Uber is sharing the fruits of its research into self-driving cars and artificial intelligence in general, releasing a new programming language called Pyro that aims to help developers create probabilistic models for AI research.
Pyro is the first public project released by Uber AI Labs, according to a company representative, and it was designed to make it easier to train computers to infer outcomes. “Pyro is a tool for deep probabilistic modeling, unifying the best of modern deep learning and Bayesian modeling,” wrote Noah Goodman, Stanford researcher and member of Uber AI Labs, in a blog post Friday.
The new programming language comes out of work done at Uber AI Labs since its creation following the acquisition of a startup called Geometric Intelligence last year. AI research is getting so expensive that only a handful of companies can hope to operate at the bleeding edge of the technology, and Uber has raised billions of dollars premised around the notion that it will be one of the companies best positioned to benefit from the advent of self-driving vehicles.
Pyro is based on Python and the PyTorch library, and it allows developers to specify probabilistic models. When married to powerful computers, those models form the basis of the deep learning techniques at the heart of current AI research.
“In Pyro, both the generative models and the inference guides can include deep neural networks as components,” Goodman wrote. “The resulting deep probabilistic models have shown great promise in recent work, especially for unsupervised and semi-supervised machine learning problems.”
Now we’ll see what happens when deep learning researchers get a chance to kick the tires on Pyro; Uber described it as an “alpha release,” which implies there’s a fair amount of work still to be done. Given the huge expense required to advance AI research, open-source projects might be a crucial way for companies and people that lack Uber’s bank account to make breakthroughs of their own.