Skip to content
Sonic
AI
Sonic
AI
Home
Discover
Ask Sonic
Projects
Use with Claude or ChatGPT
Show me around
Request source or feature
The data black hole at the center of AI, Sonic AI
Home
/
The Dwarkesh Podcast
/
The data black hole at the center of AI
The Dwarkesh Podcast
Notify me
•
Jun 19, 2026
•
11:56
Interview
The data black hole at the center of AI
From
The Dwarkesh Podcast
Dwarkesh Patel
(host)
Get the full transcript next time The Dwarkesh Podcast releases an episode
Summary, key quotes, top claims, and the searchable transcript — emailed automatically. No card needed.
Sign up
Executive Summary
Current AI progress is primarily driven by scaling data and compute, not by fundamental improvements in sample efficiency, with models requiring millions of times more data than humans to learn comparable skills.
The data-hungry nature of AI has created a multi-billion dollar industry for expert data labeling and reinforcement learning environments, which is projected to grow to tens of billions.
Despite their inefficiency, AIs are economically viable for automating common white-collar tasks because the high, one-time training cost can be amortized across billions of sessions.
The long-term strategy of major AI labs is to first automate AI research, which they hope will solve the core sample efficiency problem and unlock the automation of more complex, creative jobs.
Continue your research
Keep pulling the thread on Epoch.
The Sample Inefficiency of AI
Data as the Engine of AI Progress
The Economics of Inefficient Automation
Or ask anything across 400+ expert conversations
12
quotes
Transcript
Key Arguments
Analysis
Quotes & Entities
12
Related
Loading transcript...
Processed Jun 19, 2026
Daily intelligence brief →
yt-dlp + mlx-whisper + Gemini