Right now, when it comes to artificial intelligence, it seems impossible to avoid debating the impact AI's rise will have on humans.
In a previous post, we highlighted AI company PublicAI's plans to provide new jobs for humans post-singularity. This includes the launching a test version of the product that attracted half a million users within four months. Focused on providing AI training data services with blockchain technology, the platform uses cryptocurrency to incentivize data workers to contribute and annotate data to meet the rapid rise in demand for data across various industries.
As we continue to ponder the future of human-AI relations in the future, however, it's important to consider other aspects of our lifestyle and work. This includes the video game industry and the potential impact AI will have on gamers and gaming enthusiasts.
At its current pace, depending on the game, many artificial intelligence systems can easily defeat their human counterparts. Meta's CICERO was found to be a deceitful gamer in test runs on board game Diplomacy, placing it in the top 10% of human players. Of course, CICERO isn't the first of its kind. Previously, DeepMind's AlphaStar, an AI system specifically designed to play real-time strategy game Starcraft II, was able to take full advantage of the game's fog-of-war mechanic to feint, making human players think it was going one way when it was really going the other.
In fact, DeepMind's AI went on to achieve the highest in-game rank, besting human players by learning through a variety of algorithms. Below, we'll take a closer look at what Starcraft is, how the AI that ranked Grandmaster did it, and the implications of AI in gaming:
What is Starcraft?
Initially called "Warcraft in space" by gamers, Starcraft quickly rose to fame as the real-time strategy game in the gaming community. Developed by Blizzard Entertainment (News - Alert), the game is free-to-play with military and science fiction elements. After its initial release in 1998, Blizzard released its sequel, Starcraft II, in 2010. Despite fans being starved of new installments or remakes since the sequel's release, the game remains an esports staple due to its competitive nature. The game is also one of 22 games featured in the ongoing Esports World Cup held in Riyadh, Saudi Arabia.
Nowadays, many gamers, fans, and viewers actively participate in Starcraft betting via online platforms. While professional players attempt to gain superiority over their opponents, viewers can place bets using various odds, from American and decimal to fractional. Nowadays, fans can even bet using cryptocurrency.
In order to improve their odds of winning, fans, bettors, and players alike watch content creators and professionals within the Starcraft niche. As of July 2024, Starcraft's Twitch viewership may not seem as impressive compared to other competitive and esports titles. The game reached peak viewership in July 2022 at 45,699 and has a smaller but dedicated number of channels and content creators.
Starcraft's AI Grandmaster
Now that we know what Starcraft is, it's important to understand how AlphaStar, the AI that achieved a Grandmaster level in the game, did it. In 2019, Google (News - Alert) DeepMind's AlphaStar team detailed its findings. A preliminary version of AlphaStar in January 2019 challenged two of the world's top players in Starcraft II. Afterwards, the DeepMind team focused on pushing the AI to play the full game at a Grandmaster level under professionally approved conditions.
To achieve this, the team used various general purpose machine learning techniques. This included neural networks, self-play via reinforcement learning, multi-agent learning, and imitation learning. Therefore, the AlphaStar AI was able to learn directly from game data. Later on, AlphaStar ranked above 99.8% of active players on Blizzard's Battle.net platform. The AI was also able to achieve a Grandmaster rank for all three Starcraft II races: Protoss, Terran, and Zerg. To conclude their real-time strategy experiment, the DeepMind team states that they expect these methods to be applied to many other domains.
The implications of AI in gaming
So now you may be thinking, so what? It's important to note that AI in gaming isn't a wholly new concept. Casual and competitive chess players have struggled against their AI counterparts for decades now, and lots of modern games use some sort of AI system or algorithm to enhance, improve, or tailor gameplay to different players.
In terms of Starcraft II specifically, however, the AI experiments may lead to more real-life consequences than we think. Namely, the US Army recently announced they were looking into AI to enhance their battle planning capabilities. There is a particular focus on AI's knack for battle planning in Starcraft II here, with recent experiments having showcased "promising results." At the same time, however, experts remain cautious about the tech's application in real-world combat situations. A study looking at AI models such as OpenAI, Anthropic, and Meta found that these AI models had an inclination for rapidly escalating conflicts, which can sometimes lead to the deployment of nuclear weapons.
Ultimately, there's no telling the future of AI in video gaming. At the same time, insights on what can be drawn from AI trying to learn to get good at video games. However, results continue to impress, and it'll be interesting to see further real-life implications of this development in the future.
