I often catch myself mourning the normal that hasn’t ended yet.
I’m living in Hyde Park finishing school after a year in Berkeley. The maple trees by my apartment have turned red, as usual. I like them.
I don’t dread the future when I mourn. I don’t find high-variance futures in the abstract very scary. I’m worried AIs would destroy humanity, but I also have hope. I expect I won’t be mourning this moment in ten years, whatever happens.
Mourning the present is like mourning a friendship when I move countries, because I know it’s not going to be the same. I start noticing our little inside jokes; our conversations linger in my memory longer. I live more, because I expect this moment to end soon.
I.
Last month my history of science professor took us to a museum to look at Egyptian artefacts. One of the archive items was the notebook of James Henry Breasted, whom UChicago appointed as America’s first professor of Egyptology 121 years ago. The yellowed pages held his cursive transcriptions of hieroglyphics.
An archaeologist at the museum, turning the pages carefully with a wooden stick, joked that one of her dreams was for her notebook in 100 years to be displayed like this.
My first thought was pretty uncharitable: wow, it’s crazy to expect this profession to still exist in 100 years. Probably we’ll have definitive answers to most answerable archaeological questions. Probably robots will be excavating the banks of the Nile and deciphering all the tablets in the world, if a compute-rich person or AI cares enough about this. Knowledge won’t be a slow and careful endeavour. The mathematicians are feeling it already.
Then these thoughts passed and I felt a tender admiration for her. She had recently deciphered embalming recipes from the inscriptions on the outside of excavated jars. She showed us one, and the symbols looked entirely foreign to me. I imagined it must have taken years of patient effort to get to her level of knowledge.
I knew a few people who wanted to be archaeologists. When they talked about the academic path ahead, they seemed happy to spend their lives mastering the esoteric; this pursuit gave them meaning. I know I’m romanticizing the ivory tower here. Still, I think deep domain knowledge is a little sacred to those who spend a life acquiring it.
I don’t know how the archaeologist would react if the sacred spring of knowledge ran dry and came back like the flooding Nile. In her position I’d probably be confused. Angry, too, not at the flood but at myself and the people who summoned the flood, because how can you be angry at a force of nature? I might think, “what beauty is there in almost knowing, when an oracle can compute in one step what it would take me a life to figure out?” Maybe I’d recover, and find something else that I’d consider sacred. Maybe I’d worship the flood like the ancient Egyptians.
II.
I went to my math professor’s office hours to chat.
One student sat with an old textbook open. Another student brought a Claude transcript. I used to be the first student. I’d borrow old seminar notes at the math library and work through them on a blackboard. Now I mostly have Claude tutor me. It’s more efficient, but the purist in me feels a little guilty.
I expected the professor to prefer the textbook over Claude, but if he did, he certainly didn’t show it. He remarked, Claude, she’s a colleague. I simply advise you that you always make sure you understand what she’s doing. I was surprised by how steady he seemed.
We started talking about AI and math. I asked him what the probability of an AI solving P vs NP by 2028 was. I was curious, since he’s one of the most respected living mathematicians in his field. I expected him to give a number in the low double digits, given that he had just called Claude a colleague, and Claude had been improving fast.
He made a circle with his thumb and index finger. 0%.
Part of me wants to believe him. It’s not implausible to me that AIs solve all other Millennium Problems and make scant progress on this one. Part of me doesn’t want to bet against the exponential.
A week later OpenAI and Anthropic announced within hours of each other that their internal models had found a sub-exponential algorithm for 3-SAT, refuting the Exponential Time Hypothesis. We’re of course still a long way from P vs NP, but if I internalize the exponential, P vs NP being solved by 2028 feels increasingly plausible.
Two years ago when I tabled for the AI safety group at the club fair, I told skeptical students that superintelligence would probably be developed within our lifetime; o1 can solve 83% of high school math competition problems, I said, and a few years ago these models struggled with simple math problems. Today I’d drop the “probably”.
III.
In September I took the California Zephyr with my mum from Berkeley to Denver. In the observation car she chatted with a lady whose daughter was about my age and wanted to pursue antitrust law. The lady showed us pictures of her daughter. In one video she was singing opera.
I smiled and nodded along to the conversation. I don’t really know if law as we know it will exist when she graduates from law school. Probably there will still be lawyers. She seemed bright and driven, and I’m positive that in most worlds without catastrophe she will be fine. I couldn’t say anything to the lady. I wasn’t there to argue, and it would’ve been cruel if she believed me and presumptuous if she didn’t.
I don’t feel misunderstood or smarter or desperate when I talk to people like her. Mostly I feel affection, with a tad of grief.
I’ve spoken to my mum a lot about AI safety. She used to talk about my future like this lady. Now she tells me that whatever happens, she’ll be by my side.
IV.
On autumn mornings I enjoy walking a fixed loop. The highlight of my walk is a modernist quadrangle that once housed the Lutheran School of Theology. A quarter of this building is office space for data science and AI, and I used to work from there every day.
Last year, the sight of the maple tree by its window aflame in the sun was often overwhelmingly beautiful. When I sat beneath its red leaves this year, it felt like looking at a childhood photograph.
Now the maples are growing bare, and I will be leaving Hyde Park to return to the Bay soon. I’ll miss UChicago. In Berkeley, I’m surrounded by people who feel time ticking as I do, but it doesn’t have the same spirit of serious play: people inventing elegant little games and dancing within them as if they had all the time in the world. When I was back for school this spring, I got to LARP as an Italian Renaissance cardinal and do complexity theory. The fun felt luxurious.
When I first started college, I lived as if I had time, even though the arguments seemed right. I indulged in the present, so I could not have mourned it. Now I’m spending what’s left of the normal working on AI safety, and I get to mourn clear-eyed. There are miles to go before we get to sleep. I want to see many more cycles of maple leaves, red, brown, and green.
The maple tree in the quad
Thanks to Connor Blake, Keshav Shenoy, Kenneth Sun, and Matthew Khoriaty for helpful comments and discussions on the piece.
I often catch myself mourning the normal that hasn’t ended yet.
I’m living in Hyde Park finishing school after a year in Berkeley. The maple trees by my apartment have turned red, as usual. I like them.
I don’t dread the future when I mourn. I don’t find high-variance futures in the abstract very scary. I’m worried AIs would destroy humanity, but I also have hope. I expect I won’t be mourning this moment in ten years, whatever happens.
Mourning the present is like mourning a friendship when I move countries, because I know it’s not going to be the same. I start noticing our little inside jokes; our conversations linger in my memory longer. I live more, because I expect this moment to end soon.
I.
Last month my history of science professor took us to a museum to look at Egyptian artefacts. One of the archive items was the notebook of James Henry Breasted, whom UChicago appointed as America’s first professor of Egyptology 121 years ago. The yellowed pages held his cursive transcriptions of hieroglyphics.
An archaeologist at the museum, turning the pages carefully with a wooden stick, joked that one of her dreams was for her notebook in 100 years to be displayed like this.
My first thought was pretty uncharitable: wow, it’s crazy to expect this profession to still exist in 100 years. Probably we’ll have definitive answers to most answerable archaeological questions. Probably robots will be excavating the banks of the Nile and deciphering all the tablets in the world, if a compute-rich person or AI cares enough about this. Knowledge won’t be a slow and careful endeavour. The mathematicians are feeling it already.
Then these thoughts passed and I felt a tender admiration for her. She had recently deciphered embalming recipes from the inscriptions on the outside of excavated jars. She showed us one, and the symbols looked entirely foreign to me. I imagined it must have taken years of patient effort to get to her level of knowledge.
I knew a few people who wanted to be archaeologists. When they talked about the academic path ahead, they seemed happy to spend their lives mastering the esoteric; this pursuit gave them meaning. I know I’m romanticizing the ivory tower here. Still, I think deep domain knowledge is a little sacred to those who spend a life acquiring it.
I don’t know how the archaeologist would react if the sacred spring of knowledge ran dry and came back like the flooding Nile. In her position I’d probably be confused. Angry, too, not at the flood but at myself and the people who summoned the flood, because how can you be angry at a force of nature? I might think, “what beauty is there in almost knowing, when an oracle can compute in one step what it would take me a life to figure out?” Maybe I’d recover, and find something else that I’d consider sacred. Maybe I’d worship the flood like the ancient Egyptians.
II.
I went to my math professor’s office hours to chat.
One student sat with an old textbook open. Another student brought a Claude transcript. I used to be the first student. I’d borrow old seminar notes at the math library and work through them on a blackboard. Now I mostly have Claude tutor me. It’s more efficient, but the purist in me feels a little guilty.
I expected the professor to prefer the textbook over Claude, but if he did, he certainly didn’t show it. He remarked, Claude, she’s a colleague. I simply advise you that you always make sure you understand what she’s doing. I was surprised by how steady he seemed.
We started talking about AI and math. I asked him what the probability of an AI solving P vs NP by 2028 was. I was curious, since he’s one of the most respected living mathematicians in his field. I expected him to give a number in the low double digits, given that he had just called Claude a colleague, and Claude had been improving fast.
He made a circle with his thumb and index finger. 0%.
Part of me wants to believe him. It’s not implausible to me that AIs solve all other Millennium Problems and make scant progress on this one. Part of me doesn’t want to bet against the exponential.
A week later OpenAI and Anthropic announced within hours of each other that their internal models had found a sub-exponential algorithm for 3-SAT, refuting the Exponential Time Hypothesis. We’re of course still a long way from P vs NP, but if I internalize the exponential, P vs NP being solved by 2028 feels increasingly plausible.
Two years ago when I tabled for the AI safety group at the club fair, I told skeptical students that superintelligence would probably be developed within our lifetime; o1 can solve 83% of high school math competition problems, I said, and a few years ago these models struggled with simple math problems. Today I’d drop the “probably”.
III.
In September I took the California Zephyr with my mum from Berkeley to Denver. In the observation car she chatted with a lady whose daughter was about my age and wanted to pursue antitrust law. The lady showed us pictures of her daughter. In one video she was singing opera.
I smiled and nodded along to the conversation. I don’t really know if law as we know it will exist when she graduates from law school. Probably there will still be lawyers. She seemed bright and driven, and I’m positive that in most worlds without catastrophe she will be fine. I couldn’t say anything to the lady. I wasn’t there to argue, and it would’ve been cruel if she believed me and presumptuous if she didn’t.
I don’t feel misunderstood or smarter or desperate when I talk to people like her. Mostly I feel affection, with a tad of grief.
I’ve spoken to my mum a lot about AI safety. She used to talk about my future like this lady. Now she tells me that whatever happens, she’ll be by my side.
IV.
On autumn mornings I enjoy walking a fixed loop. The highlight of my walk is a modernist quadrangle that once housed the Lutheran School of Theology. A quarter of this building is office space for data science and AI, and I used to work from there every day.
Last year, the sight of the maple tree by its window aflame in the sun was often overwhelmingly beautiful. When I sat beneath its red leaves this year, it felt like looking at a childhood photograph.
Now the maples are growing bare, and I will be leaving Hyde Park to return to the Bay soon. I’ll miss UChicago. In Berkeley, I’m surrounded by people who feel time ticking as I do, but it doesn’t have the same spirit of serious play: people inventing elegant little games and dancing within them as if they had all the time in the world. When I was back for school this spring, I got to LARP as an Italian Renaissance cardinal and do complexity theory. The fun felt luxurious.
When I first started college, I lived as if I had time, even though the arguments seemed right. I indulged in the present, so I could not have mourned it. Now I’m spending what’s left of the normal working on AI safety, and I get to mourn clear-eyed. There are miles to go before we get to sleep. I want to see many more cycles of maple leaves, red, brown, and green.
The maple tree in the quad
Thanks to Connor Blake, Keshav Shenoy, Kenneth Sun, and Matthew Khoriaty for helpful comments and discussions on the piece.
Also posted on Substack.