Data31 Aug 2026
The Biggest Problem With AI Nobody Is Talking About
Sunil Nair
Co-Founder, Investments and Business Development, Clairva
Singapore
The video plays from YouTube when you press play.
What we talked about
Sunil explains why better models still need representative real-world data. He describes turning consented video into structured contextual data, Western-heavy training sets, contributor networks, copyright and provenance, contributor payment, synthetic-data risk and AI colonisation.
Key takeaways
- 01Western-heavy data can make AI misunderstand the Global South.
- 02Video teaches physical context and behaviour that text cannot.
- 03Only useful segments of long footage may be needed once labelled.
- 04Consent, copyright and provenance must travel with the data.
- 05Synthetic data can repeat existing blind spots.
“We are in the pix and shovels business of AI.”
About Sunil
Sunil Nair is a co-founder of Singapore-based Clairva, working across investments and business development. Clairva supplies licensed real-world video and audio data for multimodal AI, world models and embodied systems.
If you liked this, watch next.
5 Oct 2026 · Technology
The Real-World Test 99% of Voice AI Startups Skip
Raj Jethwa, Voice AI Strategist, Speechmatics
6 Jul 2026 · Legal
How AI Engineers Work Beyond ChatGPT
Ridhima Garg, Senior AI Engineer
28 Sept 2026 · Software
I Built 15 Apps With Vibe Coding
Henry Greer, Product Owner and former full-stack developer, 3Cloud
21 Sept 2026 · Gaming
How This Studio Builds 1,000 Games a Year With AI Agents
Tatu Petersen-Jessen, Co-Founder, Odd Latent
Free AI Masterclass · Mon 16 Nov
Turn ideas like these into one working AI workflow.
A free 90-minute live session for business leaders.
Free · Live on Zoom · Replay included