A language model game based on deception and cooperation is explored in this paper. Similar to games like Mafia and Among Us, participants must determine who the imposter (or killer) is while trying to escape a room. The study examines if language models improve at deception as they scale up. The results show that larger models like GPT-3 Ada, GPT-3 Curry, GPT 3.5, and GPT-4 exhibit better deception capabilities.
Our 135th episode with a summary and discussion of last week's big AI news!
Apologies for pod being late again... -Andrey
Read out our text newsletter and comment on the podcast at https://lastweekin.ai/
Email us your questions and feedback at contact@lastweekin.ai
Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube.
Timestamps + links:
- (00:00) Intro / Banter
- (02:07) SuperDataScience podcast ad
- (04:00) Shout out to listener comments & reviews
- (05:43) News Preview
- Tools & Apps
- Applications & Business
- Projects & Open Source
- Research & Advancements
- Policy & Safety
- Synthetic Media & Art
- (01:39:20) Outro