If you want to see where human-machine competition is fiercest, look to games. Complex, rules-based games present the ideal environment to train AI systems, where the industry's top startups compete against the world's best players.
Enter GPT-3: it's like GPT-2, but much bigger
The many experimental applications of GPT-2, the most advanced text generator to date
Salesforce's economic simulations have created some very unexpected tax models
Microsoft Flight Simulator 2020's AI maps and real time weather is worth knowng about, even for none flight sim fans
DeepMind sets its sights on an annoying and maybe world-threatening phenomenon: specification gaming
Watch an AI train itself in Super Mario Bros. 2 live
Games have been an AI proving ground since the 1960s, when early bots competed with humans in checkers.
Today, leading AI startups train their technology on some of world's the most competitive and complicated video games, including Starcraft II and Dota 2.
Often, the training culminates in a real match against top players. Google DeepMind and OpenAI have both competed with human eSports champs.
Below are several high profile events:
- OpenAI's OpenAI Five: OpenAI's AI system beat the world's top players at Dota 2 in April, 2019.
- Google DeepMind's AlphaStar: Starcraft II is a real-time strategy game. In December, 2018, DeepMind's AlphaStar program beat a top-25 Starcraft II player.