Best Games for making Artificial Intelligence Better: Despite all the hype around artificial intelligence, many people may feel like they have been waiting for a future that will never really come.
We can all remember the hype at the turn of the last decade, where every opinion piece and op-ed imaginable either sang the praises of AI or, conversely, issued stark warnings about the dystopian future we would inherit in due course as we everything becomes automated.
Neither prophecy has come true, and most parts of life and the economy can feel untouched by AI for much of the general public.
However, there are still plenty of reasons to be excited about this emerging technology, even if it’s not rolling out as quickly as some tech gurus had hoped. According to this article in Info World, AI has already made great strides in fields like logistics, data analytics, and autonomous vehicles.
One of the most effective ways to train AI algorithms remains gaming testing, where they are put through popular games enjoyed by humans to see if they can beat us. With that in mind, here are some of the games currently being used to make AI smarter.
When one thinks of AI going head-to-head with humanity, chess is probably the first game that comes to mind. Chess has long been considered the gold standard when it comes to any measure of intelligence, which is why we elevate the most skilled players to the level of “grandmaster” and shower them with awards and accolades.
The first major milestone in this genre of AI testing dates back to 1996, when IBM’s Deep Blue supercomputer bested the world’s reigning chess grandmaster, Garry Kasparov, in 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 handily defeat the world’s chess grandmasters time and time again.
In the years after AI triumphed over chess, pundits 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 the AI came close to beating the best Go players in the world.
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 abandoned the Go scene altogether, commenting that it wasn’t even worth trying to play in a world where the AI could predict your every move and counter accordingly. For many AI watchers, this moment was a real game changer.
Few people think that blackjack is a game that has much use in AI development. After all, this guide to automated online blackjack explains how every move is mostly down to sheer luck, and the player can only make marginal decisions on their next move based on the hand they are dealt. Furthermore, the guide explains exactly the odds of blackjack and how the house edge works in this classic casino game.
However, this Towards Data Science guide demonstrates how blackjack has become increasingly useful for training so-called neural networks, which can reduce the house edge by always making the optimal move based on the current hand being played. has distributed the AI. By doing this, the AI has always been able to beat the house at blackjack, although its success rate is not as high as pure strategy games like Chess and Go.
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 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 biggest AI advances in history.
In 1992, Gerald Thesaurus of 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” its own knowledge of it. Applications since then have been significant.
If you plan to train an AI to think for itself and develop new, reactive responses to external stimuli, these classic games are a great place to start. By pitting your AI against these games, you can develop important features and applications that can be used in a wide variety of use cases.
YOU MAY ALSO LIKE:
Source: Analytics Insight