Better Data is All You Need — Ari Morcos, Datology

In this episode of the Latent Space Podcast, Alessio and Swyx interview Ari Morcos, CEO and co-founder of Datology, about data curation in machine learning. Ari discusses Datology's mission to improve data curation, enabling faster, better, and smaller model training. He shares his background in neuroscience and his transition to AI, driven by the realization of data's paramount importance over inductive biases. The conversation explores the undervaluing of data in research, the impact of self-supervised learning, and the challenges of open-source datasets. Ari also addresses the economics of data curation, the role of synthetic data, curriculum learning, and the potential of smaller, more efficient models. He emphasizes the need for diversity in data and Datology's focus on valuing data with respect to downstream use cases.