Why Humans Are Still Powering AI [Sponsored]

Episode Link

https://share.snipd.com/episode/b26a7d70-152b-47d3-9594-fa438ac8ff19

Episode publish date

November 3, 2025 12:42 AM (UTC)

Last edit date

Nov 4, 2025 3:21 PM

Last snip date

November 4, 2025 3:20 PM (UTC)

Last sync date

November 4, 2025 3:21 PM (UTC)

Show

Machine Learning Street Talk (MLST)

Show notes link

https://podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Why-Humans-Are-Still-Powering-AI-Sponsored-e3adil7

Snips

Warning

⚠️ Any content within the episode information, snip blocks might be updated or overwritten by Snipd in a future sync. Add your edits or additional notes outside these blocks to keep them safe.

‣

Episode show notes

Your snips

‣

[00:15] Human Data Is The Hidden Core

‣

[02:17] Code-Review Experience Informed Matching

‣

[03:46] Prioritize Vetting And Long-Term Incentives

‣

[06:30] Design For Relationships, Not One-Off Tasks

‣

[08:39] Direct Feedback Improves Trust

‣

[10:12] Force Clear Specs And Train Participants

‣

[11:53] Prefer Real-World Experts Over Pro Raters

‣

[16:01] Two-Tower Matching For People And Tasks

‣

[16:46] Centralization Creates Geopolitical Risk

‣

[18:10] AI as a Teacher

‣

[18:37] Use AI To Amplify Junior Engineers

‣

[20:07] Agents And Humans Will Collaborate

‣

[22:20] Synthetic And Human Data Are Complementary

‣