S3 E5 Berkeley Prof Raluca Ada Popa: How to keep your data confidential when training AI
The Robot Brains Podcast The Robot Brains Podcast
13.6K subscribers
2,563 views
0

 Published On Apr 12, 2023

S3 E5 Berkeley Prof Raluca Ada Popa: How to keep your data confidential when training AI

What's in this episode:

00:00:00 - Raluca Ada Popa
00:01:27 - sponsors: Index Ventures, Weights and Biases
00:02:34 - tension between privacy and data needs for AI model training
00:05:57 - confidential computing: encrypted computing, cloud enclaves
00:09:25 - neural network training with cloud enclaves
00:13:15 - differential privacy
00:17:54 - advertising tech in absence of cookies
00:19:05 - Opaque founding story to today
00:25:48 - PreVeil, sharded keys, red button
00:34:40 - junction of AI and Security: training on secure data, limiting what models can reveal about training data
00:42:33 - Raluca’s growing up to today
00:47:50 - motivation
00:50:10 - what looking for when recruiting students
00:57:14 - what does Raluca do to unwind and relax

Guest links:
https://people.eecs.berkeley.edu/~ral...
https://confidentialcomputingsummit.com/
https://opaque.co/
https://www.preveil.com/

Sponsor Links:
https://www.indexventures.com/
https://wandb.com

Subscribe and listen to our podcast today:
Apple: http://bit.ly/3zPWbVs
Spotify: https://spoti.fi/3zTN3iG
Amazon: http://bit.ly/3KVrowE
Google: http://bit.ly/3KVBTAq
Acast: http://bit.ly/3moJqhI

Host: Pieter Abbeel -- https://people.eecs.berkeley.edu/~pab...
Production: Bo Obradovic

show more

Share/Embed