Chamath made a provocative post on Twitter:
Aug 26, 2024
"Curious about AI’s impact on society? Look at online poker as an interesting, early window into the future.
Why?
Online poker is a multi billion dollar economy and will be worth exactly $0 over the next few years. The fall of this economy will be a harbinger of what’s possible as generalized AI advances are applied to the “low hanging fruit” markets.
Agents are becoming smart enough – reasoning, memory, chain of thought – that poker playing agents will run over the human population in legal poker sites and will make money for their creators.
But won’t these bots be stopped?
These aren’t bots. They are agents – and will know how to lose certain hands, creating an illusion of normalcy. They will learn to “shear the sheep” vs “skin the sheep”.
You can expect that online poker leaderboards will flip to names you’ve never heard of and those names won’t be real people but groups of programmers making money.
I suspect that most people stop playing online for money pretty quickly and the economic value of this market implodes."
– "Does the online poker market eventually turn into a C-robots type bot war for money? No human players, just bots tasked with better training and prompting fighting it out for money. Basically it's the new cockfight."
-" Roughly yes. I could see various agent farms competing against each other. This would make the most sense at the highest stakes, essentially killing these games for humans. There is a cost threshold for training such that agents are unprofitable at some level but that level will decrease significantly over time, approaching $0."
– "Will this add value to traditional physical casinos, where live poker is played?"
– "I think this is a logical outcome. Over time, players will also likely have to pass some kind of “metal detector” to prove they also don’t have a Neuralink-like implant.
Casinos will want to maximize the experience for people who don’t have implants. People will likely pay more and play more knowing their opponents are natty."
– "Can it be done with the stock market too?"
– "No. It’s not a game (to everyone) where a finite result happens in some small window of time."
– "Help me make one [a bot]. Let’s get some money."
– "I don’t think there that much money in it. That’s why it will be interesting to observe. It will be a quick race to the bottom."
The World of Statistics account, which has nearly 4 million followers on Twitter, responded to Chamath's post with encouraging data, showing that rumors about the death of online poker are premature:
"The global online poker market was valued at $2.9 billion in 2020 and is projected to reach $4.6 billion by 2027.
The number of online poker players worldwide is estimated to be around 100 million.
The average age of online poker players is 31 years old."
Well-known professionals also do not share Palihapitiya's pessimistic view.
- Patrick Howard (Mobius Poker)
"This view is too pessimistic. Bots are a real issue, but poker solvers, machine learning, and AI have been around for years, yet major bot problems only plague unregulated sites that don't care.
Your analysis overlooks the severe penalties for getting caught. Making money in poker requires capital and time to overcome variance. If you're caught botting, all your funds are confiscated, and if a bot farm has any detectable fingerprint, the entire operation collapses like a house of cards (no pun intended).
At high stakes, most online players know each other, so if an unknown account starts crushing, it's a major red flag. The combination of these factors makes makes the risk-to-reward unfavorable in well-regulated environments."
Chamath proposed a specific example:
– "I will play online and have a separate machine tape my screen in real time, process it and give me GTO recommendations on a separate tablet. I’ll then click on the right play. Every time. How will you catch me? Be specific."
– Patrick Howard: "I'll ask you to record yourself playing from multiple angles for a few days and analyze your play style before/after. PokerStars often requires this from pros (see attached screenshot below)."
"Anyone with an anomalous win rate or unusually accurate stats could be asked to do the same. In the future, some sites might even require pros to always have webcams on during play, but we’re not there yet.
Also, if you’re just reading outputs from a screen, your timing will likely show it. You’d need to mimic human-like timing in your actions to avoid detection."
- Joe Ingram
"This has been going on for many years and online poker is bigger than it has been in a long time (debatably) – heads up displays with any type of player data you wanted to display in real time were one of the first versions of this that players used to make decisions
Agents do battle with other agents regularly and variance still exists so it doesn’t mean they are unbeatable. Players of all skill level want to compete for the massive tournament prize pools and anyone can get lucky at one of those provided the operator isn’t programming the “game” against you.
Operators run their own programmable “agents” who can be designed to interact with other “agents” at the table to create the illusion of liquidity and a real marketplace to win at whatever rate they are designed for.
Players have their own versions of “agents” giving real time feedback on what decisions to make and they are becoming easy to access for everyone.
Smart “agent assisted” “players” have been creating the rollercoaster effect for years at live and online poker while printing bb/100
I don’t see the market going to 0 with all of the content creators being incentivized to promote the game 24/7 and talk about this topic 0% of the time – people will still want to gamble for big scores so if an arena can create events like that, people will show up."
- Matt Berkey
– "I've played with him probably seven or eight times in high-stakes games, and he's a brilliant man. That's abundantly clear. But I also kind of feel like he lives in a bit of a techie Silicon Valley silo, you know? A bit of an echo chamber perhaps, where he might not get enough vantage points outside of that. Now, to be fair, he's talking about AI in this instance, so what do we know compared to the people he surrounds himself with? I do think that there might be a lot of siloed group-think when it comes to these types of things.
The fact that he brought poker into the conversation now allows us to talk about this a little bit because I think that we can at least have some opinions on what the impact of AI has been and will be. We can make some predictions moving forward."
"So far, I mostly agree with this. If we're going to unpack this line by line, I don't know that online poker will be worth exactly zero. I think that there are a lot of ways that operators can protect themselves. I think there are a lot of ways that they can pivot as technology grows, and we've seen a lot of those efforts being made currently.
I think it might be real names. I think it might be 360-degree cameras. I think there are a lot of unfortunate methods that are going to need to be implemented.
What I do agree with here is that generalized AI will be applied to low-hanging fruit markets. I think that was a really well-stated sentence, and I think it is something that we can anticipate. Not just as online poker falls into that category, but looking even further into other categories that are very big markets like this and easy low-hanging fruit for AI. Sports betting comes to mind.
We've already seen this in day trading. Day trading has been automated for years."
