The AI Sandbox

Dig Into Monte-Carlo Tree Search is now Available!

It's an exciting time to be involved in artificial intelligence! New algorithms are surfacing after decades of research, modern hardware is making a huge impact, and companies/developers are taking it more seriously than ever. One of those emerging algorithms is Monte-Carlo Tree Search, or MCTS for short.

This algorithm was discovered in 2006, and its application to board games has been extremely succesful so far. Now MCTS is making its way into AAA video games and simulations too. Over the next few days, we'll be digging into MCTS in depth as we release The AI Sandbox™.

Bonuses Available for 2014-8-29 12:00:00 GMT+00:00...

Yes, it's finally launch week! Thanks to those of you who helped with testing and feedback in our Community Alpha, and everyone else for your patience. All details are revealed; find out more about The AI Sandbox™ and our first release called Dig Into MCTS below...

1. So Why MCTS?

2. Common Challenges

3. About The AI Sandbox

This video covers the current state of AI, where and why Monte-Carlo Tree Search fits in, the inspiration for our work on The AI Sandbox™, and what's included in Dig Into MCTS.

Launched on Wednesday, August 27th

1. Why Monte-Carlo Tree Search?

The MCTS algorithm is the first simulation to be released within The AI Sandbox™ and the team has spent many months on the topic — developing, refining and tuning the implementation.

For what reason should you dig into Monte-Carlo Tree Search? We've covered at least four great reasons in the past few videos...

Powerful Solution
MCTS has revolutionized search algorithms in board games, particularly Go, and is now being applied into AAA real-time strategy games via TOTAL WAR: ROME II.
Innovative Algorithm
Monte-Carlo Tree Search was only discovered in 2006, but has already seen serious adoption in research and now industry. It's cutting edge, but just the tip of the iceberg!
Future Applications
Beyond what developers are doing now, MCTS is proving to be such a powerful algorithm that it has huge potential for other applications. It's a great career move to make now.
Fun at Your Fingertips
As an algorithm with some emergent properties, Monte-Carlo Tree Search is also extremely fun to play with — and provided us with hours of entertainment :-)

In short, MCTS seems like the ideal algorithm for AI enthusiasts! It's certainly been keeping us busy over the weekends and evenings too...

2. Common Challenges

Part of the challenge applying any advanced artificial intelligence solution is that it can take significant time and effort to get up-and-running. Over the many years since we started our research projects, we've bumped into many problems. (It can be both depressing and annoying.)

Here's a quick guide — partly as a warning — of things you'll likely encounter:

  1. Research Frameworks — Most likely you need a quick survey of possible solutions!
  2. Building Dependencies — Setting up your system isn't difficult, but it takes time...
  3. Creating Environment — Now finally start building a world... from scratch.
  4. Implementing Tools — You'll need some utilities to understand your own AI.
  5. Fixing Odd Bugs — Once you start pushing your system, things will break.
  6. Maintenance Work — Regularly, you'll need to perform updates or apply patches.

Luckily, the state of modern game engines is improving fast and there are many libraries & tools for scientific computing too. This reduces the time you'll spend on infrastructure to months, depending how many compromises you're willing to make!

3. The AI Sandbox

Over the past years, we've been experimenting with a variety of artificial intelligence techniques and applying them into games. We made many mistakes in the process, as it took a lot more work to build certain prototypes that we would have hoped... but at lest we suffered so you don't have to!

In the process, we've built up an engine and supporting tools that have significantly reduced our turn-around times for building prototypes or doing research. It's broken down into three parts:

Simulation Engine
Combining open source graphics and physics engines implemented in C++, it's a good way to build efficient and interactive 3D simulations.
Scientific Computing
Languages such as Python are revolutionizing modern science, providing libraries for data-mining, machine learning, or search-based AI.
Visualization Tools
Browser-based graphics toolkits that display interactive figures with SVG and Javascript are making it very easy to visualize algorithms and their results.

Combining these three separate worlds together has been an interesting challenge! But it's also been a very rewarding process...

4. Dig into Monte-Carlo Tree Search

The first simulation we're releasing within The AI Sandbox™ focuses on Monte-Carlo Tree Search. The product itself has multiple levels, depending how involved you want to get!

Dig Into MCTS — Spade (€19)
  • Content: Interactive tutorial in a dozen notebooks.
  • Source: “Coordinated Search” demo.
  • Access: “Monte-Carlo Tree Search” module.
  • Access: The AI Sandbox™ Community Alpha
Dig Into MCTS — Tractor (€39)
  • Content: Full course as recorded videos. [September 3rd]
  • Source: “Monte-Carlo Tree Search” module.
  • Everything as above from previous Spade level.
Dig Into MCTS — Bulldozer (€79)
  • Content: Case Studies from games as in-depth interviews.
  • Source: “Hunt The Flag Carrier” demo.
  • Everything as above from previous Tractor level.
Dig Into MCTS — Excavator (€149)
  • Content: Conversations with the experts (audio). [September 1st]
  • Support: 25 minutes Skype consulting and advice. [September]
  • Access: Email assistance on the topic of MCTS.
  • Everything as above from previous Bulldozer level.

Note: The AI Sandbox supports Windows (from 7 to 8.1) and Linux (as .deb on Ubuntu officially).

Launched on Wednesday, August 27th

Dig Into Monte-Carlo Tree Search

Enter your primary email address, and we'll send you everything as it becomes available this week:

NOTE: Make sure you click the confirmation link once you receive our email...