Page 5 - summer of ai

SUMMER OF AI JULY 2019
But what does it mean for machines to learn from experts,
data, and simulations?
In the first article, Dr. Rajkumar Bondugula discusses “Learn from Experts,” artificial intelligence that creates knowledge bases as
the foundation for expert systems. At Equifax, this type of artificial intelligence is ideal for data ingestion and improving data quality. You take raw, unlabeled, and unstructured data, find hidden patterns and build knowledge bases for machines to create rules.
Dr. Matthew Turner talks about “Learn from Data” – commonly known as machine learning. This type of artificial intelligence makes up a fair share of the AI portfolio at Equifax — nearly 90% of the Data Science Lab’s patent applications, according to Chris. The intense focus on “Learn from Data” represents the high importance customers place in the next stage of the insights supply chain — model development and insights generation. While many companies in banking and lending rely on logistic regression to predict future behavior, the patented NeuroDecision® Technology uses neural networks increase predictive performance and generate explanations at the consumer level for a credit decision.
John Fenstermaker describes “Learn from Simulations” – best exemplified by leveraging the enormous computing power available to the lab, and by far the newest AI technology at Equifax. Recognizing the need for a dynamic AI solution that automates the modeling
DATA & ANALYTICS 5
process from model governance to model updates, Equifax invented
an adaptive AI technology for the final stage of the insights supply chain – model deployment. Chris often compares adaptive AI to how some
of the popular navigation apps identify the fastest route through traffic. Adaptive AI simulates thousands of combinations to build the fastest and most valuable model constrained by a customer’s governance and business demands.
These AI technologies help customers look for people who have
never been scored before. As well as people with thin files that appear “invisible” in the traditional financial models built with logistic regression. If we can help customers identify new consumers to score utilizing better data coupled with sound predictions, we can help more people have access to credit they otherwise would not have. In doing so, businesses can reach more people without adding risk or at times even reducing risk in their portfolio.
Now, it is your turn to ask: where can AI be used to solve my challenge?
CHRISTOPHER YASKO VICE PRESIDENT DATA SCIENCE LAB
If we can help customers identify new consumers to score utilizing better data— we can help more people have access to credit they otherwise would not have.