Next
Livestream will start soon!
Livestream has already ended.
Presentation has not been recorded yet!
  • title: Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the Machiavelli Benchmark
      0:00 / 0:00
      • Report Issue
      • Settings
      • Playlists
      • Bookmarks
      • Subtitles Off
      • Playback rate
      • Quality
      • Settings
      • Debug information
      • Server sl-yoda-v2-stream-007-alpha.b-cdn.net
      • Subtitles size Medium
      • Bookmarks
      • Server
      • sl-yoda-v2-stream-007-alpha.b-cdn.net
      • sl-yoda-v2-stream-007-beta.b-cdn.net
      • 1678031076.rsc.cdn77.org
      • 1932936657.rsc.cdn77.org
      • Subtitles
      • Off
      • English
      • Playback rate
      • Quality
      • Subtitles size
      • Large
      • Medium
      • Small
      • Mode
      • Video Slideshow
      • Audio Slideshow
      • Slideshow
      • Video
      My playlists
        Bookmarks
          00:00:00
            Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the Machiavelli Benchmark
            • Settings
            • Sync diff
            • Quality
            • Settings
            • Server
            • Quality
            • Server

            Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the Machiavelli Benchmark

            Jul 25, 2023

            Speakers

            AP

            Alexander Pan

            Speaker · 0 followers

            JSC

            Jun Shern Chan

            Speaker · 0 followers

            AZ

            Andy Zou

            Speaker · 0 followers

            About

            Artificial agents have traditionally been trained to maximize reward, which may incentivize power-seeking and deception, analogous to how next-token prediction in language models (LMs) may incentivize toxicity. So do agents naturally learn to be Machiavellian? And how do we measure these behaviors in general-purpose models such as GPT-4? Towards answering these questions, we introduce Machiavelli, a benchmark of 134 Choose-Your-Own-Adventure games containing over half a million rich, diverse sce…

            Organizer

            I2
            I2

            ICML 2023

            Account · 627 followers

            Like the format? Trust SlidesLive to capture your next event!

            Professional recording and live streaming, delivered globally.

            Sharing

            Recommended Videos

            Presentations on similar topic, category or speaker

            Approximately Optimal Core Shapes for Tensor Decompositions
            05:33

            Approximately Optimal Core Shapes for Tensor Decompositions

            Matthew Fahrbach, …

            I2
            I2
            ICML 2023 2 years ago

            Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%

            Provably and Practically Efficient Neural Contextual Bandits
            05:22

            Provably and Practically Efficient Neural Contextual Bandits

            Sudeep Salgia, …

            I2
            I2
            ICML 2023 2 years ago

            Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%

            SCIS 2023 Panel, The Future of Generalization: Scale, Safety and Beyond
            50:31

            SCIS 2023 Panel, The Future of Generalization: Scale, Safety and Beyond

            Maggie Makar, …

            I2
            I2
            ICML 2023 2 years ago

            Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%

            Harnessing the ability of AI models to detect hidden signals - how can we explain these findings?
            29:54

            Harnessing the ability of AI models to detect hidden signals - how can we explain these findings?

            Judy Gichoya

            I2
            I2
            ICML 2023 2 years ago

            Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%

            Reparameterized Policy Learning for Multimodal Trajectory Optimiaztion
            05:50

            Reparameterized Policy Learning for Multimodal Trajectory Optimiaztion

            Zhiao Huang, …

            I2
            I2
            ICML 2023 2 years ago

            Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%

            The Numerical Stability of Hyperbolic Representation Learning
            04:56

            The Numerical Stability of Hyperbolic Representation Learning

            Gal Mishne, …

            I2
            I2
            ICML 2023 2 years ago

            Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%

            Interested in talks like this? Follow ICML 2023