State of PPL 
How Probabilistic Programming is the next big thing in AI/ Data Science

A report on the State of Probabilistic Programming in Industry

Probabilistic Programming is a new paradigm that allows one to express Bayesian Statistics in computer code. The key strengths of these models is that it allows you to build scalable models that incorporate uncertainty leveraging modern research as Hamiltonian Monte Carlo.

Probabilistic Programming is widely used in academia, there are currently close to 200 papers using PyMC3 (a PPL) in various fields, including astronomy, chemistry, ecology, psychology, neuroscience, computer security, and many more.

It is also used to solve various business problems by large and small companies. One common use case we have heard is for A/B testing, which makes sense as uncertainty plays a big role. Another problem well solved by Probabilistic Programming is supply chain optimization and modelling loss ratios in Insurance.

We question over 100 professional data scientists including some from Space-X, Uber, Hotels.com and Quantopian on how they are using probabilistic programming and provide some insight into the exponential progress and opportunities for the next 12 months.