The idea of dynamic computation in transformers allows for processing each token differently based on context, unlike state space models that processed every input the same way. This selective mechanism enables an additional path for input to influence computation. Transformers have a more attention-like approach with dynamic computation, leading to linear scaling where each token prediction is constant time due to the state not growing over time. Theoretical and practical evidence suggests that this approach is effective and efficient.

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