For #transformationtuesdays I’m sharing one of the most common use cases for Machine Learning that a lot of people who are young or young at heart can relate to – computer games.
For today’s example, it’s how AI Bots can be better than humans at playing Starcraft.
Why Starcraft? Because its complexity and the multitude of factors to consider that affect gameplay is closer to the number of factors a human brain considers in real life when making decisions, as compared to other games like Chess for example, where players take turns.
Why is this important? Because implementing Machine Learning Algorithms on games like Starcraft will allow us to learn more about how we can build better Bots that will be applicable to the real world.
Read on for more details and to see the actual use case.
Before I show you the use case, allow me to give you an idea just how complex it is to build one of these Bots that can beat human players in Starcraft.
To give you some perspective, Google’s computer processors that crunch through Machine Learning Algorithms like Google Translate are called TPUs (Tensor Processing Units).
To beat world-class chess players, the AlphaZero algorithm needs 4 TPUs.
On the other hand, AlphaStar, the algorithm designed to beat professional Starcraft players, uses a whopping 32 TPUs! Thats a lot of processing power.
I’m sure by now you’re interested to find out more about the use case, so I won’t delay it any further. Check it out now.
Why Starcraft is the perfect battleground for testing AI
How it worked is they trained the Machine Learning algorithm by having it watch 44 days worth of human game replays for each of the 3 Starcraft races – Protoss, Zergs, and Terrans, for it to identify patterns and develop its own gameplay strategies.
It got so good, that what they’re looking at now is how to reduce the required processing power to accomplish the same tasks.
Taking on challenges like this teaches us humans on how we can be better at creating new AI tools that work in a multitude of environments and with a lot of variables, which are closer to the number of complexities and factors humans actually experience and deal with in the real world.
Ending Note
I hope you found this to be of value.
If you have any questions, reply with a comment. I’d love to hear from you.
And if you think this is helpful and you’d want to get more, then subscribe for updates, and get a free copy of my book – The Business Optimization Blueprint.