DeepMind, Google’s artificial intelligence subsidiary, has revealed a groundbreaking innovation called PathNet. This neural network model, inspired by the human brain, enables artificial intelligence to learn and adapt in a way that mimics human learning. Essentially, PathNet allows AI to transfer knowledge from one task to another, making it a significant leap in machine learning.
PathNet is a modular system comprising several neural networks or ‘agents’. Each of these agents is capable of learning a different task, and their combined knowledge can be applied to new tasks. This process, known as ‘transfer learning’, is similar to how humans apply knowledge from past experiences to new situations.
This revolutionary technology could have far-reaching implications. For one, it could make AI more efficient and flexible, allowing it to adapt to a wide range of tasks. Moreover, PathNet could potentially enable AI to learn continually without the need for constant retraining, thereby saving time and resources.
Despite these exciting advancements, DeepMind acknowledges that PathNet is still a work in progress. The team is currently testing the system on video games to evaluate its learning capabilities. However, the company is optimistic about the future, believing that PathNet could pave the way for more sophisticated AI systems.
Go to source article: https://medium.com/@thoszymkowiak/deepmind-just-published-a-mind-blowing-paper-pathnet-f72b1ed38d46#.8nr5b935q