Google lost the battle for machine learning to Meta, insiders say. Now it’s betting the future of its own products on a new internal AI project

• Google was a trailblazer in machine learning, releasing one of the first general-use frameworks.
• TensorFlow has since lost the hearts and minds of developers to Meta’s AI framework, PyTorch.
• Google is now betting on a new AI project internally to replace TensorFlow called JAX.

Google, in 2015, essentially created the modern-day machine learning ecosystem when it open sourced a small research project from Google Brain in 2015 called TensorFlow. It quickly exploded in popularity and made the company the steward of mainstream AI products. 

But the story is very different today, where Google has lost the hearts and minds of developers — to Meta. 

Once an omnipresent machine learning tool, Google’s TensorFlow has since fallen behind Meta’s machine-learning tool PyTorch. First developed at Facebook and open sourced in beta form in 2017, PyTorch is increasingly coming to be seen as the leader.