The recent victory of DeepMind’s artificial intelligence (AI) over professional gamers in StarCraft II, may prompt many to wonder why Google’s parent company, Alphabet, is investing hundreds of millions of dollars a year to play games.
While DeepMind innovates in many fields, games like StarCraft and Go before that, demonstrate a computer’s ability to be intuitive. In this case, intuition means that the computer is able to act unconsciously, non-rationally, and quickly, surpassing ordinary processing to deeply understand the information and the situation at hand. Given that these games have a nearly infinite number of moves, DeepMind’s successes show that the AI is aware of its environment and of other players.
This intuition is valuable beyond the gaming environment and is expected to be replicable in many aspects of society. It will be something that people will eventually interact with on a daily basis in their jobs. Just as one can play alongside the computer in a game, the same will be possible in the workplace.
In developing this technology, the AI improves through imitation learning, which imitates past trajectories, and self-play, observing the play of people and then playing against different versions of itself, each time learning from the best player in previous iterations. It’s much the same as humans learn, except that the AI learns much faster.
It’s not as simple or straightforward as it may seem. Given the escalating increase in computing power, this most recent DeepMind endeavour may be the largest scale AI project that has ever existed, requiring a tremendous amount of talent, computing power, and money. Without the resources of Google behind it, such an accomplishment would be unlikely.
While DeepMind is focused on AI research, other companies are working to leverage such progress and innovation to create everyday uses, particularly in the enterprise space. AI is already used by companies to optimize their organizations, and provide data that influences how roles are performed.
In observing how players win at games, the AI learns from their successes and failures. Similarly, AI will be used to identify how employees succeed and make mistakes in their jobs. The AI will then be able to help employees in preventing mistakes before they happen. This could be as simple as making sure that all aspects of a form are filled out correctly, or as complex as creating the foundation for an investment bank’s financial model.
This technology won’t remove the human component entirely, but it does have the potential to automate a significant portion of the work involved. In observing how the best employees do their job, AI will be able to create a framework for how that job should be performed. Just as AI in a video game can observe winning players’ behavior and then use the most effective strategies and tactics to win the game.
The difference, of course, is that real life isn’t a game. People will still play an important part in all functions within a business. The real-world changes constantly, and that requires someone to make sure that any AI-created framework is suitable for particular circumstances.
At the same time, even video games are dynamic and subject to variation, and DeepMind has demonstrated that AI is capable of reacting to and excelling at changing conditions. Advances in AI by DeepMind have so far been used to help clinicians and patients, so it will be exciting to see how their latest victories in gaming will transform other sectors and other jobs.