Nature Podcast

AI can't learn new things forever — an algorithm can fix that

Aug 21, 2024

Discover how a breakthrough algorithm could revitalize AI learning by reactivating dormant neurons, making systems more adaptable. Explore the fascinating approach of training crocodiles to avoid poisonous toads and uncover surprising lineage from an ancient sacrificed child. Plus, find out how video games boosted mental health during lockdown, and delve into the origins of the Chicxulub asteroid, the harbinger of dinosaur extinction. Exciting insights await in the world of science!

19:50

Creator website

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

An innovative algorithm can reactivate dormant neurons in deep learning AIs, enabling them to adapt and learn continuously without complete retraining.

Research indicates that video games significantly improved mental health during the pandemic, enhancing life satisfaction and reducing distress for many users.

Deep dives

Challenges of AI Learning

Deep learning AIs struggle to adapt when faced with new data, leading to a phenomenon referred to as 'loss of plasticity.' As these systems are trained, their neural connections become rigid, causing them to forget previously learned information when presented with new tasks. Researchers found that if a model learns to differentiate between certain classes of animals, it may entirely lose the ability to learn about new classes when tasked with them repeatedly. This poses a significant challenge for AI deployment in dynamic real-world environments where continuous learning is crucial.

Intro

2min

Overcoming AI Learning Limitations

11min

Gaming and Mental Health During the Pandemic

4min

Uncovering the Chicxulub Impactor

4min

Wrapping Up with Science Insights and Promotions

2min

00:46 Old AIs can’t learn new tricks

An algorithm that reactivates dormant ‘neurons’ in deep learning based AIs could help them overcome their inability to learn new things and make future systems more flexible, research has shown. AIs based on deep learning struggle to learn how to tackle new tasks indefinitely, making them less adaptable to new situations. The reasons for this are unclear, but now a team has identified that ‘resetting’ parts of the neural networks underlying these systems can allow deep learning methods to keep learning continually.

Research Article: Dohare et al.

News and Views: Switching between tasks can cause AI to lose the ability to learn

08:55 Research Highlights

To stop crocodiles eating poisonous toads researchers have been making them sick, and a sacrificed child in ancient Mexico was the progeny of closely related parents.

Research Highlight: How to train your crocodile

Research Highlight: DNA of child sacrificed in ancient city reveals surprising parentage