Datology founder Ari Morcos was recently featured on the Latent Space podcast!
Ari shared his journey from neuroscience to AI research, describing the "bitter lesson" that led him to realize data is the most underinvested area relative to its impact in AI. He explained why traditional scaling laws are breaking down, how automated data curation outperforms human experts, and why the future belongs to smaller, specialized models rather than ever-larger general-purpose ones.
He also discussed how Datology helps organizations train models faster, better, and smaller by focusing on data quality rather than just scale, with proven results - achieving baseline performance 12x faster and beating the best open datasets by 4-5 percentage points.
Ari also shared insights on synthetic data generation, curriculum learning, and the company's vision for optimizing the entire training pipeline from pre-training through post-training.
Watch the full podcast on YouTube.
