For some time now, Artificial Intelligence has been cited as being the next great technological leap forward. Some fear it, others embrace its possibilities, but many of us have yet to see it really in action.

So there has been considerable excitement with the news that researchers at Sony have developed an AI program called GT Sophy that has mastered how to beat 95% of human opponents on the hugely popular PlayStation game Gran Turismo.

Although computer programs have existed for a long time that can take on and beat us at everything from chess to poker, the dynamic nature of a driving game like this meant that mastering it was an altogether trickier proposition for a piece of software.

In fact, the process turned out to be quite a simple one. GT Sophy was pitted against experienced and skilled players of the game and had to gradually build up experience-based knowledge of how to win. As it turned out, within two days it had mastered much of the strategy and showed itself capable of beating all but the very best opponents.

It then continued to improve, gradually cutting more and more time off each lap. While this experiment is obviously very exciting news for the scientists looking into the applications of AI, it’s likely to be equally useful for games developers. That’s because the knowledge gained will help to make single-player games even more realistic.

It may even spill over into other forms of gaming such as the kinds of online slots UK players spent an estimated £2.9 billion on between April 2020 and March 2021 according to official figures. With new titles being released almost weekly and games developers committed to creating ever better and more immersive playing experiences, they will undoubtedly be interested in exploring how AI can help them to achieve this. For example, just like video games and movies, new slots have planned release dates. By profiling the kinds of games that particular players favor, AI could ensure that details of an imminent new game are brought to their attention just before launch.

What has made this experiment into AI so interesting is that GT Sophy also had to work out the optimum strategy to not just beat opponents, but also to continue reducing lap times. This encompassed mastering skills including choosing the optimum time to overtake as well as selecting the best line to take when cornering. Compared with the more logic-based decision-making of chess and poker computers, this called for a far more nuanced approach. The fact that the game also imposes penalties for infringements meant that yet another variable had to be included in the calculations.

Beyond the gaming world altogether, the experiment also has great potential significance for the future of self-driving cars. There are hopes that it can be applied not to enable driving at speed but rather to make sure the car is always being driven in the safest and most economical way, constantly refining and improving based on learned experiences.

And who knows? Perhaps the actual world of motor racing will one day feature driverless cars with the programmers and developers being celebrated on the winner’s podium instead of the traditional champagne-spraying drivers!

Categorized in: