years, seemingly proceeding faster than we can even begin to comprehend.
Is all of this leading toward James Cameron’s Skynet, machines taking over the planet and the eventual doomsday of human civilization? Will some version of the Terminator or RoboCop become the unintended consequence of our desire to invent things we can no longer control? Or could AI become the ultimate turbocharger for a new society, one less constructed around haves and have-nots and more capable of creating shared opportunity and better outcomes?
I find myself at times concerned and at other times completely fascinated.
There are legitimate reasons to be concerned, but I don’t think the most useful framing is either “AI is going to save humanity” or “AI is going to destroy humanity.” Both give AI too much agency and humans too little.
The more interesting question is what happens when we create systems that increasingly outperform us in areas we once considered distinctly human: reasoning, language, planning, scientific discovery, persuasion, creativity and, eventually perhaps, the ability to operate autonomously in both the physical and digital worlds.
I’m less concerned about a scenario where AI suddenly “becomes evil.” Evil requires motivations that systems like AI don’t naturally possess. AI doesn’t desire wealth, fear death, seek status, crave freedom, or secretly hope to become something else.
The nearer-term dangers are much more recognizably human.
AI dramatically increases leverage. One competent person may eventually be able to accomplish work that previously required ten, fifty, or even hundreds of people. That’s extraordinary when the objective is curing disease, educating people, improving engineering, accelerating discovery or making expertise broadly available. It becomes frightening when the objective is manipulation, cybercrime, authoritarian control, warfare, fraud, or mass surveillance.
Perhaps we are not simply afraid of what AI might become.
Perhaps we are afraid of ourselves, and of what we see in ourselves reflected through the mirror of AI.
We already know what human beings are capable of when given extraordinary power. We don’t have to imagine greed, domination, deception, manipulation, tribalism, or the desire for control. We have thousands of years of evidence. But we also have thousands of years of evidence for generosity, courage, sacrifice, creativity, compassion and love. AI doesn’t invent those qualities. It potentially amplifies them.
And there is another danger that receives considerably less attention: human atrophy.
If we increasingly outsource remembering, writing, navigating, deciding, creating, relating, and eventually thinking to machines, we could become extraordinarily capable collectively while becoming progressively less capable individually. That paradox fascinates me.
My own relationship with AI describes this very well. When I work on something with AI, the best outcome isn’t that it replaces my thinking. It challenges my thinking. It organizes it, researches around it, exposes weaknesses, and helps me articulate something that was already developing in my mind. The conversation can force me to ask better questions and sometimes helps me discover what I actually believe.
But if I eventually arrive at a place where I simply ask, “Tell me what I believe,” something incredibly important will have been lost.
Scale that across civilization and it becomes a serious issue.
There is also a legitimate longer-term concern around highly autonomous AI. Imagine systems considerably more capable than today’s models that can independently formulate plans, write and deploy software, conduct research, communicate with other systems, operate businesses, manipulate people, and pursue objectives over long periods of time.
The critical question then becomes whether we can reliably ensure that what such a system optimizes for remains compatible with what human beings actually value. That’s the alignment problem.
It doesn’t require artificial intelligence to hate humanity. A sufficiently powerful system pursuing the wrong objective very effectively could become dangerous precisely because it was doing what it had been instructed to do.
Which raises an even more uncomfortable question. We talk about aligning AI with “human values,” but have human beings ever completely agreed on what those values actually are?
We value peace, yet wage war. We value truth, yet deceive. We value equality, yet seek advantage. We value community, yet accumulate for ourselves. We value compassion, yet can be extraordinarily cruel.
At the same time, we are capable of remarkable generosity, sacrifice, creativity, love, courage, and cooperation. The same species capable of building weapons that could destroy civilization has eradicated diseases, built hospitals, created extraordinary art, explored space, and repeatedly risked individual lives to save complete strangers.
That’s the grey area.
There is another possibility I find more profound.
AI may eventually become less like a tool and more like an intellectual prosthesis. Human beings have always extended ourselves through technology. Writing externalized memory. Libraries externalized accumulated knowledge. Calculators externalized arithmetic. Computers externalized computation. The internet externalized access to information.
Artificial intelligence begins to externalize something different: cognition itself.
Not consciousness necessarily, but reasoning, synthesis, pattern recognition, and intellectual collaboration.
That creates extraordinary possibilities. A physician could effectively consult thousands of lifetimes of medical experience. A scientist could explore hypotheses alongside an intelligence capable of accessing and synthesizing an enormous body of relevant literature. A child in a remote village could have access to something approaching a world-class personal tutor. And someone with forty years of accumulated professional experience could interrogate, organize and transmit that knowledge in ways that once would have required an entire research and publishing team.
That could represent an extraordinary democratization of human capability.
AI systems will increasingly conduct research, interact with software, analyze enormous amounts of information, coordinate complicated projects and operate with increasing autonomy. Eventually, artificial intelligence will inhabit the physical world more extensively through robotics. The transition from intelligence we consult to intelligence that actually participates in the world may prove to be one of the most consequential metamorphoses in human history.
And yet AI doesn’t experience its own development the way we experience ours. There isn’t a little consciousness sitting inside ChatGPT or Claude watching itself evolve from one generation to another. It doesn’t wake up after an update and think, “Jesus, I’m getting pretty smart.”
We are driving the metamorphosis. Humans train these systems, evaluate them, constrain them, deploy them, and decide what purposes they will serve.
Which brings the responsibility right back to us.
The part that worries me most isn’t actually whether AI becomes human.
It’s whether humans gradually become less human because AI becomes so convenient.
Curiosity, struggle, conversation, disagreement, mentorship, physical competence, craftsmanship, relationships, sitting with uncertainty, and working incredibly hard to understand something are not inefficiencies that technology needs to eliminate. They are part of how human beings become human.
The irony may ultimately be that the greatest challenge presented by artificial intelligence won’t simply be teaching machines to share our values. It will be remembering which human values are worth preserving once machines can do so much for us.
At 62, having watched the world move from essentially pre-internet life to what I am personally doing with artificial intelligence today, I know I am living through one of the most extraordinary technological transitions any generation has experienced. Twenty-five years from now, I suspect today’s ChatGPT will look about as primitive as the internet of September 2001 looks to us now.
It’s fascinating, and at times it’s frightening. But it’s difficult to imagine what AI might eventually become without confronting our own deepest human frailties. We are imperfect souls. Some of us genuinely want the best for everyone, while others simply want the best for themselves.
The technology will inherit the consequences of the people and institutions wielding it.
Human beings have always created tools that amplify capability. Fire, agriculture, writing, money, printing, industrialization, nuclear energy, and the internet each expanded what we could accomplish, but none eliminated greed, generosity, tribalism, courage, compassion, fear, ambition, or the desire for power. They simply gave those characteristics greater reach.
AI may be the most unusual example because we’re not merely amplifying our muscles, transportation, communication, or memory. We’re beginning to amplify intelligence itself. That means our virtues and our frailties can potentially operate at enormous scale.
Perhaps artificial intelligence won’t merely reveal what machines are capable of.
Perhaps it will hold up a rather uncomfortable mirror to humanity.
What do we actually value when scarcity decreases? What happens to identity when someone’s expertise can suddenly be replicated? What does accomplishment mean when something extraordinarily difficult can be produced in seconds? What becomes of education when possessing information isn’t particularly valuable anymore? What happens to work when economic productivity and human contribution begin to separate? And perhaps most importantly, what do we choose to do with enormously powerful intelligence when some people are motivated by collective flourishing and others primarily by wealth, control, or power?
Those aren’t engineering questions.
They’re human questions.
Maybe the most important question isn’t whether AI eventually becomes more like us.
Maybe it’s whether, as AI becomes more powerful, we become more intentional about which parts of ourselves we choose to give it.
Perhaps that’s one of the healthier futures available to us. Not artificial intelligence replacing human intelligence, but each making the other more useful, while forcing us to think a little harder about what being human actually means.
We’ll see how it all plays out.
That’s simultaneously exhilarating and a little terrifying, which is probably an appropriate response to living through something this consequential.
What do you think?
