Inside the 20-year quest to build bots to beat pro poker players, and how recent breakthroughs in AI research could threaten the online gambling industry
Context & Ripple Effects
This 2017 Bloomberg feature by Joshua Brustein opens an arc the rest of the coverage completes: after two decades of research aimed at beating professional poker players, the pieces were falling into place for real-money consequences. Poker was the hard case among game-playing AIs precisely because it hides information and rewards deception, unlike the perfect-information games traced from Deep Blue to AlphaGo in the history of game-playing AI.
What followed vindicated the warning. In 2019, Facebook and Carnegie Mellon announced Pluribus, the first bot to beat top pros at six-player No-Limit Texas Hold 'Em; by 2022, professional players were using solver-generated optimal strategies to augment their own play; and by 2024, Bot Farm Corporation had turned the research into a business, deploying advanced poker AI across gambling sites for millions in profit.
First-order effects
- Online poker operators face a direct integrity problem: bots indistinguishable from skilled humans can extract money from recreational players, undermining the trust the rake-based business model depends on.
- Professional players lose their moat — once solvers produce near-optimal strategies, the edge that justified their livelihood migrates to whoever runs the software.
Second-order effects
- Sites are pushed into a detection arms race against bot operations like Bot Farm, spending on countermeasures whose cost comes straight out of margins.
- The same strategy tools become legitimate products sold to players, splitting the market between sanctioned training aids and illicit table-side deployment — with only enforcement distinguishing them.
Third-order effects
- If the pattern holds, real-money skill games restructure around verification: platforms must prove the human on the other side of the table is human, making bot detection a core infrastructure layer rather than a security afterthought.
- Game-playing AI follows a recurring path — research demo, then superhuman benchmark, then commercial exploitation — and gambling, where performance converts directly to cash, is the first market to feel it at scale.
The trend: Game-playing AI is moving from laboratory benchmarks to commercial deployment, with online gambling as the first market where superhuman play translates directly into extracted revenue.