Machine Learning and the Importance of Rapid Experimentation: A Conversation with Moody Analytics’ Keith Berry

Keith Berry has been through it all in the tech world: from a programmer, to an electrical engineer, to a data systems head, to a program director for one of the top analytics companies. Through his career, Berry has accumulated over twenty-five years of experience and is a source ripe with knowledge when it comes to the evolution of technology, especially during an innovative time in the industry. According to Berry, “the cloud has been transformational” in revolutionizing the way we use modern technology, and thanks to Microsoft, Google, and other tech companies’ abilities to host millions upon millions of servers, companies can store incomprehensible amounts of data at a low cost. Machine learning and artificial intelligence have arisen as companies attempt to use this mass amount of data. Further, programming languages like Python and TensorFlow, commonly used for machine learning, have become increasingly popular of late, as has the open-source website Github which allows programmers to share code and pool ideas with others across the world. The combination of machine learning along with artificial intelligence has, in Berry’s words, “untapped potential,” as these AI-based machines can now act based on real-time company info and datasets. Berry discussed how AI is being used today, and the examples are mind-blowing – Microsoft uses AI to map all physical structures across the Earth using satellite imaging alone, tech companies are pooling their APIs to create more advanced solutions, and the rise of Blockchain has had a tremendous effect on the financial markets.

Berry’s current position as director of Moody’s Analytics Accelerator gives him the unique opportunity to use machine learning and AI to assist many different startups in their attempts to get off the ground. He uses “The Lean Startup” technique popularized in Eric Ries’ book of the same name. This methodology for building companies from the ground up is based on the idea of rapid experimentation; companies must build hypotheses, test assumptions (smartly and efficiently), and limit risk by making decisions quickly and at a lower cost. Berry stressed that it is important to attempt many small actions a few times in order to limit risk and allow yourself to learn from mistakes, rather than try and take down an enormous problem in one fell swoop.

In today’s world, progress is based around data collection and analysis, and it must be done quickly and without error. That need for instant knowledge and decision-making means that machine learning and AI will ultimately become more and more necessary as our society progresses.