Fine-tuning language models using crowdsourcing similar to Wikipedia

4min Snip

00:00

Play full episode

Summary

Comments

Transcript

Episode notes

The idea of fine-tuning language models through a collaborative, crowd-sourced approach similar to Wikipedia is a compelling one. It aims to train models to be safe, aligned, and responsible. By allowing a broader community to participate in the fine-tuning process, it aligns with the ethos of open sourcing technology and acknowledges the challenges of infrastructure, community management, and product perspectives. However, this approach raises many questions and requires further research to determine the best way to produce aligned AI models. The success of Wikipedia in handling controversial topics with balance and grace serves as inspiration, although it is not a perfect system. In the pursuit of truth, knowledge, and wisdom, AI models must strive to achieve similar outcomes.

Mark Zuckerberg is CEO of Meta. Please support this podcast by checking out our sponsors:
– Numerai: https://numer.ai/lex
– Shopify: https://shopify.com/lex to get $1 per month trial
– BetterHelp: https://betterhelp.com/lex to get 10% off

EPISODE LINKS:
Mark’s Facebook: https://facebook.com/zuck
Mark’s Instagram: https://instagram.com/zuck
Meta AI: https://ai.facebook.com/
Meta Quest: https://www.meta.com/quest/

PODCAST INFO:
Podcast website: https://lexfridman.com/podcast
Apple Podcasts: https://apple.co/2lwqZIr
Spotify: https://spoti.fi/2nEwCF8
RSS: https://lexfridman.com/feed/podcast/
YouTube Full Episodes: https://youtube.com/lexfridman
YouTube Clips: https://youtube.com/lexclips

SUPPORT & CONNECT:
– Check out the sponsors above, it’s the best way to support this podcast
– Support on Patreon: https://www.patreon.com/lexfridman
– Twitter: https://twitter.com/lexfridman
– Instagram: https://www.instagram.com/lexfridman
– LinkedIn: https://www.linkedin.com/in/lexfridman
– Facebook: https://www.facebook.com/lexfridman
– Medium: https://medium.com/@lexfridman

OUTLINE:
Here’s the timestamps for the episode. On some podcast players you should be able to click the timestamp to jump to that time.
(00:00) – Introduction
(05:38) – Jiu-jitsu competition
(23:01) – AI and open source movement
(35:32) – Next AI model release
(47:48) – Future of AI at Meta
(1:08:25) – Bots
(1:23:53) – Censorship
(1:38:34) – Meta’s new social network
(1:45:20) – Elon Musk
(1:49:25) – Layoffs and firing
(1:56:55) – Hiring
(2:02:48) – Meta Quest 3
(2:09:45) – Apple Vision Pro
(2:16:00) – AI existential risk
(2:22:23) – Power
(2:25:55) – AGI timeline
(2:33:17) – Murph challenge
(2:38:33) – Embodied AGI
(2:41:39) – Faith