Wednesday, January 8, 2020

Review of 2019 & Plans for 2020


SUMMARY:
  • Space Engineers released from Early Access 
  • GoodAI published new Badger Architecture 
  • Oranžérie (The Orangery) - our new office building

People are interested in knowing what our progress was in 2019 and what our plans are for 2020. In this post you can learn about the highlights of last year and our plans for the future.


Highlights for GoodAI Research


  • Badger architecture - we published a paper describing the framework and principles of our AI architecture. Badger is the culmination of our efforts over the last five years. It’s the main direction in which we will be focusing our research in the following years. Badger is an agent made from multiple agents (aka experts) that share the same policy (neural network) but each has unique internal state. They communicate and coordinate during the inner-loop (lifetime of the agent), and the result of this is adaptation and task solving. The search for a better expert-policy is performed in the outer loop, and only there the policy is modified. The advantage we see in Badger is the ability to automate the search for properties such as meta-learning, learned credit assignment, recursive self-improvement (via communication), gradual learning, transfer learning, and build up to AGI.

 

  • In December our team visited NeurIPS and used this opportunity to discuss Badger with experts and to recruit new colleagues. What we took from this experience is that Badger is perceived as an original and promising direction, and people are looking forward to seeing more demonstrations and comparisons on standardized benchmarks. We also received some new research ideas. Overall, the trip was very successful:
    • we organized a meetup on Multi-Agent Reinforcement Learning (MARL) which was attended by about 100 people,
    • our Senior Research Scientist Jan Feyereisl joined Matthew Crosby, from The Leverhulme Centre for the Future of Intelligence, and Hiroshi Yamakawa, from The Whole Brain Architecture Initiative (WBA), to show the results of the Animal-AI Olympics which tested AI-agents on common animal cognition tasks.
 

  • New researchers have already started to join our team. We aim to hire at least 10 new researchers in 2020 which would lead to scaling up our R&D efforts, allowing us to run more experiments on Badger architecture. If you are interested check out the open positions here.
  • New advisors joined our team (Thank you!) & More will join later:
    • Šimon Šicko and Lucia Šicková: Šimon and Lucia have been working with GoodAI since September. Simon is the CEO and Lucia is the CLO (Chief Learning Officer) of Pixel Federation, a successful and growing free-to-play gaming company from Bratislava, Slovakia. They are both experienced entrepreneurs, we have been friends for years and they have already given GoodAI valuable input. 


    • Petr Šrámek - Petr Šrámek has joined GoodAI as a Strategic Advisor. He is a serial entrepreneur with a track record in the AI world which includes founding the AI Startup Incubator.

  • GoodAI is proud to have become a supporter of Prg.AI. We are happy to be involved with this initiative which aims to develop Prague as an AI superhub and help the Czech Republic become a major AI center of excellence in Europe.

    GoodAI’s Director of AI Policy and Social Impacts, Marek Havrda, was invited to speak at the launch event in May. The initiative was founded by The Czech Technical University (ČVUT), Charles University, Czech Academy of Sciences, and the City of Prague, and is supported by GoodAI as well as a range of other local and international organizations.
  • Patent - we have filed a patent application for the core of our technology. We strongly believe that this technology is the most likely way AGI will be created in the future.


Highlights for GoodAI Applied


  • This year GoodAI Applied started two sister companies to specialize in specific industries.
  • GoodAI Solutions was launched in October in order to take over specific projects from GoodAI Applied with a focus on predictive maintenance, visual inspection, and asset management in the manufacturing and energy industries.
  • GoodAI Digital was also founded last year to focus on the automation of human conversations using the latest disruptive technologies such as Natural Language Processing.
GoodAI researcher demonstrates image recognition technology for pressure ulcer prevention.


Highlights for Keen Software House


  • Space Engineers was released from Early Access in February 2019
  • Reviews are above 9/10 across all board - Steam reviews and games press
  • The revenue doubled in a year-to-year comparison

  • “Concurrent players” has grown significantly over the years.


  • In November, we shared the details about where we are with the Space Engineers Xbox One port by running the live stream gameplay session.

  • Space Engineers was awarded as the best game in the Technology category at the 2019 Central and Eastern European Game Awards. 


New office: Oranžérie (The Orangery)


My long-term dream has always been to have our own office building which we can fit out the way we need it, and how we like it.

This year we were able to achieve this goal. We finally found a building in Prague that has all the attributes we needed. We chose to call it Oranžérie.


Oranžérie's office space is 1000 square meters and the garden is 2000 square meters (plus there's a large public park with a lake in front of it). The new office also includes an underground garage and a large conference room. We will also have our own gym, a fully fitted kitchen, a fireplace, and a garden hang-out spot complete with grills for summer barbeques!

Right now, a construction company is finishing the interiors, making Oranžérie cozy and comfortable, for our teams at Keen Software House and GoodAI. We believe that what surrounds you impacts what you create.

We are planning to move to Oranžérie in May 2020.

History: Oranžérie was built in 1760 and originally used as a winter greenhouse, it was later converted into a residential building, and then an office for the municipality.

This is the first time we are publicly sharing these photos!





Plans for GoodAI


These are our plans for 2020:
  • Badger architecture
    • Solve the training difficulties = make it easier to train, so that we can do more experiments
    • Benchmark = Use standard Machine Learning benchmarks to compare with other architectures. Demonstrate that Badger performs better than other models or that it can learn and solve tasks that are not possible by non-Badger architectures (we will be mostly measuring adaptability to novel tasks, not task performance)
    • Business cases = investigate where Badger can be implemented already, without waiting for better curricula
  • AI Game - if things go well, we will work on an experimental game that’s built around adaptable AI characters. This technology could be later imported to Space Engineers.
  • 3rd Round of General AI Challenge - launch a new round, focused on expanding Badger ideas and proofs of concept. The winners will be invited free of charge and all expenses paid to our Summer Workshop.
  • AI Governance Congress in Prague - we will be playing a crucial role in the 1st AI Governance Congress which will take place in Prague from 16-18 April. 
 
  • Summer Workshop - a fully expenses paid multi-disciplinary retreat, organized by us and hosted in Oranzerie. We will invite experts who can contribute to research on Badger-like architectures. From the multi-day retreat we expect: new insights, and new proofs of concept (demonstrations).
    • Disciplines: Multi-Agent Reinforcement Learning, Meta-Learning, Network Science, Complexity, Social Intelligence, Biologically inspired neural networks, etc.



Plans for Keen Software House


  • Keen Software House has continued a process of adaptation and development which we believe will ultimately bring great benefits for Keen Software House, as well as for our community. Rather than having a divided focus and spreading limited resources over multiple projects we are now devoting 100% of our development time to the much anticipated Xbox release of Space Engineers and to continuing the development of Space Engineers for PC.
  • Continue supporting and developing Space Engineers. One major update every few months
  • Release Space Engineers for Xbox One
  • Perform an internal study on what game engine is best for the future of Space Engineers?
  • Run an internal hackathon - new game mechanics, features, stories for Space Engineers
  • Start working on something new & big - more details later!


Conclusion


See here for a review from 2018.

2019 was great, and truly game-changing for Keen Software House and GoodAI. 2020 will bring even more changes and growth.

Thanks to everyone who is keeping their fingers crossed for us!

Thank you for reading this blog!

Best,
Marek Rosa
CEO, Creative Director, Founder at Keen Software House
CEO, CTO, Founder at GoodAI

For more news:
Space Engineers: www.SpaceEngineersGame.com
Medieval Engineers: www.MedievalEngineers.com
General AI Challenge: www.General-AI-Challenge.org
AI Roadmap Institute: www.RoadmapInstitute.org
GoodAI: www.GoodAI.com
Keen Software House: www.keenswh.com
GoodAI Applied: https://www.goodai.com/goodai-applied

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 its best-seller Space Engineers (4 million copies sold). Both companies are based in Prague, Czech Republic.

 Marek has been interested in artificial intelligence since childhood. He 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.

 GoodAI started in January 2014 and has over 30 research scientists, engineers, and consultants working across its Research and Applied teams.

At this time, Marek is developing both Space Engineers and Medieval Engineers, as well as leading daily research and development on recursive self-improvement based general AI architecture - Badger.
Posted by at 89 comments:

Thursday, December 19, 2019

Space Engineers: Holiday Update


SUMMARY:
  • The Holidays are here!
  • Free update for all!
  • Small window blocks
  • Large open cockpit
  • Additional catwalk variants into the Deco Pack 2 DLC
  • Space Engineers comic
  • Festive skybox moon
  • Bug fixes and optimizations

Hello, Engineers!

I’ll try to keep today’s blog short. My team and I would like to celebrate the Holidays with all of you, our players, whether you celebrate Christmas, Hanukkah, Kwanza or any other holiday during this festive season. Christmas is the biggest holiday for the Keen Software House Team and we usually give a lot of presents to our families and friends during this holiday. We also spend time with our loved ones and it’s a time when we slow down after another fantastic year of working on Space Engineers!

The reason why I’m writing this is very simple. Looking back at the Space Engineers in 2019, the year was very successful and very busy and there were a lot of changes and new things for the game.

We released Space Engineers out of Early Access at the end of February, but it was not the end. We did not stop moving forward.

In April, we came with the first DLC - Decorative Pack which was received very well by our community. I think I won’t exaggerate if I say that we were surprised by your reactions and the very positive feedback we received from you.

We decided to continue with releasing optional DLCs (designed to not split the Space Engineers playerbase), which always include a free update and paid cosmetics.
We went live with the second DLC - Style Pack - at the beginning of June and then with the third one - Economy - in August.

We wrapped up the 2019 DLCs with Decorative Pack II, which went live on the sixth anniversary of Space Engineers in October.

And again, we did not stop here. We decided to share the details about where we are with the Space Engineers Xbox One port by running the live stream gameplay session on November 28th. With our special guests, Cosmic Chimp, Captain Jack and Dr Craz, we streamed three hours of  Space Engineers gameplay on three Xbox Ones, having fun and answering questions from those members of our community who are excited to play Space Engineers on Microsoft’s console.

Xbox One Gameplay Reveal Stream (November 28, 2019)

What is special about today?


Today’s release is a special one, because we are bringing you a number of “presents”. The first one, the Small Window Blocks have been requested by many players, so we decided to add them to the game. There are 20 small window block variations in total, so I’m sure you’ll unleash your creativity even more now.
Small Window Blocks

Another one is the Large Open Cockpit, which has also been super popular with our community. This cockpit is an addition to already existing seats, but again, we listened to our community and we decided to give it to you as a Christmas present.

Large Open Cockpit

Finally, we are adding additional catwalk variants into the Deco Pack 2 DLC so they look good in even more situations. This upgrade is free for everyone who already owns the Deco Pack 2.

New Catwalk variants
We are also releasing several improvements with this release; all of the details can be found in the change log.

The Space Engineers Comic

With Christmas just around the corner, I am very happy to share great news about the Space Engineers comic, which was written by Jesse Baule. We have received a final version of the artwork from Mike Hamlett and I’m thrilled that we have the very first Space Engineers digital comic book coming out today!

Discover the story of Doctor Petra Halas and her intern Reginald Simmons, which brings you to a top-secret research facility where strange things are happening.
You can download the Space Engineers comic here.


If you liked some of the character skins (Miner suit), safe zone skins (Voronoi) and faction logos featured in the comic, you can purchase the Economy Deluxe and enjoy these cosmetic items while playing Space Engineers.

All players who own the Deluxe version of Space Engineers will get access to the comic book concept art as part of the free addition to the Deluxe package.

Please let us know if you liked the story, as well as the art featured in this comic and if you would like to see more Space Engineers comics in the future.

A festive skybox moon for the holidays


From December 19 until January 2nd, there will be the legendary red ship (Santa and his reindeer are too busy delivering presents) being pulled by a faithful team of drones, which will be visible in the Space Engineers skybox in-game and help us all to remember to celebrate the holidays with our loved ones.

Thank you and happy holidays!

Thank you for being such an awesome community! My team and I wanted to release this update for free, with no paid DLC attached to it, as a way to thank you for all your support.

I hope you’ll enjoy this update and that you’ll have a great time playing Space Engineers during the holiday season.

If you want to let me know your feedback, please get in touch via my personal email address marek.rosa@keenswh.com, or use our Keen Software House support site. I welcome all of the feedback we receive and we will use it to learn and provide better services to our players.


Thank you for reading this blog!

Best,
Marek Rosa
CEO, Creative Director, Founder at Keen Software House
CEO, CTO, Founder at GoodAI

For more news:
Space Engineers: www.SpaceEngineersGame.com
Medieval Engineers: www.MedievalEngineers.com
General AI Challenge: www.General-AI-Challenge.org
AI Roadmap Institute: www.RoadmapInstitute.org
GoodAI: www.GoodAI.com
Keen Software House: www.keenswh.com
GoodAI Applied: https://www.goodai.com/goodai-applied

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 (3 million copies sold). Both companies are based in Prague, Czech Republic.

 Marek has been interested in artificial intelligence since childhood. He 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.

 GoodAI started in January 2014 and has over 30 research scientists, engineers and consultants working across its Research and Applied teams.

 At this time, Marek is developing both Space Engineers and Medieval Engineers, as well as daily research and development on recursive self-improvement based general AI architecture.
Posted by at 5 comments:

Wednesday, December 4, 2019

GoodAI: Badger architecture and learning procedure

READ THE FULL PAPER HERE 



Summary:

  • Badger
    • An agent is made up of many homogeneous experts
    • All experts share the same expert policy, but have different internal memory states
    • Behavior, adaptation and learning to adapt emerges from the interactions of experts inside a single agent.
    • Functionally it captures the notion of ‘how can homogenous experts coordinate together to learn to solve new tasks as fast as possible’.
  • Source code of a prototype attached 
  • Current state & next steps
  • I am doubling the funding for GoodAI
  • We are looking for new colleagues. Join our team!


Intro

GoodAI started 5 years ago with a single mission: “develop general artificial intelligence - as fast as possible - to help humanity and understand the universe.”

Back then we had only a vague idea of what to do, what the requirements of general AI are, and what should be our milestones.

During the years we tried many different approaches and architectures - from biologically inspired to artificial neural networks and deep learning. Luckily for us, the field of AI has also dramatically accelerated.

We grew to an understanding that what we want is an agent that can quickly adapt to novel tasks. The follow up question was, what are the building blocks of an agent that can adapt its internal structure as it learns new tasks? Can we meta-learn these building block(s)?


Badger

All this effort led to the Badger architecture, which is a set of design principles and a learning protocol. 

Summary of badger

Badger = an architecture and a learning procedure where:
  • An agent is made up of many experts
  • All experts share the same communication policy (expert policy), but have different internal memory states
  • There are two levels of learning, an inner loop (with a communication stage) and an outer loop
  • Inner loop – Agent's behavior and adaptation emerges as a result of experts communicating between each other. Experts send messages (of any complexity) to each other and update their internal memories/states based on observations/messages and their internal state from the previous time-step. Expert policy is fixed and does not change during the inner loop. 
  • Inner loop loss need not even be a proper loss function. It can be any kind of structured feedback guiding the adaptation during the agent’s lifetime.
  • Outer loop – An expert policy is discovered over generations of agents, ensuring that strategies that find solutions to problems in diverse environments can quickly emerge in the inner loop.
  • Agent’s objective is to adapt fast to novel tasks
Exhibiting the following novel properties:
  • Roles of experts and connectivity among them assigned dynamically at inference time
  • Learned communication protocol with context-dependent messages of varied complexity
  • Generalizes to different numbers and types of inputs/outputs
  • Can be trained to handle variations in architecture during both training and testing
Initial empirical results show generalization and scalability along the spectrum of learning types.

A Badger agent is made up of many experts. All experts have the same expert policy, but have different internal memory states. Experts communicate and coordinate together in order to adapt the agent to novel tasks.

The crucial question is - how to get this expert policy? One approach is to handcraft it. The other approach is to meta-learn it, to automate the search for it. This approach can be framed as “multi-agent meta-learning” (in our case, multi-expert meta-learning). 

For the motivation behind Badger, more details, preliminary experiments, literature, please see the paper:




Why are we publishing this paper now?

Research on Badger architecture is in its infancy. 

The biggest bottleneck to our progress is the sheer amount of hypotheses that we need to develop and test. 

This is why need more people working on Badger. We need to grow our team!

We are looking for collaborators:
  • People who want to join our team, either in Prague, or remotely (so they can stay where they live). We are planning to hire 10 new researchers in 2020.
  • people who can't join our team, but still want to collaborate (university students, professors, entire research teams, R&D companies who are open to collaboration)
We will prefer researchers who have experience with Badger-like architectures, such as:
  • Multi-agent learning
  • Multi-agent reinforcement learning (MARL)
  • Learned communication 
  • Emergent language
  • Minimum viable environments for evaluating Badger-like agents
  • Meta-learning / meta-reinforcement learning
  • etc.
If you are interested, please get in touch: marek.rosa@goodai.com 

By publishing the Badger paper, we are not aiming for attention or publicity. We do it for a simple reason: we want to start speaking with people outside of our team, we need to grow our team.


Source code

Together with the paper, we are also releasing an accompanying demo source code that reproduces the ‘Attentional Architecture’ experiment, shown in Figure 7 in the paper. This should allow anyone interested in playing around with one particular instantiation of the Badger architecture, to get a sense of what some of the ideas in the paper are about.


Why are we opening up Badger research? Why not keep it in stealth mode?

General AI is years in the future, and the sooner we all get there, the better for all of us.


Why do we call it Badger?

First, we saw the video Honey Badgers Don’t Care which led us to see the documentary Honey Badgers: Masters of Mayhem.

Honey badgers are an inspiring example of scalable intelligence and life-long learning. Much like our desired general AI agent, they are able to come up with creative strategies and solve tasks way beyond their typical environment, thanks to general problem-solving, planning and fast adaptation abilities.

Honey badgers are also known for their grit, tenaciousness and “can-do” attitude, which we believe are indispensable qualities for any team set to achieve “impossible” goals.

For these reasons, we couldn't help but fall in love with Honey Badgers and admire their creative thinking, problem-solving and determination.


Current state

Our present prototypes are very basic proofs-of-concept, realized on toy tasks.

Badger principles show promise, but the real prototypes proving specific abilities such as lifelong learning, representation learning, transfer learning, more complex tasks, etc. are still ahead of us. 

We still have to analyze what are the business cases where a Badger-like architecture, in its current simple form, can bring benefits today. The same applies for benchmarking Badger on standard AI / ML datasets.



Next Steps

  • We are going to grow our GoodAI team. 
  • We are going to find new collaborators.
  • We are going to launch a third round of our General AI Challenge, focused on Badger.
  • See more ‘next steps’ in the Badger paper.

Funding

I believe that now is the right time to speed up and scale up our R&D, I am doubling the funding that I provide every year to GoodAI.

Conclusion


If you are interested in discussing Badger related architectures and frameworks (MARL, meta-learning, etc), MVE tasks (minimum viable environments), would like to discuss joining our team, starting any kind of cooperation, or have any kind of feedback, please get in touch: marek.rosa@goodai.com 

Thank you for reading this!

Best,
Marek Rosa
CEO, Creative Director, Founder at Keen Software House
CEO, CTO, Founder at GoodAI

For more news:
General AI Challenge: www.General-AI-Challenge.org 
AI Roadmap Institute: www.RoadmapInstitute.org  
Space Engineers: www.SpaceEngineersGame.com
Keen Software House: www.keenswh.com

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 its best-seller Space Engineers (4 million copies sold). Both companies are based in Prague, Czech Republic.

Marek has been interested in artificial intelligence since childhood. He 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.

GoodAI started in January 2014 and has over 30 research scientists, engineers, and consultants working across its Research and Applied teams.

At this time, Marek is developing both Space Engineers and Medieval Engineers, as well as leading daily research and development on recursive self-improvement based general AI architecture - Badger.
Posted by at 7 comments:
Subscribe to: Posts (Atom)