Latent Space: The AI Engineer Podcast cover image

Latent Space: The AI Engineer Podcast

2024 in Post-Transformers Architectures (State Space Models, RWKV) [LS Live @ NeurIPS]

Dec 24, 2024
Dan Fu, an AI researcher soon to join UCSD, and Eugene Cheah, CEO of Featherless AI, delve into the future of post-transformer architectures. They discuss innovations like RWKV and state-space models, highlighting their collaborative and open-source nature. The duo examines the challenges of multilingual training and computational efficiency, while also exploring advancements in non-transformer models like Mamba and Jamba. Tune in for insights on scaling models, 'infinite context,' and how new architectures are reshaping the AI landscape!
43:02

Episode guests

Podcast summary created with Snipd AI

Quick takeaways

  • The podcast emphasizes the significant advancements in non-transformer architectures, particularly highlighting contributions from Together AI and Recursal AI in developing efficient models.
  • Challenges related to scaling model parameters and context lengths are crucial, as researchers explore innovative alternatives to traditional attention mechanisms for improved efficiency.

Deep dives

The Evolution of Transformer Architectures

The discussion covers the state of transformer alternatives, focusing on advancements in non-transformer architectures, particularly the work from Together AI and Recursal AI. These organizations have been pivotal in developing models that range from kernel and systems programming to intricate mathematical abstractions. Noteworthy contributions included Red Pajama V2 and Flash Attention 3, which demonstrated enhanced model architectures and inference algorithms. This evolution signifies a critical shift in AI development, pushing the boundaries of what is possible with architectural alternatives.

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