The World’s Largest AI Processor with Cerebras CEO Andrew Feldman
No Priors: Artificial Intelligence | Technology | Startups
Embracing Training and Open Source Model Game
The decision to engage in the training and open source model game was driven by having significant compute power to demonstrate ease of training, building systems to train the biggest and fastest models, as well as distributing work effortlessly across a GPU cluster. The ability to train and serve multiple models easily led to releasing seven GPT models to the open source community, surpassing others in the field. Noteworthy milestones include designing a model at three billion for mobile devices and considering the challenges of scaling to networks larger than 175 billion, which pose high training and serving costs.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
