One of the most pressing problems in machine learning research these days is reproducibility of results. A 2017 study of 1500 researchers showed that over 70% of scientists have failed to reproduce the results from a published study, and 90% of scientists in that same survey stated that they believed reproducibility was in crisis. Considering how integral AI has become in academic research, it is critical that researchers be able to efficiently and effectively reproduce results not just for peer review, but also for furthering research in an open-source manner such that AI becomes more “understandable” and accessible to everybody.
plAIgrounds is a decentralized, open-source data science platform that addresses the reproducibility crisis by adapting the immutable computational architecture of Resilient Distributed Datasets into a distributed software platform, allowing researchers to publish their code on a decentralized network that gives others the ability to verify models and methods step-by-step on any machine. plAIgrounds allows anybody to utilize published models under open-source licenses in future research or commerce by branching existing projects, preventing the need for time-consuming data engineering and model reproduction, and providing smaller institutions and organizations access to cutting edge research that they otherwise would not have access to.