Alphabet subsidiary DeepMind makes even more AI projects open-source

Alphabet’s research group on the development of artificial intelligence, DeepMind Technologies, is going to share more of its work with the world. Some of the important algorithmic components is made open-source. These are algorithms associated with what the company describes as some of its most successful AI initiatives.

The algorithms are housed in the TRFL library and are aimed at helping researchers working on reinforcement learning projects to train neural networks. In essence, reinforcement learning is a method of trial and error, which improves the accuracy of an algorithm. If an AI makes a correct decision in a training environment, it is virtually rewarded, so that it makes good choices more often.

Reinforcement learning

TRFL, which is pronounced as truffle, includes the implementations of mathematical operations used by algorithms for reinforcement learning. They are linked to components that DeepMind says can perform more advanced calculations. In addition, they contain other building blocks that are needed to ensure that AI training runs successfully and smoothly.

The entire collection is built to run on the popular TensorFlow deep learning engine developed by Alphabet subsidiary Google. Researchers can interact with TRFL via a programming interface. This makes it relatively easy for DeepMind to link components in techniques and concepts.

More construction bucks

The goal behind TRFL goes beyond simply simplifying individual reinforcement learning projects. The hope is that more building blocks will be laid out for developers and researchers to work with and strengthen their AI. The basic idea is that the same components are used more often in certain projects. This will enable researchers to work more quickly, which should accelerate the development of artificial intelligence.

TRFL is available on GitHub and is part of a longer series of open-source projects that DeepMind and Alphabet have released in recent years.