[Reader’s Note: my earlier posts were either fiction, i.e., meant to entertain, or non-fiction, intended to inform. In this post, I try to combine the two. Does it work? Please let me know in the comments below. FYI – Renata and Juan are significant players in my latest thriller NOT SO GONE. All the tech mentioned here is current, true, and accurate. Enjoy!]
The early morning light filters through the blinds in Juan and Renata’s third floor Park Slope Brooklyn apartment. Juan rolls over in bed and kisses Renata on the back of her shoulder. “I need some coffee badly,” he says.
“Tough night? I didn’t even hear you come in.”
“Sorry I was so late.”
“Studying?” Renata asks.
“Sorta. Exams next week. Took a break and stumbled on an article about Pluribus, a program created by Carnegie-Melon and Facebook. This sucker beat five top poker pros at No-Limit Holdem.”
Renata turns around to face him, her long brown hair spread across her pillow. “Fascinating…” she says as she suppresses a yawn.
Undeterred, Juan continues. “Remember when IBM’s Big Blue beat Gary Kasparov, the world chess champion twenty years ago? That was a big deal….proved that a computer could be programmed to beat the best human at something.”
“How did it do that?” Renata perks up.
“From what I know, it learns and can calculate millions of possible moves several steps ahead to plot the optimal next play. The best human player can only figure maybe eight moves ahead.”
“OK, the computer is a better, faster calculator than a human. So what?”
Juan sighs, exasperated… “It’s not just a ‘stupid computer trick.’ A couple of years ago, Google’s AlphaGO program beats the Korean world champion at GO, an even more complex and subtle game.”
“OK, genius. Get dressed and we’ll make your emergency coffee run. Do I need to call an ambulance or can we just walk there?”
Juan laughs and throws a pillow at her, “Smartass.”
They leave their apartment and walk a block to Cafe Grumpy on 7th. The aroma of coffee wafts through the air while Bob Dylan’s, Don’t Think Twice is playing in the background. A line of people waits to get to the counter, all noses pointing down to glowing cell phone screens.
Juan taps his foot and fidgets, “Maybe if making coffee could be done by computer, I’d get my fix before midnight.”
“Calm down, hombre. Hey, let’s finish the computer game discussion… About the poker program, sounds like more numbers and more calculations to me. The computer can see the whole board in chess and GO, then calculates the moves from there, right?” Renata says.
Juan’s eyes light up, “Yeah, that’s true, and that’s why the Pluribus program beating Holdem is such a big deal.” He waits for a response, but Renata is taking in her surroundings, the exposed brick walls, the driftwood tables, millennials on one side and the retiree coffee klatch on the other.
She turns back to Juan, “Go on.”
“With Holdem, the computer and the players can’t see all the cards until the final card is exposed. Imagine trying to play chess if you could only see half the other player’s pieces. How would you do?”
“I’d get crushed.”
“Exactly! That’s why most poker players are losing players. It’s a game of incomplete information. The players take the information they do know, like the cards in their hands, calculate their odds, and bet or bluff accordingly. That’s also why there’s no bluffing in chess or GO.”
“Can’t you feign a bluff in chess or GO, making it look like you’re doing one thing when your real plan is a zinger that your opponent doesn’t see? I don’t get it. Why no bluffing?”
“It’s not that easy. As you said, in chess and GO you can see all the pieces. You can’t pretend or bluff that you have a better position than is readily apparent on the board. But in Holdem, you can’t see my cards or the cards to come on the flop. So, I can pretend or bluff like I have a big hand, make a large bet, and maybe get you to fold a better hand. All because of incomplete information.”
“Kinda like life, huh? We make decisions all the time without knowing all the facts.”
Counterfactual Regret Minimization
“Yep, we take risks. We try to figure the odds of the best outcome, but it’s still a gamble.”
“Like our relationship?”
“Um, I hope not. I mean I’m all in on this relationship, but there are no guarantees, are there?”
“Nope. We don’t know the future, but we’re both taking a gamble.”
“And enjoying the game. Don’t forget that part.”
“OK. So how does this Pluribus beat the pros with incomplete information? Maybe we can learn something.”
“Like the chess and GO programs, it learns. But it does something else called Counterfactual Regret Minimization.”
“What the hell is that?”
“It learns from every hand, especially the ones it loses. It says like, “OK I fucked up that hand. What can I do better next time?’ And it does it over and over again thousands, even millions of times until it perfects a better strategy than the pros.”
“Something like learning from your own mistakes…”
“Right and with the power of a computer, it can crunch a lifetime of learning into a few days.”
“Reminds me of something our friend Sam says, ‘I hope I live long enough to benefit from all my mistakes.’”
“Ain’t that the truth.”
They reach the counter. Renata orders her favorite cinnamon dolce creme and Juan orders a large iced coffee with an extra shot.
“So why are you thinking about this regret minimization stuff now?” Renata asks.
The redheaded barista rings up their order. “That’ll be $8.32,” she says looking at Juan.
Juan shuffles his feet and turns to Renata, his head bowed. “Because I lost all my money at poker last night. Mistakes were made. Can you pay?”
Renata’s face flushes. She inserts her Visa debit card into the reader. “Now I understand the regret part,” she snickers. “I’m practicing some Counterfactual Regret Minimization myself right now.”