I’d promised I would keep myself under control this time. But the night I took home the grand prize at the forecasting tournament, all thoughts of my primary romantic partner left my mind. I laid down a couple bounties, and within an hour, a gaggle of top-rated cuddle therapists were in my penthouse hotel suite.
Leaning back in my chair, I took a long rip of synthetic oxytocin, which I probably didn’t need because Krystal was massaging my shoulders while Kredence was rubbing my feet and Klarity was gently calibrating my priors. Around midnight the cops came because somebody complained about the noise. I greeted them shirtless and asked the officers to hold my martini while I scribbled out a map of 90% confidence intervals for the year-over-year violent crime rates this coming summer broken down by precinct. They thanked me profusely and left without so much as a warning.
One of the cuddle therapists whispered in my ear, “I think I’m experiencing limerence toward you.”
I railed another line of crushed-up modafinil off my copy of The Age Of Em and tried to remember her name. “Sweetheart,” I murmured, “I hope you feel compersion for the game.”
To keep her from getting attached, I wandered off to where the others were snapping photos with the trophy for Mastodon. “Russia-Ukraine ceasefire, February 2024,” I declared.
Kredence and Krystal came over and perched on my lap.
“One hundred thousand people use AI tax attorneys in 2026, order of magnitude.”
Krystal stroked my arm. My System 1 tingled.
“Republicans hold the Senate majority in 2038. 72%.”
“What about AI timelines?” asked Kredence.
“Priced in.” Maybe the tingling was just my Emsam patch. “2045: Israel and a pan-Arab state each suffer 100-plus casualties clashing over competing airspace around Jupiter’s moons.”
“I didn’t know you knew so much about astropolitics,” said Krystal.
“I don’t. I deduced it from freshman chem, first principles, and the surprisingly telling physiognomy of the Crown Prince of Oman.”
Just then, there was a sharp knock at the door. I figured it must be room service delivering Foreign Affairs. But when I opened the door, a hulking figure stood in the hallway: it was my archnemesis, Nathaniel Knight.
“I feel surprise at your presence,” I said coolly. “I notice you aren’t on the guest list.” He stormed past me anyway.
“I couldn’t help but overhear you doing some after-hours forecasting,” he said, adding under his breath, “As could this entire floor.”
“Scout grindset never stops, bud. There is no offseason.”
“Two sig figs fifteen years out doesn’t sound very scoutlike to me.”
“If you’re so smart, why don’t you have a trophy?”
Knight sighed. “The approach to registering one’s predictions that maximizes accuracy in expectation is not, of course, the approach that maximizes the chances of taking first place. Impressively… strategic. Although to prioritize the latter suggests a rather carefree attitude toward” – he eyed with disdain the fresh tattoo of Thomas Bayes’s face spanning my chest – “epistemic hygiene.”
Klarity caressed the boundaries of my conscious awareness. I tucked three crisp impact certificates into the waistband of her harem pants. “Systematized winning, baby.” I wiped at the powdered magnesium L-threonate covering my nose. “This is what good judgment looks like.”
“You’re making forecasts you have no business making. Your confidence intervals are wildly undersized.”
“I can’t help it. I’m a maggie fiend.”
“That is not your main problem!”
I turned my back to him and snaked an arm around Klarity, talking even louder than before. “2031. Ninety percent of newborns in Western countries will be extensively genetically edited for strength, intellect” – wink – “and looks.” I stood up to admire my reflection in the floor-to-ceiling glass in front of the balcony until Knight punched it, narrowly missing me, and it shattered. Klarity screamed.
“Hey!” I shouted. “That window is priceless!”
He lunged at me.
“It’s Overton, Knight!”
“Not till I’m through with you it isn’t,” he growled. “You’re overindexing on the inside view.” He hauled me by the scruff of my neck onto the balcony. “Consider” – leaning me out over the railing – “the outside view.”
With a sharp finger, he drew a two-by-two grid on my chest. “Top left. Information built into your model–”
“–I know, I know, unknown unknowns.”
“You don’t know, you don’t know! You don’t know what you don’t know!”
“Unhand me, good sir,” I implored sweetly, feeling the ashwagandha hit. “An agent must act, no? Which requires some implied probability, however uncertain? Ontologically, I see no bright line distinguishing a hypothetical with some likelihood of occuring from a set of hypotheticals some of whose members are unknown. After all, randomness is randomness – the uncertainty of a known type of event ultimately breaks down into uncertainty about sub-events that can’t be anticipated.”
“The necessity of action is precisely why you have to be rigorous about your uncertainty in the underlying model! If you can’t know a certain kind of information, you can’t act on it, so you don’t bake that into how you proceed, but when you’re reporting your probability estimate, you’ve got to specify the places your world-model lacks confidence so that I don’t incorporate your random guesswork into my update from learning your model plus its internal estimate.”
“You think I’m operating on just a single–”
But I was cut off midsentence by a loud flapping noise. I ducked just in time to avoid being sideswiped by a massive wing. When I stood back up, looming behind Knight was an enormous black swan.
“What’s the problem, boys?” the swan asked. It had the deepest voice I had ever heard on an animal. “You both know Aumann’s agreement theorem. Two rational agents can’t disagree after reflection.”
“There’s no more reflection,” I said. “He broke the window.”
“This madman was entirely neglecting epistemic modesty!” Knight protested.
“Epistemic modesty, my beak,” the swan chuckled. “Here’s what epistemic modesty looks like. Sit down, the two of you.” He gestured at the balcony chairs and unfurled a long scroll from under his wing onto the table. It was blank except for the top, which read: PROBABILITY DETROIT LIONS WIN A SUPER BOWL BEFORE CALIFORNIA HIGH-SPEED RAIL COMPLETED.
“Now, you’re a couple of smart fellows,” said the swan. “The perfect pair for an adversarial collaboration. Instead of arguing, why don’t you each enumerate all your assumptions, subquestions, and priors. Then you can iteratively identify the cruxes of your disagreement and reach a consensus.”
I felt myself beginning to sober up. Knight sheepishly wiped some of the blood off his knuckles. We each pulled out a pen and started to map out a web of scenarios.
Half an hour later, the two of us strolled back into the hotel room triumphantly waving a mutually-endorsed tree of hypothetical cases, academic meta-analyses, granular sub-probabilities, and arrows indicating causality. Then I looked up and dropped the scroll.
All the champagne bottles were empty. My dossier of alpha and Billy Beane action figure were nowhere to be seen. The trophy was gone. The only sign of the room’s previous inhabitants was a smattering of inky feathers, blonde hairs, and rose petals on the pillows. Thinking, Fast and Slow was filled with dog-ears.
Somewhere below, a car revved its engine. I ran back out to the balcony just in time to see the black swan speeding away in a convertible with Krystal riding in the passenger seat and Kredence and Klarity in the back. “Excellent work, boys!” he hollered over the screech as the car drifted around the corner. “You’ll take the econ blogosphere by storm. Keep on systematically winning! You’re perfect Bayesian agents!” As they disappeared from sight, I could hear his derisive shout fading into the distance: “Hail to the futaaaarchs!”
When they got onto the freeway, the swan turned to Krystal. “What’s your read on whether self-driving will make this thing obsolete by 2027?”
Hehehehe. This is awesome stuff.