Techly NewsGet the app

Get the latest tech news

Executing programs inside transformers with exponentially faster inference


We build a computer inside a transformer — executing arbitrary C programs for millions of steps with exponentially faster inference via 2D attention heads.

None

Get the Android app

Or read this on Hacker News

Read more on:

Photo of Transformers

Transformers

Photo of faster inference

faster inference

Photo of Executing programs

Executing programs

Related news:

News photo

Transformers know more than they can tell: Learning the Collatz sequence

News photo

Why can't transformers learn multiplication?

News photo

Sakana AI's CTO says he's 'absolutely sick' of transformers, the tech that powers every major AI model

« China's ByteDance gets access to top Nvidia AI chips in Malaysia, WSJ reports.
Moscow’s Internet Outages Drive Sales of Pagers and Paper Maps »