Scaling LLMs and Accelerating Adoption with Aidan Gomez at Cohere
Gradient Dissent: Conversations on AI
The Importance of Scalability in Neural Networks
So another experience that I've had with neural networks is that it seems like the exact details of the architecture often doesn't matter. And we end up kind of picking these architectures based on trying to replicate previous papers and not wanting to mess something up. How fundamental do you think transformers are? And maybe to make it a more well formed question, if you've ran back history a thousand times, how many of those times do you think you would get the transformer Architecture specifically? Yeah. That's a really interesting thought experiment. I don't think it's that fundamental. All you really need to do is saturate compute in a good way. You need to come up with an architecture that
Remember Everything You Learn from Podcasts
