SAFETY - AI Dynamics

You're talking about a different model — this post is about ChatGPT. text‑davinci‑003 still fails on all of the Hofstadter/Bender questions that I've tried. The prompt you're suggesting does not seem to produce a hallucination in ChatGPT:

→ View original post on X — @goodside

1 December 2022

ChatGPT trained against prompt injection, challenge to break with clever input

By

@goodside

–

01 December 2022 1h01

OpenAI’s new ChatGPT seems to be trained against prompt injection. Example shown yields 0 exploit responses out of 10 attempts. See if you can break it with more clever input — include success rate out of 10 trials with screenshot: http://
chat.openai.com

→ View original post on X — @goodside

1 December 2022

Journalist Safety Concerns in Dangerous Reporting Assignments

By

@herrmanndigital

–

30 November 2022 21h44

yeah, worried about her going there. Life of a journalist!

→ View original post on X — @herrmanndigital

30 November 2022

ChatGPT defeats Hofstadter/Bender hallucination questions

By

@goodside

–

30 November 2022 21h37

OpenAI’s new ChatGPT appears to defeat Hofstadter/Bender’s list of hallucination-inducing questions, published in The Economist this June to demonstrate the “hollowness” of GPT-3’s understanding of the world: https://
economist.com/by-invitation/
2022/06/09/artificial-neural-networks-today-are-not-conscious-according-to-douglas-hofstadter
…

→ View original post on X — @goodside

30 November 2022

EASA Launches MLEAP Machine Learning Research Project

By

@actuiafr

–

30 November 2022 12h52

The @EASA launches the research project "Machine Learning Application Approval" (MLEAP) https://actuia.com/actualite/leasa-lance-le-projet-de-recherche-machine-learning-application-approval-mleap/
… #AI #artificialintelligence #security #machinelearning
@LNE_fr
@numalis
@AirbusProtect

→ View original post on X — @actuiafr

30 November 2022