The impact of Artificial Intelligence in the esport and Gaming industry.
Ai will very soon have an outstanding impact on the esport industry and the way pro-gamers train for competition and how they can progress in the long and short-term. Eventually, those AI could turn into robots and train athletes from different traditional sports such as basketball, soccer and such. The advancement of technology will allow to gather and process data at a rapid speed and will allow a professional athlete to train in a much more efficient way which could bring the level of competitiveness up.
The paper seeks to explore the role that AI will take within the gaming and esport industries. In doing so, I will explore the world of gaming and what has already been done in AI and how both realms can be combined to help the growth of the exploding esport ecosystem. I will start by looking at the work from OpenAI that recently showcased a demo of their AI at a video game championship tournament and look into what would be necessary to build an AI that will learn every game on the book. The goal here is to find out the roadmap of AI in gaming and how esport organizations will use it in their businesses, as a training tool for example.
The esport industry has been growing tremendously in the last 5–10 years with the explosion of gaming and the rise of competitive online platforms. The best gamers in skills are now considered athletes and are now contracted professional gamers for organizations based all around the world. Today, only a few companies have a professional league and a stable ecosystem around them. Riot Games with League of Legends and Valve with Counter-strike Global Offensive are the two biggest names in the scene today and will be the example I’ll use for my essay. For the AI part of the project, I will explore the realm of OpenAI, a recently launched startup by Elon Musk with a mission to explore “safe” AI. Being the first and only company to show results or testing to the public, I will talk about how they got started in this field, where they are now and what direction they should follow.
Nowadays, esport is its own industry with its teams, athletes, and fans. The industry has almost gathered $1 billion in revenue with an audience size of 380 million worldwide and these number are projected to double within 2022. Different games have reached the stage of having their own professional league such as League of Legends and Counter-Strike: Global Offensive (CSGO). The way professional gamers are trained today isn’t optimal and is usually resulting in quick burnout from very talented players. These days, pro-gamers train by playing the game a lot but they don’t get the best return of the investment of their time since they match with different players all the time and for them to gather data and feedback, nothing is very out there that is optimal. There is definitely a gap about how pro-gamer train and how traditional sports athlete train every day. A boxer or a basketball player would train less time in terms of hours but will be making the most of every minute since so much data will be generated and tracking tools will help athletes for future competitions. Stephen Curry, for example, has a team of analytics that gathers the data of his shots, passes, rebounds, etc.. and they make sure that he works on his specific weaknesses during a given practice but in gaming, the whole spectrum is different. To train a game like League of Legends, you have to get matched with 9 other players and play a full 40–50 minutes game non-stop and every game is different. It’s different in a way that you might be playing a different character, you may be playing against a different character, the rules and factors are always different. Today if I want to train my 3-point shots, I just need a ball and a basket but with gaming, things are just too complicated to be isolated into specific tasks. An esport organization will have an entire coaching staff to support its pro-gamers, a gameplay, an analyst, a mental and a physical coach for the biggest teams out there. The way training works is really similar to what we see in traditional sport, gamers train for hours and hours, review their gameplay and train to work on their weaknesses to eventually become better. It takes years for gamers to master a game and as the game continues to change with patches and updates, the same pros have to keep up with an extremely high pace of practice runs. Whereas I don’t think that AI will replace the pro-gamers as this will have very low entertainment value, I think it could drastically affect the efficiency of training sessions and make competitiveness higher.
According to esport industry experts, AI will be one the major disrupting tool in the realm of competitive gaming. Ai is the most exciting thing for pro-gamers since it will change the whole training process for each of them and make everyone’s time more efficient and limit the burn-out scenarios from young talent that are today playing 8 to 10 hours every day.
OpenAI is a non-profit artificial intelligence research company that is currently developing AI in different areas. The company wants to tackle gaming as one of his main projects because it could be of a great use and a great example of how AI could be taken advantage of in tomorrow’s world in a “safe” way. They began their research by teaching the AI chess until it beat the world champions and then they targeted videos games starting with Dota2, which is a direct competitor to League of Legends as they are both from the same genre. Last year, at The International 2017 which is basically the Dota2 World Cup, OpenAI showcased its AI and challenged the best Dota2 players in a 1v1 with a mirror matchup (both the AI and the human had to play the same champion/character). To everyone’s surprise, especially the pros, the AI won every single match with perfect mechanics and ultra-fast learning of the enemy movement and plays. The most remarkable thing was the AI’s capability to adapt to the opponent’s game style and win with a minimum of casualties. The power of machine learning is how things were capable, the processing of massive data and the learning aspect of the game was done with this technology, the bot literally learned the game by itself in a matter of months. Machine learning is done with massive data processing and thanks to gaming engines, data has already been gathered and stored which is one of the reasons why Elon Musk wanted to explore AI with gaming to start with. That being said, Elon Musk and the OpenAI team want to take another approach to machine learning that I found fascinating. According to the Open AI blog, the program “does not use imitation learning” which is one of the most common machine learning approaches out there. The reason why this is important for gaming is that it allows robots to learn games on their own and not necessarily copy moves from pros that are only good at a certain standard. For games like chess, imitation learning makes sense since the game can be broken down to a certain number of algorithms but for video games, things are much more complex and therefore should be treated differently.
As much as this is a major milestone, Elon Musk wants to push things to the next level and build a team of 5 AIs to play against a team of 5 pros, because Dota2 is originally a 5 versus 5 game. The company hasn’t shared any details on how they will process the synergy between artificial intelligence and how the communication will occur between one another, but I found it fascinating to think about the fact that AIs will use each other and synergize to take on a full team.
Games like Dota2 require teamwork to win as it is impossible to win on your own against 5, whereas machine learning takes care of the bots individually, something will have to teach them how to work together and build momentum together. As I previously mentioned, the company didn’t share the technology they will use to build a team of bots but I believe this could be achieved from a more elaborated machine learning where each bot will just take into consideration its teammates movements and actions and try to capitalize on certain actions or that a single AI will control 5 characters in the game and work them like puppets to ultimately win the game. Both of those approaches are being discussed on different forums on the topic and I think both are plausible to certain extinct. That being said, when it comes to knowing which will be more efficient and “stronger” in the game, I don’t know the answer.
Now that we understand better the spectrum of how Artificial intelligence will soon impact Esport, I want to know if esport organization will see value in it being a training tool. That being said, the solo AI is already an amazing tool for pro-gamers to train duel mechanics and learn from their mistakes because the most inconvenient part in training is that you never want to give information about what champion you want to practice to the enemy team before the big day and the fact that now someone can practice with a computer is increasing the competitiveness and the element of surprise during tournaments. Once OpenAI gets to create a team of AI in data to compete and assuming that it works, that they win a full game going through all the phases the game has to offer, then everything will be different from now on.
OpenAI will probably license this technology to game and esport organizations and become the best training tool pro-gamers could ask for. It will adapt as the game adapts (updates, patches, etc.) and learn game style from different pro-gamer in the scene giving you the opportunity to train against the best in the scene but this isn’t for tomorrow. Also, OpenAI targeted gaming as a challenge but their roadmap has ultimately nothing to do with gaming and esport as they just want to explore ways AI could be used in a safe environment.
According to some research I conducted online on platforms like Reddit and Quora, some self-proclaimed expert about the topic and gaming are very skeptical about the mission from OpenAI, some define it as a marketing stunt to impress people and gained traction to the company. The first argument that I want to share from them is the fact that OpenAI Dota2 achievement was done in a one versus one context with a mirror matchup as a stated earlier but the champion that was chosen has the easiest mechanics that the game can offer and Dota2 happens to have more than 100 different champions ranging from easy to extremely hard with complex mechanics and very versatile gameplay. Whereas the company did reach a milestone beating pros, they are far from beating an entire team with different champions on the map as this will multiply the amount of data to be learned by an immense number. The second argument that I thought was interesting was that the robot doesn’t really know how to play the game of Dota2 but only knows how to play one champion in a specific scenario in a specific timeframe (the duel was limited to 10 minutes). One might criticize the headlines that were made after the event took place and say that it misleads people to understand that OpenAI created a robot that can play the video game like a human but just learn faster. All that are opinions and whereas I agree with comments like such that I found of online forums, Elon Musk gave us all a promise that at the next The International 2018, OpenAI will showcase the promised 5v5 situation against pros. The event will take place August 20th and so far, we don’t have further information about the progress done on the AI.
To conclude, OpenAI has shown great results already creating an AI that could win duels in Dota2. Whereas this is already a great milestone that allows individual skills to be trained, there’s still massive work that needs to be done for Dota2 and other games. Dota2 fills in the MOBA genre of video gaming but what about all the other competitive genres like CSGO (first-person shooter), or Hearthstone (virtual card game). This could obviously become the best if not the only necessary tool for pro-gamers to train on an everyday basis and will impact the industry tremendously. But few questions are still on hold: How long does an AI need to master a game that came out today? Who will have access to the AI? Will the AI be in symbioses with Robotics one day and affect traditional sports the same way (an AI robot that plays like Lebron James, etc…) Although the progression has been remarkable going from beating the best chess player in the world to duels in Dota2, I think there’s still an immense way to go and a lot to learn before esport welcomes AI as its new training tool.
Sources:
OpenAI at The International 2017 from OpenAI Blog
https://blog.openai.com/dota-2/#more
Topic: Esports Market — E-sport industry revenue and audience in 2018 and projections
Steve Fuller — https://www.statista.com/topics/3121/esports-market/
What Openai Victory in Dota 2 Means For the Future Of Esports
Jack Crosbie — https://www.inverse.com/article/35449-elon-musk-dota-2-openai-the-international-dendi-1v1
How did OpenAI create the Dota 2 game bot that beats professionals?
Lucca Fehn, BA Cognitive Computer Science in process.
https://www.quora.com/How-did-OpenAI-create-the-Dota-2-game-bot-that-beats-professionals
