Arago's Hans-Christian Boos Talks While Their AI Plays Freeciv at Disrupt London 2016 | Summary and Q&A

Summary

This panel discussion is about playing the game Civilization using artificial intelligence (AI). The guest, Christiane Boos from Arago, explains how they have developed an AI platform that can play games like Civilization by learning from human players and making decisions based on the information available to the player. The AI platform is a combination of machine reasoning and machine learning, and they have used it not only for gaming but also for IT system administration and other business processes.

Questions & Answers

Q: How does the AI platform play the game Civilization?

The AI platform only sees what the player sees and acts based on what the player could act on. It starts by learning the basic gameplay mechanics and then learns from human players to improve its strategy. It can adapt to different situations and make decisions based on the knowledge acquired from human players.

Q: Why did Arago choose to use the free version of Civilization?

Arago chose the free version of Civilization because it had an open API, which made it easier for them to integrate their AI platform with the game without spending too much time on programming interfaces. They wanted to focus more on learning the gameplay and improving their AI's strategy.

Q: How did Arago teach the AI platform to play the game?

Arago started by teaching the AI platform the basic gameplay mechanics, such as the map and cities. They then reached out to the free Civ community and asked for their help in teaching the AI platform. The community provided valuable feedback and experience, which helped improve the AI's gameplay. Now, they are playing against the top players in the community and learning from them.

Q: What were some unexpected things that the AI platform did while playing the game?

The AI platform did some unexpected things, such as building a trade route from Pyongyang to Washington and putting a submarine into Lake Michigan to counter a terrorist attack. These actions were based on the AI's reasoning and strategy, which sometimes resulted in unconventional decisions. However, through post-mortem analysis, they were able to understand why the AI made those choices and improve its gameplay.

Q: Can two AI platforms play against each other?

Yes, two AI platforms can play against each other. Arago uses this approach for machine learning purposes. They can have multiple AI platforms play against each other with the same or different knowledge sets to evaluate which players give the best experience for different game scenarios. It helps them understand the effectiveness of their AI in different situations.

Q: What does the future of AI look like in the next few decades?

According to Christiane, the future of AI will involve embedding AI into various business processes. They have already seen success in IT automation and believe that many other business processes can be automated using AI. The focus will be on teaching AI systems specific knowledge and allowing them to transfer that knowledge between different domains. Natural language processing is still a challenge, but progress is being made in that area.

Q: How has using AI in gaming benefited Arago's business?

Using AI in gaming has been beneficial for Arago's business because it has provided them with a platform to showcase the cognitive abilities of their software. The short feedback cycles and the ability to demonstrate the AI's decision-making capabilities have impressed their business clients. It has also allowed them to attract potential customers who are interested in embedding AI into their own products.

Q: Besides gaming, where else has Arago used their AI platform in business?

Arago has used their AI platform in real estate decision-making with commercial clients. They have found that the decision-making process in real estate is similar to that in the game Civilization, making it an ideal use case for their AI platform. They anticipate that the knowledge and experience gained from gaming can be applied to other business processes, similar to how Minecraft is used in education.

Q: Can the AI platform be used to provide strategy for other games like football?

Yes, the AI platform can be used to analyze and provide strategy for other games. The flexibility of the AI platform allows it to handle randomness, which is a key feature of games like Civilization. While Civilization focuses on strategy, the AI platform can be trained to adapt to different game mechanics and provide guidance based on the knowledge it has acquired.

Q: What is the true test of an AI system?

Christiane believes that the true test of an AI system is its ability to understand and adapt to human values. While natural language processing remains a challenge, the goal is to teach AI systems human values so that they can make decisions aligned with those values. The focus is on automating various business processes and providing cognitive abilities to AI systems.

Takeaways

The panel discussion highlighted how AI can be used in gaming to not only improve gameplay but also demonstrate the potential of AI in other business processes. By learning from human players, the AI platform developed by Arago showed impressive decision-making abilities in the game Civilization. The future of AI is expected to involve embedding AI into different business processes and teaching AI systems human values. While challenges remain, progress is being made in automating various tasks and providing cognitive abilities to AI systems.