Minecraft is no stranger when it comes to be used for educational purposes. As you might recall, back in May, Microsoft created a new version of the game called Minecraft: Education Edition, which was designed for teachers to use in the classroom.
Well, as it turns out, the game can also be used to teach more than just humans: it can teach artificial intelligence, too. As evidence of that, Microsoft announced a public release for Project Malmo, the company's system for testing AI software in Minecraft.
The concept of using Minecraft to train AI might seem new to some people, but Microsoft has been engaging in this practice for quite some time. Back in March, Microsoft unveiled Project Malmo, known as Project AIX at the time, and other than the name, the key difference between then and now is that only a few academic researchers had access to it.
That now changes with the public introduction of Project Malmo, which allows anyone to modify the underlying code of the Minecraft server and introduce AI elements to the virtual world. All you have to do is access the code, which is readily available under an open-source license on GitHub.
Of course, even with the general public behind the wheel, the advantages of this are still the same. Minecraft provides a safer and more meaningful environment than anything that the real world could ever offer. For example, researchers trying to teach a robot to climb a hill would have to replace or repair it every time it fell into a river. However, if the same happens in Minecraft, all the user has to do is to redo the test.
Furthermore, since the Minecraft server controls the perceived passage of time, researchers can speed up their simulation if time is not an important variable in the experiment. In addition, since everything is simulated, the server can measure and monitor every detail for later, replay and analyze it, making it easier for other researchers to reproduce published results.
So, what is the point of all this? Yes, as mentioned before, part of it is for the AI to learn and hopefully be able to hold conversations, make decisions and complete complex tasks in the future. However, more importantly, the goal is to have the AI learn about the meaning behind those actions. AI is getting increasingly proficient at talking and parsing human language, written and spoken, but generally is just parroting it off and has no clue of what it means. Throwing it into a simulated environment where it can associate words and actions will give it the opportunity to learn what those words really mean, and hopefully, make the intelligence a little less artificial.
