Universal Data Income: Paying Contributors on-Chain | David Dao

AI was trained on us. How do we value data and pay the people who create it? David Dao, co-founder and Chief Scientist of Gainforest, joins host Jonathan Schwartz to unpack data valuation, Shapley values, which can be applied to address this issue, and how a green data marketplace could fund conservation while improving AI.

We break down:

What “valuable data” means in practice (task-specific, fair, budget-aware)
Shapley values, KN-N shortcuts, and “one-run” data valuation at scale
Data provenance, wallets, and paying contributors on-chain (universal data income)
Why biodiversity data is missing where it matters—and how to fix it
Privacy, leakage, and differential privacy tradeoffs in real systems
How decentralized science (DeSci) can finance nature with verifiable impact

Chapters:
00:00 – Intro: Velocity of Change & today’s topic
00:20 – Guest intro: David Dao & Gainforest
05:52 – What is data valuation? (task-specific, fair, budget-aware)
07:57 – Shapley values explained (from game theory to ML)
09:54 – Scaling challenges: retraining, sampling, influence functions
16:42 – KNN-Shapley and closed-form speedups
23:51 – “One training run” Data Shapley (batching; LLM relevance)
31:21 – Time-based Shapley & ongoing on-chain payouts
50:59 – Privacy, provenance, and royalties for data
01:23:07 – Universal Data Income & where this goes next

Links
Follow GainForest on X
https://x.com/GainForestNow

Follow GainForest on Bluesky
https://bsky.app/profile/gainforest.e...

Follow David on X
https://x.com/dwdow

Follow Jonathan on X
https://x.com/j_schwartzz

About Protocol Labs
Protocol Labs is an innovation network of 600+ teams, projects, and movements advancing frontier computing.

Follow Protocol Labs
Website: https://www.protocol.ai

Bluesky: https://protocollabs.bsky.app

X: https://x.com/protocollabs

LinkedIn:   / protocollabs