Artificial intelligence (AI) can be used for myriad of things. Traditionally, it has been used to take on repetitive tasks, like in manufacturing. We’ve seen AI beating game masters and now we have virtual assistants on our smartphones and in our homes. But there is a lot of research and experimentation being done right now to enhance audience’s entertainment experiences too.
Aside from AI challenging board and digital game champions (Chess, Go, Miss Pacman, DOTA 2 etc.) recently very little has been done to truly exploit all what AI can do to really amaze audiences. In this introduction session, we’ll explore what is artificial intelligence, what are its limitations and how it works.
There are two types of artificial intelligence (AI):
Companies however love to call their applications “artificially intelligent” or “smart” but up until now, very few products have AI running. Here is how you can tell if an advertised AI qualifies as narrow AI: it learns from contact with users. Nowadays, this unfortunately means you need to know whether the AI advertised is using some form of machine learning algorithm in its programming to improve its performance as it does the job it is designed to do.
At its core machine learning is the programming that allows AI software to learn how to do its job efficiently. Programming machine learning is art just as much as it is technical. Some types of algorithms used to achieve machine learning are artificial neural networks and deep learning. Various mathematical data analysis methods are used as well from decision trees, to regression analysis to clustering algorithms. The AI algorithms learn by trial and error through a programmed feedback process that often includes a human being directing the learning process. With clever programming, the learning process, or conversion, can take hold on its own without the presence of human feedback and iterated several million times until a certain acceptable success threshold is reached. Often, that threshold is the success rate of a human expert in the same area of expertise.
Machine learning process:
Narrow AI is present in the entertainment space but for the time being, mostly experimentally and in beta phases. There are some examples of AI in the music industry, creating full sheet music based on an author’s specifications and AI musician robots capable of improvising based on the style being played.
In the arts, IBM’s Watson is known as an art critique and Watson has even been involved in the creation and identification of baking recipes.
In the game industry, several recent attempts to beat niche game champions through specialized AI met with success, including the AlphaGo artificial intelligence that beat the World’s Go champion, Microsoft’s AI beating one of the hardest video games ever made, Miss Pacman, and Elon Musk’s OpenAI beating DOTA 2 world champions. We should see some level of narrow AI in the next generation games thanks to these experimental efforts.
AI will soon infiltrate all entertainment media to both enhance audience experiences and to adapt to individual audience members in its performance.