Anonymize Data Locally Before Using ChatGPT #shorts

Description

Chapters

Transcript

0:01 [laughter]
0:04 I ask my local LLM to take my
0:06 confidential data and anonymize any
0:08 names, date of birth, addresses, phone
0:10 numbers. I then take the output and feed
0:13 that into an online model to do the
0:14 work. So Todd, that's what we're doing.
0:16 So at StoryVine, we've built a
0:18 deidentification system. So, so what's
0:21 happening at StoryVine is
0:24 when a video comes into the system, we
0:26 take the transcript or we we we take the
0:29 video, we run it through a local
0:33 multimodal model that can watch the
0:36 video that can understand the words. It
0:38 makes a transcript. We then take that
0:41 transcript and we've got a
0:43 deidentification program that replaces
0:46 all of the personal with
0:49 variables with code.
0:51 We then take that transcript with all
0:54 the code in it and no personal
0:56 information and we send that out to chat
0:58 GPT which generates a blog post and just
1:02 a bunch of other And then we bring
1:04 it back into our system and we rehydrate
1:07 it. So, we take all those codes where it
1:09 was like Mary Smith and then we put Mary
1:12 Smith's name back in it once it's back
1:14 on our local machine. So, watch the full
1:16 replay at community.alon.ai.