Letter to a recent college graduate
I know you're of hearing about AI. I am, too.
I talk about AI all day at work, then I talk about it all night with my family and friends, with parents in my community, with the public at these AI safety workshops. We’re all sick of it, believe me.
But it feels dishonest to not talk about it now, in a letter to a young person who is about to kick off their career. I’m not here to hype AI, I’m definitely not here to give you five tips to AI-proof your career path. But I do want you to see what I see.
I’ve been building and selling and living and breathing AI for 20 years now, and here’s the honest truth as I see it: AI is going to massively, massively impact and disrupt your life and plans for the future. Much of that impact will be negative. This is not fair.
We didn't always know that AI would go like this. It wasn’t clear when I started working on it two decades ago; it wasn’t even clear when ChatGPT came out five and a half years ago. At the time, it felt really plausible that either the current approach would hit a ceiling and we would need further breakthroughs that buy us time; and/or, we’d get a strong regulatory AI safety framework in place.
Unfortunately, neither of those have happened, and even more unfortunately, there is very little evidence that either of those will happen soon. Instead, AI is continuing to advance and improve incredibly quickly, on the back of hundreds of billions of dollars being spent every year. Here's a brief reminder of the progress over the last five years.
Five years ago, ChatGPT launched. It was text in, and text out. It did the AI equivalent of saying the first thing to pop into its head. It was amazing and engaging for a time, but it wasn’t actually useful for anything. Since that time, it has improved:
- You can speak to it and it responds in a natural voice.
- It can deeply “understand” images, music, video, and documents..
- It can produce images, music, video, and documents, and these are often indistinguishable from something a person would create.
- It can design, generate, and test computer code.
- It can use tools like calendar and email, and it can even control your whole computer on your behalf.
- It can reason, think aloud, and grind away on nuanced tasks for hours and hours.
That’s an almost unbelievable amount of progress in five years. It’s fair to say that AI has advanced more in the past five years than it did in the previous six decades. And critically, it’s not slowing down. AI improvements are actually accelerating.
We have some quantitative data that shows this. One metric is from a group called METR. They have been doing independent tests for a while now on how good AI is at completing tasks of varying complexity. The details of the measurement aren’t worth going into here, but the punchline is that it’s increasing exponentially. The complexity of tasks that AI can handle doubles every seven months. Even that pace seems to be increasing lately as the frontier labs themselves use AI to accelerate their work.
What is this building towards? We don’t really know, but one milestone we expect soon is “AGI”. This stands for Artificial General Intelligence, which turns out to be really hard to define because we have no good definition for “intelligence” and sticking even more terms on it doesn’t help. But intuitively, AGI is a computer that is as smart as a human; concretely, it's an “AI that is better than the median human at every knowledge work task”. (Knowledge work being basically anything you do in front of a computer, and median human being math-speak for basically “typical human”.)
There’s a lot of discussion and debate on the AGI timeline, but there’s general consensus that we’ll cross it in the next few years. Let that sink in a little.
At some point soon, maybe while younger students you know are still in college, we’ll cross a point where AI is better than most people at anything done on a computer. And of course, once this line is crossed, it will never be uncrossed. It will be the new reality, forever. Crossing the AGI threshold will be completely destabilizing to us as individuals, to our society, and to the world order, and we are profoundly unprepared.
Ok. So what do we do with that? At the societal level, there’s a bunch of stuff. Safety regulations, risk frameworks, universal basic income, improved social support, all kinds of stuff. But I’m not talking to a politician; I’m talking to a smart, curious, engaged young person who wants to earn a living and live a life that’s meaningful to you. What can you do?
I don’t have any silver bullets, I don’t have any tips and tricks, I don’t have much reassurance, but I do have hope. And the hope is this:
In a world where AI is better than the typical person at everything, do not be a typical person. Do not do the thing that anybody else in your position would do, because that is exactly what AI will do, but cheaper and faster than you. Do not compete with AI at being a typical programmer, or a typical graphic designer, or a typical marketer, or a typical salesperson, or a typical writer. You will lose.
Imagine a Venn diagram with two overlapping circles. The first circle is You, the other circle is Everybody Else. This circle of Everybody Else is exactly what was used to train the AI in the first place. It’s mastered and commoditized it.
The intersection in the middle, the part where those circles overlap? This is the part of you that does things the way anyone else would have done them, and the economic value of this part is exactly zero. Your livelihood will all come from the margin, the crescent moon part that is uniquely You. Your weird and quirky and unpopular opinions, friends, background, scars, experiences, intuitions… that is where your paycheck will come from.
Said differently: AI will master conventional wisdom; your wisdom will need to be unconventional.
But now I'm swerving dangerously close to a cliche. What do we do in a world of AI?
- Dance like nobody’s watching!
- Be the you-est you you can be.
- Be yourself – everyone else is taken.
- In a world of Cheerios, be a Fruit Loop.
We say these things, we put them on posters in kindergarten classrooms, we venerate some of the people who actually do it.
But in practice, it’s been really, really hard to be yourself, to own your unconventional wisdom, because that is not what the System actually wants.
What “The System” actually is is different for all of us, but we all feel it. For you, the System may be capitalism, or the patriarchy, or Western culture, or institutional racism, or your own cultural background, or straight folks, or your parents, or neurotypicals, or normies.
Whatever the System is for you, it thrives on predictability. It wants predictable people, with predictable opinions, buying predictable products, taking predictable commutes to predictable jobs.
And in return for making yourself predictable, it offers you at least the illusion of safety and comfort. It offers approval and reassurance you’re doing the right thing.
Navigating and balancing between what the System wants from my life and what I want from my life has been the central tension of my experience as an adult, and I think it is for many of us. It’s hard to understand what our true desires and goals are, distinct from what the System wants, and even harder to have the guts to pursue them.
There’s a great Ursula Le Guin short story about exactly this tension, it’s called The Flyers of Gy. There’s this race of people that evolved from birds. Most of them are basically human, doing normal human things, and they just happen to be covered in feathers. The fact that they were originally birds doesn’t come up much. They live in houses, they drive in cars, they go to work.
But one in a thousand, when they go through puberty, they also sprout wings. Huge wings. They’re painful growing in, but the people who grow them can fly. They can soar! Of course, there’s a real cost to the wings. There’s risk and danger inherent to flying. There’s stigma, jealousy, pity, mistrust, and resentment from those without the wings. And they’re a huge hindrance in everyday life! They take up a ton of space. The winged people don’t fit in well, literally and figuratively.
The story focuses on two characters with wings who deal with those wings very differently. Arciadia embraces his wings. He works part time delivering packages, but lives a happy life of freedom on the wing, as a flier. There’s risks to flying, but rewards too, and that’s what he’s chosen. But there’s another winged character, unnamed, who has chosen a normal life as a member of a respectable law firm.
He’s got a nice wife who loves him in spite of his wings, good kids, a mortgage, all that stuff. In the words of Le Guin, his wings are “flattened, bound, and belted down.” The closing of the story are from an interview with this lawyer, and they read:
“Shortly before I left, I asked him, ‘Do you ever dream of flying?’ Lawyerlike, he was slow to answer. He looked away, out the window. ‘Doesn’t everyone?’, he said.”
This story broke my heart, because everyone of us has wings and can soar, every one of us dreams of flying. And almost every one of us flattens and binds our wings down in service of the System. That’s the deal with the devil we all make, every day. We bind our wings and suppress our dreams of flying, and the System says “Good job, you’re doing the right thing, here’s a paycheck.”
But AI has changed that. Now with AI, the System is breaking that contract with us. It’s breaking up with us. It has no more need for normal, predictable people, it’s got AI for that.
So for us, there’s no more conflict, no more tension with trying to fit into the System. There is no more motivation for you to pretend that you’re less weird, less unique, less yourself than you truly are. There is no longer any value in binding your wings while you look out the window and dream of flying.
In this sense, AI has freed us, has liberated us. Not out of good will or intention or care for the human spirit, but by conquering and commodifying everything that is normal, is typical, is “median”. Now it’s just up to us, up to you, to unbind your wings and seize that freedom.
This liberation is exciting, but it’s also unmooring. It's scary. There’s two things to say here.
The first is that, while this is all scary, for me too, there’s no other option. It’s scary to leave the nest not knowing if your wings will carry you, but the nest is burning. AI has set it alight, and there is no safety there any longer. Unfolding your wings and flying away is no longer something for the bold, admirable few; it’s the rational response to rapidly changing conditions.
And finally, wear your weirdness on your sleeve. Don’t mask and pretend you’re normal. You’re not, and there’s no longer anything to be gained by pretending otherwise. Keep your wings unfurled where others can see them and love you for them. Cheer for others when they take off, especially if their early flights are ungainly or barely get off the ground. Make your parents and family see that this is their role now: to celebrate your effort, your intention, your desire to truly soar.
That’s what I’m trying to do now in my life, and it’s working. I’ve got a weird, wonderful wife who loves my wings as I love hers. We’ve got three amazing and weird kids, and we’re building a weird life together. I’ve got weird parents. I’ve got weird friends and we’re sending weird texts on a daily basis.
It’s great. And/but I’m also 43 years old, and my therapist will tell you that it’s been a long way to get here and I’ve still got a long way to go. I want you to speedrun that. As the System turns its back on us to embrace AI, I want you to turn your back to it, unfold your beautiful wings and soar.
I’ll look for you up in the clouds.
Sincerely, Nate Nichols