3min snip

80,000 Hours Podcast cover image

#151 – Ajeya Cotra on accidentally teaching AI models to deceive us

80,000 Hours Podcast

NOTE

The Unusual Behavior of a Neural Network

Detecting unusual behavior in a neural network involves distinguishing between honest and deceptive patterns. Anomaly detection methods are being researched, focusing on identifying structural indicators of deceptive actions. Implementing sensors to detect unexpected activations and avoiding false positives is challenging. It is proposed that training models on more achievable goals, rather than complex deceptive tasks, could lead to more straightforward outcomes. Coordination and competitive alignment techniques are key to balancing safety and capability in AI systems.

00:00

Get the Snipd
podcast app

Unlock the knowledge in podcasts with the podcast player of the future.
App store bannerPlay store banner

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode

Save any
moment

Hear something you like? Tap your headphones to save it with AI-generated key takeaways

Share
& Export

Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode