Shreya Shankar: Machine Learning in the Real World

ML Evolution

The evolution of machine learning has been marked by significant shifts, particularly from traditional methods to deep learning and generative AI. Shreya Shankar highlights the dual histories of AI/ML modeling and the tools and data management systems that support them. She notes the impact of the 2014 technical debt paper, which emphasized the importance of data management in ML systems 1. This shift in focus from solely training models to considering the entire system has been pivotal.

There's this great technical debt in machine learning paper from 2014 that really sparked the intersection kind of of the two fields converging.

--- Shreya Shankar

Daniel Bashir and Shreya also discuss the changes in educational ecosystems, such as the shift at Stanford from classical machine learning to deep learning, reflecting broader trends in AI research 2.

   

Educational Shifts

Educational trends in machine learning have evolved, making research more accessible and engaging for students. Shreya recalls the excitement around constructing datasets for pretraining models, a challenging yet rewarding task that has become easier with advancements in fusion learning and pretrained models 3. This accessibility has allowed even high school students to engage with ML concepts, contributing to a more competitive and collaborative environment.

It's more accessible and there are more ways people can contribute.

--- Shreya Shankar

Shreya's own journey into ML was sparked by the emergence of generative AI and deep learning classes at Stanford, which offered practical applications and interdisciplinary opportunities, such as using RNNs to generate music 4.