English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

RL is increasingly being used to control robotic systems that interactclosely with humans. This interaction raises the problem of safe RL: how toensure that a RL-controlled robotic system never, for instance, injures ahuman. This problem is especially challenging in rich, realistic settings whereit is not even possible to clearly write down a reward function whichincorporates these outcomes. In these circumstances, perhaps the only viableapproach is based on IRL, which infers rewards from human demonstrations.However, IRL is massively underdetermined as many different rewards can lead tothe same optimal policies; we show that this makes it difficult to distinguishcatastrophic outcomes (such as injuring a human) from merely undesirableoutcomes. Our key insight is that humans do display different behaviour whencatastrophic outcomes are possible: they become much more careful. Weincorporate carefulness signals into IRL, and find that they do indeed allowIRL to disambiguate undesirable from catastrophic outcomes, which is criticalto ensuring safety in future real-world human-robot interactions.

Imitating careful experts to avoid catastrophic events