Here are some of the games that are currently being used to make AI more intelligent

For all of the hype around artificial intelligence, many people may feel like they have been waiting for a future that will never actually arrive.

We can all remember the hype earlier in the last decade, where every think-piece and op-ed imaginable would sing the praises of AI, or conversely, issue stark warnings about the dystopian future we would inherit in due course as everything becomes automated.

Neither prophecy has come true, and most parts of life and the economy may feel untouched by AI to much of the general public.

However, there is still plenty of reason to be excited about this emerging technology, even if it isn't rolling out as quickly as some tech gurus might have hoped. According to this article in Info World, AI has already made huge strides in fields such as logistics, data analysis, and autonomous vehicles.

One of the most effective ways to train AI algorithms remains game testing, in which they are subjected to popular games enjoyed by humans to see if they can beat us. With that in mind, here are some of the games that are currently being used to make AI more intelligent.

Chess

When one thinks of AI going head-to-head with humanity, chess is likely the first game to come to mind. Chess has long been considered to be the gold standard when it comes to any measure of intelligence, which is precisely why we elevate the most skilled players to the level of "grandmaster" and shower them with prizes and accolades.

The first major milestone in this genre of AI testing came all the way back in 1996 when IBM's supercomputer Deep Blue bested the world's reigning chess grandmaster, Garry Kasparov, at a tournament watched by millions in Philadelphia. Since then, more sophisticated AI algorithms have been programmed to process thousands of moves per second, and have been able to comfortably defeat the world's greatest chess grandmasters time and time again.

Go

In the years following AI's triumph over chess, experts focused on the classic strategy board game Go as the next milestone. Due to the nature of the game, which is incredibly complex and involves a very long-term, holistic strategy to win, few people thought that AI was anywhere close to beating the world's top Go players.

However, that all changed in 2016 when Google's DeepMind AI defeated South Korean Go champion Lee Se-dol four games to one. A few years later, Se-dol threw in the towel and quit the Go scene entirely, commenting that it was not worth even trying to play in a world where AI could predict your every move and counter it accordingly. For many AI watchers, this moment was a true game-changer.

Blackjack

Few people think of blackjack as being a game that has much use in the development of AI. After all, this guide to automated online blackjack explains how each move is mostly down to pure luck, with the player being able to make only marginal decisions on their next move based on the hand they are dealt. In addition, the guide explains exactly the odds of blackjack and how the house edge works in this classic casino game.

However, this guide from Towards Data Science demonstrates how blackjack has become increasingly useful to train so-called neural networks, which can reduce the house edge by always making the optimal move based on the current hand that the AI has been dealt. By doing this, AI has consistently been able to beat the house at blackjack, although its success rate is nowhere near as high as it is for pure strategy games like chess and Go.

Backgammon

Backgammon is one of the oldest board games still in use today, with hundreds of years of history. The game continues to hold the position of the national game for a wide range of countries, including Egypt, Greece, Lebanon, and Israel. The game has also played a pivotal role in one of the most important AI advances in history.

Back in 1992, Gerald Tesauro from IBM developed his own backgammon algorithm called TD-Gammon. The algorithm incorporated a then-unused learning mechanism known as Temporal Difference (TD), which is now widely used in AI applications around the world. Using this form of deep learning, the algorithm was able to develop a series of backgammon moves that had not been recorded at any point in history, meaning that it was essentially able to "invent" knowledge of its very own. The applications since then have been significant.

If you're planning on training up an AI to think for itself and develop new and reactive responses to external stimuli, then these classic games are a great place to start. By pitting your AI against these games, it can develop important features and applications that can be used across a wide variety of use cases.