all AI news
Text Embeddings Reveal (Almost) As Much As Text
Jan. 8, 2024, 5:22 a.m. |
Simon Willison's Weblog simonwillison.net
Text Embeddings Reveal (Almost) As Much As Text
Embeddings of text - where a text string is converted into a fixed-number length array of floating point numbers - are demonstrably reversible: "a multi-step method that iteratively corrects and re-embeds text is able to recover 92% of 32-token text inputs exactly".
This means that if you're using a vector database for embeddings of private data you need to treat those embedding vectors with the same level of protection as the original …
ai array embeddings floating point inputs numbers privacy security string text token
More from simonwillison.net / Simon Willison's Weblog
Printing music with CSS Grid
1 day, 12 hours ago |
simonwillison.net
Introducing the Claude Team plan and iOS app
2 days, 11 hours ago |
simonwillison.net
Jobs in AI, ML, Big Data
AI Engineer Intern, Agents
@ Occam AI | US
AI Research Scientist
@ Vara | Berlin, Germany and Remote
Data Architect
@ University of Texas at Austin | Austin, TX
Data ETL Engineer
@ University of Texas at Austin | Austin, TX
Lead GNSS Data Scientist
@ Lurra Systems | Melbourne
Lead Data Modeler
@ Sherwin-Williams | Cleveland, OH, United States