Wednesday, February 15, 2017

First round of General AI Challenge just launched: Gradual Learning – Learning like a Human

  • First 6-month warm-up round just launched
  • Focus on “gradual learning” – because it’s an architecture property that enables gradual accumulation of skills, making learning more efficient
  • $50k in prizes in this round; $5mil in total prizes over the multi-year General AI Challenge

Today, we at GoodAI have launched our first warm-up round of the General AI Challenge.

It’s one of many stepping stones on our mission to develop general artificial intelligence - as fast as possible - to help humanity and understand the universe.

This first round focuses on “gradual learning” – which is the ability to gradually accumulate skills and use existing skills to learn new skills more efficiently.

The reason why we started with “gradual learning” is that we have identified it as an architecture property that will enable the efficient inclusion of additional properties. In other words, if you use your existing knowledge to learn to solve new problems, you should be more efficient than if you always have to start with zero experience.

The gradual learning round is not concerned with how good an agent is at solving a particular task (e.g. highest score in a game). Gradual learning is about how efficient an agent is at learning to solve new and unseen task. Using less training data and fewer computational resources is  among the criteria for better agents.

Gradual learning requires a combination of at least these abilities: compositional learning, meta-learning / learning to learn, continuous learning, life-long learning, learning without forgetting, transfer learning and more. More info about “gradual learning” and other required properties for general AI is available in our framework document.

How this round works:
  • Today, teams start developing their AI / AGI agents
  • They can develop, test and train their agents on training tasks provided by us
  • All these tasks were designed with “graduality” in mind – which means that each task builds on skills acquired in previous tasks; each new task reuses skills learned in previous tasks
  • After 6 months, teams will submit their pre-trained agents / models and code
  • We will start evaluating the agents on non-public evaluation tasks
  • We will test the agent’s ability to learn gradually and to not-forget skills
  • The environment that we use for this round is a version of CommAI-Env. It is based on byte inputs and outputs and has text-like properties.
  • The training tasks are based on the CommAI-mini set recently proposed by Baroni et al., 2017 (
  • It may look like the agent is learning to communicate with the environment, however, our ultimate goal is not to build agents for this kind of environment. We chose it now because this type of environment makes it easier and more intuitive for people to understand why one task builds on top of a previous task. If we chose more complex and noisy environments (e.g. computer games), you would have a much harder time identifying when agents build skills on top of previously acquired skills.
  • However, our plan with the General AI Challenge is to scale to this level of complexity sooner or later.
  • Another reason for this environment was that during our road-mapping process we had identified “learned communication” as one of the essential skills which can speed up the acquisition of more advanced skills, and therefore increase the efficiency of learning (but this is something for later stages anyway)
Where are we aiming for with the General AI Challenge? We have identified a set of open problems that we consider to be key milestones in achieving general AI. Our teams here at GoodAI are working hard on solving these milestones but we also think that “outsourcing” to the greater community of researchers, programmers, and hackers can both speed up this process and diversify the avenues of research. We are also hoping that we may find new talented colleagues within the participating teams.

What if no team passes the evaluation in 6 months? Well, we will probably restart this same round again, perhaps with modified definitions and rules, or maybe slightly different evaluation tasks. There’s also a hope that if the participants get another chance, they will build on top of experience they gained in the first attempt which would eventually lead to  a solution for gradual learning. In principle, the teams would then be gradually learning how to gradually learn :-)

We have allocated $5mil in total prize money for this multi-year challenge. We plan to distribute this pool of money to the participants of various rounds throughout the following years. We hope that we get to general AI before the money runs out :-)

Thank you for reading!

Marek Rosa
CEO and Founder of Keen Software House
CEO, CTO of GoodAI

For more news:
General AI Challenge:
AI Roadmap Institute:
Space Engineers:
Medieval Engineers:

Personal bio: Marek Rosa is the CEO and CTO of GoodAI, a general artificial intelligence R&D company, and the CEO and founder of Keen Software House, an independent game development studio best known for their best-seller Space Engineers (2mil+ copies sold). Both companies are based in Prague, Czech Republic. Marek has been interested in artificial intelligence since childhood. Marek started his career as a programmer but later transitioned to a leadership role. After the success of the Keen Software House titles, Marek was able to personally fund GoodAI, his new general AI research company building human-level artificial intelligence, with $10mil. GoodAI started in January 2014 and has grown to an international team of 20 researchers.

Thursday, December 15, 2016

Space Engineers Entering the Beta Development Stage

Today, we are pushing all the features from last week’s enormous update to the stable branch. Just in case you missed the news, this includes the total block redesign, new multiplayer netcode, a tutorial campaign and much more! We wanted to make this great update available to all players, as you’ve been asking for it all week! 

This update marks the transition of Space Engineers entering into the Beta phase.

Throughout its time in Early Access, Space Engineers has become the first space sandbox game of its kind and size; as a result, we have successfully established ourselves as one of the leaders in the engineering genre. Very few games allow you to build such a huge variety of ships, stations, voxel based planets, and mods. It has one of the biggest workshops on Steam thanks to its amazing modding community, with more than 240,000 items and counting!

At the moment, a great multiplayer experience is our primary goal. We are proud of the strides we’ve made with our newly implemented multiplayer and are planning to improve it even further in the future.  

So what do you need to know about the Space Engineers Beta?

The most important thing to remember with this announcement is that there are still more improvements, optimizations, and contents to come. As an example, I can confirm that one of the things being worked on currently is a new HUD which should significantly increase immersion in game. Beta simply means that we feel the game now has a solid foundation. Some of the new models will be receiving adjustments because of feedback from the community. Thanks to feedback from the Space Engineers community, we were able to clean up many, many bugs and crashes that our players were experiencing regarding multiplayer and online play.

If you are a current player and fan, you can look forward to fewer bugs, more optimizations, an updated user interface (new HUD), and constant improvements in all aspects of the game.
If you would like to help, please leave your reports and log files in the bug section of our forums here:

What if I’m new to Space Engineers?

In case you are a potential new player, there is now a tutorial campaign to help familiarize you with basic gameplay once you purchase the game. Sandbox games can be overwhelming in the beginning because of their sheer scope of possibilities. The campaign experience can now introduce you to the world of Space Engineers comfortably, without the need to read guides or go through clunky tutorial missions. It also serves as a showcase for what all you can achieve with our visual scripting tool. We cannot wait for what will be the golden age for modders in Space Engineers history!

What about us modders?!

If you are a modder, we really want to thank you for sticking with us during this crazy ride. At the beginning of the development we wanted to make our game as mod-friendly as possible. We even published the source code on GitHub for our modders to poke around in and the Space Engineers workshop ended up blowing our minds as a result. There is constant contact between us and our modders. Some of their creations were even implemented into the core game, such as the fighter cockpit, corner lights, and corner LCDs. There may be time for additional mods after we polish and optimize the game during the Beta stage.

We know modding an early access game can be frustrating from time to time, because things are constantly changing and we want to make you happy. From now on you can look forward to a more stable environment for modding as well as official, updated modding guides. We also have plans for cleaning up and polishing the MODApi interface during the Beta stage.

You guys aren’t looking so small these days…

It’s true! Keen Software House has now grown into a company of 50+ employees! However, we still consider ourselves an indie company. Testers can talk to the leads, artists can walk over and discuss something with programmers, designers can call out to producers, etc. We maintain an atmosphere of openness and communication at all times, and enjoy the aspects of being an independent company. The plan is to continue this way now and for the foreseeable future.

The bottom line:

Our goal from the beginning has been to deliver the vision from the minds of our highly talented team to the people. More than anything else, the Beta period will focus on stabilizing and optimizing in order to realize that goal. Keep in mind that this is still a period of development. Over the next few months, we plan to make Space Engineers the best it can be so that we are fully prepared to launch the full version of the game. We have been humbled by the support of our awesome community, and we can’t thank you enough for all the ways you’ve helped make this game a reality. Whether you’re a modder, a user on our forums, someone who’s uploaded content to the workshop, or a fan playing to exercise their creativity- you are the reason we are able to exist. Thank you for allowing us to pave the way for a game like no other. We’re excited for what the future of Space Engineers holds!
Marek Rosa
CEO and Founder of Keen Software House
CEO, CTO of GoodAI

For more game news, follow us on social media:

Medieval Engineers on Facebook:
Medieval Engineers on Twitter:
Space Engineers on Facebook:
Space Engineers on Twitter:

Thursday, December 1, 2016

GoodAI Releases School for AI - A New Update to Brain Simulator

In Brief:
  • We are releasing School for AI, a training and testing environment designed to teach your AI architectures new skills and to measure their performance.
  • Download School for AI, the new update to Brain Simulator.

Today I’m happy to announce a new update to Brain Simulator - School for AI. This is both a training and testing environment that allows you to teach your brain architectures new skills and at the same time measure their performance.

School for AI (or School for short) is a world within Brain Simulator, coupled with an intuitive graphical interface. The training provided by School is separated into learning tasks. A single learning task teaches or tests preferably a single new skill or ability. A collection of learning tasks forms a curriculum. An agent is expected to gradually acquire knowledge as it is progressing through the curriculum. It should acquire knowledge from one learning task and then leverage that knowledge to acquire knowledge from the next learning task faster and with greater ease. We call this accumulation of skills “gradual learning” and we talk about it in more detail in our Framework document.

The training can occur in a range of environments. We've prepared a basic 2D environment (RoguelikeWorld), an advanced 2D environment (ToyWorld), and a basic 3D environment (3D version of ToyWorld). A single curriculum can train the same architecture on multiple different environments.

School for AI does not explicitly separate the training and testing phases. It prefers life-long learning - the agent is learning continuously, without any interruption. The testing is performed during training. This of course poses certain requirements on the datasets - the environments that are used in school are generated rather than fixed. This allows a fast learning agent to progress through the curriculum in shorter time than a slow learning agent. We’re planning to include existing datasets (both fixed and interactive) in School to provide a single interface and make using different datasets a breeze.

The learning tasks that appear in the current version of School are based on our Agent Development Roadmap. They teach the agent basic skills like object detection, color classification, or path finding. We took care to ensure that it is easy to add new learning tasks to the environment. We’ll be happy if you go on, get Brain Simulator code from GitHub and create a learning task of your own!

The school communicates with the agent through a fixed interface. The interface allows the agent to move around the environment, move its focus around the scene and interact with the objects found within the environment (when applicable).

For more information, have a look at the Brain Simulator documentation page for School for AI or just download Brain Simulator and try out School for yourself!

We look forward to receiving your input! please don't hesitate to get in touch at
Read all about GoodAI's latest achievements, technology, and initiatives on our About page.
Marek Rosa
CEO, CTO of GoodAI
CEO of Keen Software House