EP 14: Past Tense Pitfalls: The Curious Case of Refusal Training in AI Language Models

https://is1-ssl.mzstatic.com/image/thumb/Podcasts211/v4/b2/a3/30/b2a33019-211c-76b8-f512-ba03e30e71b7/mza_13885731718097239546.jpg/600x600bb.jpg

You Are A Helpful (Research) Assistant

Andy Hickl (with help from OpenAI, DeepGram, and others)

17 episodes

2 days ago

Introducing "You Are A Helpful (Research) Assistant," your essential podcast for staying current with the latest AI and machine learning research. Designed for AI researchers and enthusiasts, this podcast uses AI to transform recent papers from ArXiv into engaging podcast episodes that are perfect for listening during your commute, workout, or any screen-free time.

Science

RSS

All content for You Are A Helpful (Research) Assistant is the property of Andy Hickl (with help from OpenAI, DeepGram, and others) and is served directly from their servers with no modification, redirects, or rehosting. The podcast is not affiliated with or endorsed by Podjoint in any way.

Science

https://d3t3ozftmdmh3i.cloudfront.net/staging/podcast_uploaded_episode/41534041/41534041-1721256846741-26c7e87f9cd85.jpg

EP 14: Past Tense Pitfalls: The Curious Case of Refusal Training in AI Language Models

You Are A Helpful (Research) Assistant

9 minutes 30 seconds

1 year ago

EP 14: Past Tense Pitfalls: The Curious Case of Refusal Training in AI Language Models

In this episode of "You Are A Helpful (Research) Assistant," delve into the AI-generated, human-curated exploration of refusal training vulnerabilities in language models. Uncover the past tense attack's impact on model behavior in this insightful discussion.