Learning = L7. The premise of L7 is simple: human learning and machine learning flourish in tandem. With improvements in human learning, we design machines that learn better and further inform human learning, enabling us to invent better machines, that enable better human learning, and so forth. L7 embodies this principle as a machine learning blog that emphasizes human learning. Topics may include anything from deep learning and machine learning to using machine learning to improve human learning to helping humans learn with machine learning.

Why the name L7?

For logos, I like simple, symmetric shapes: for L7, I started with a square. An “L” alone and a “7” alone are fragile shapes, but combined they form a square, with the former being sturdier. This is an allegory for human learning and machine learning. The name “L7” captures that symmetry about the diagonal. If you’re wondering what the “7” stands for in “L7”, count the letters after the “L” in “Learning.”

About the author

Curtis G. Northcutt is a grad student in Computer Science at MIT, supported by a NSF Fellowship and a MITx Digital Learning Research Fellowship working with Isaac Chuang. His work focuses on two goals: (1) characterizing and fixing (or learning in spite of) label errors in machine learning datasets, (2) using artificial intelligence to enable human intelligence. To this end, Curtis invented confident learning, a family of theory and algorithms for learning with label errors, and created cleanlab, a Python package using confident learning to find label errors in datasets, characterize label noise, and learn with noisy labels. Other fields related to Curtis’s work are weak supervision, semi-supervised learning, and online education.

His favorite rapper is PomDP the PhD rapper.

Curtis has been fortunate to receive the MIT Morris Joseph Levin Masters Thesis Award, an NSF Graduate Research Fellowship, the Barry M. Goldwater National Scholarship, and the Vanderbilt Founder’s Medal (Valedictorian). Curtis created and manages the cheating detection system used by MITx and HarvardX online course teams, particularly the MIT MicroMasters courses. While at MIT, he TA’d 6.867, a large graduate machine learning course.

Thank You for reading!