This paper:

• Explores work done by researchers since the original scaling laws paper and will revisit evidence of scaling laws’ veracity

• Suggests that in cases of significantly large data sets that come from low quality data sources, scaling laws no longer hold true and instead the law of diminishing returns takes its place

• Explores empirical investigations that show where scaling laws work and where they do not

• Suggests ways we can consider scaling laws as a component of many predictors that can contribute to model performance while ensuring data quality is at the center of the power law.

2021 American Statistical Association JSM Panel with Sarah Cox, Coinbase; Samsara Counts, Amazon; Dr. Kalinda Ukanwa, University of Southern California

My presentation on the policy regulation limitations for Generative AI.

The power of Machine Learning to increase Product Marketing

2022 NeurIPs Global South in AI Keynote Speaker presentation

Deep Learning Presentation for Deci AI - 2022

Presentation of the findings from “Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism” paper