Uploaded July 2026 | Updated November 2026, 3 hours ago
In this episode of the ODSC Ai X Podcast, Ari Morcos, CEO and Co-Founder of DatologyAI, explains why data quality may now deliver greater returns than simply building larger models or adding more compute.
Drawing on his research at DeepMind and Meta AI Research, Ari discusses how high-quality data acts as a compute multiplier, and why enterprise AI strategy may shift from renting general-purpose intelligence to owning specialized models trained on carefully curated data.
The conversation explores how companies can move from model APIs and fine-tuning toward mid-training and eventually training models from scratch, along with the tradeoffs around synthetic data, catastrophic forgetting, and enterprise data curation.
00:00 Introduction: Own Your Intelligence, Not Rent It
00:00:30 Welcome & Guest Introduction
00:01:22 Why Better Data Matters More Than Bigger Models
00:02:28 Ari Morcos' Journey: Neuroscience to DeepMind & Meta
00:04:48 Why Data Became the Biggest AI Breakthrough
00:07:00 Self-Supervised Learning Changed Everything
00:10:53 Why Ari Founded Datology AI
00:13:55 Data Quality Is the Ultimate Compute Multiplier
00:19:18 Does More Data Always Mean Better Models?
00:21:42 When Should Companies Build Their Own AI Models?
00:24:40 Fine-Tuning vs Mid-Training vs Training from Scratch
00:29:48 Is Training a Model from Scratch Worth It?
00:34:12 Build a Model Factory, Not Just a Model
00:36:33 What Is Mid-Training?
00:38:58 How to Avoid Catastrophic Forgetting
00:42:24 How Datology Curates Training Data
00:48:02 Synthetic Data: When It Helps and When It Hurts
00:52:49 The Biggest Mistakes Companies Make with AI Training
00:54:15 How Datology Works with Enterprise Data
00:55:41 Lessons from Building an AI Startup
00:57:40 Final Thoughts & Outro
Visit our website and choose the nearest ODSC AI event to attend and experience all our training and workshops: odsc.ai
To watch more videos like this, visit https://aiplus.training
Sign up for the newsletter to stay up to date with the latest trends in data science: opendatascience.com/newsletter
Follow us online!
• Facebook: facebook.com/OPENDATASCI
• Instagram: instagram.com/odsc
• Blog: opendatascience.com
• LinkedIn: linkedin.com/company/open-data-science
• X (twitter): https://x.com/_odsc
In this episode of the ODSC Ai X Podcast, Ari Morcos, CEO and Co-Founder of DatologyAI, explains why data quality may now deliver greater returns than simply building larger models or adding more compute.
Drawing on his research at DeepMind and Meta AI Research, Ari discusses how high-quality data acts as a compute multiplier, and why enterprise AI strategy may shift from renting general-purpose intelligence to owning specialized models trained on carefully curated data.
The conversation explores how companies can move from model APIs and fine-tuning toward mid-training and eventually training models from scratch, along with the tradeoffs around synthetic data, catastrophic forgetting, and enterprise data curation.
00:00 Introduction: Own Your Intelligence, Not Rent It
00:00:30 Welcome & Guest Introduction
00:01:22 Why Better Data Matters More Than Bigger Models
00:02:28 Ari Morcos' Journey: Neuroscience to DeepMind & Meta
00:04:48 Why Data Became the Biggest AI Breakthrough
00:07:00 Self-Supervised Learning Changed Everything
00:10:53 Why Ari Founded Datology AI
00:13:55 Data Quality Is the Ultimate Compute Multiplier
00:19:18 Does More Data Always Mean Better Models?
00:21:42 When Should Companies Build Their Own AI Models?
00:24:40 Fine-Tuning vs Mid-Training vs Training from Scratch
00:29:48 Is Training a Model from Scratch Worth It?
00:34:12 Build a Model Factory, Not Just a Model
00:36:33 What Is Mid-Training?
00:38:58 How to Avoid Catastrophic Forgetting
00:42:24 How Datology Curates Training Data
00:48:02 Synthetic Data: When It Helps and When It Hurts
00:52:49 The Biggest Mistakes Companies Make with AI Training
00:54:15 How Datology Works with Enterprise Data
00:55:41 Lessons from Building an AI Startup
00:57:40 Final Thoughts & Outro
Visit our website and choose the nearest ODSC AI event to attend and experience all our training and workshops: odsc.ai
To watch more videos like this, visit https://aiplus.training
Sign up for the newsletter to stay up to date with the latest trends in data science: opendatascience.com/newsletter
Follow us online!
• Facebook: facebook.com/OPENDATASCI
• Instagram: instagram.com/odsc
• Blog: opendatascience.com
• LinkedIn: linkedin.com/company/open-data-science
• X (twitter): https://x.com/_odsc










