The Complete Magazine on Open Source

Google DeepMind rivals OpenAI with open source availability

/ 4743 0

Google DeepMind

In the fast-growing world of artificial intelligence (AI), Google’s parent company Alphabet is set to take on Elon Musk-backed OpenAI by announcing the open source availability of its DeepMind. The new development is aimed to enhance AI research.

DeepMind gained huge popularity with the debut of AlphaGo, an AI agent that recently mastered abstract strategy game Go and defeat an human expert. But that was just an instance. The Alphabet team is planning to even “push the boundaries” of the present AI movement with its DeepMind.

“DeepMind’s scientific mission is to push the boundaries of AI, developing systems that can learn to solve any complex problem without needing to be taught how. To achieve this, we work from the premise that AI needs to be general,” the DeepMind team, comprises of Software Engineer Charlie Beattie and Stig Petersen, Research Scientist Joel Leido and Chief Scientist Shane Legg, writes in a blog post.

AI research on top

Developers can leverage DeepMind as a 3D game-like platform for their agent-based AI research. It helps in building virtual games and uses advanced machine learning technology called deep reinforcement learning (DeepRL). In terms of customisation, the technology comes with editor tools and enables programmatic level-creation.

Just like DeepMind, Musk’s OpenAI also comes as an AI research platform. Both are aiming to uplift the field of AI albeit using different avenues.

Last month, OpenAI received Microsoft Azure support to enable cloud computing for AI research. Microsoft offerings like Azure Batch, Azure Machine Learning and the Microsoft Cognitive Toolkit all can work with OpenAI. However, contrary to OpenAI, DeepMind uses Google Cloud resources.

The DeepMind Lab is available as a GitHub repository under an open source license. It comes with assets as well as the ioq3 engine and documentation to ease AI developments.