Tools

Explore

Videos Channels Figures

Atmrix

Tools

Explore

Videos Channels Figures

Atmrix

JD

Jeff Dean

05/12/25

@ Sequoia Capital

The future of computing will involve running incredibly large neural networks with high performance and low power consumption, requiring different solutions for training and inference workloads.

Video

Google’s Jeff Dean on the Coming Era of Virtual Engineers

@ Sequoia Capital

Google’s Jeff Dean on the Coming Era of Virtual Engineers

05/12/25

Related Takeaways

a16z Cast

03/21/25

@ a16z

The future of computing will involve asynchronous workloads where AI systems perform extensive research before returning results, necessitating robust infrastructure to handle increased loads.

Lex Fridman Cast

02/03/25

@ Lex Fridman

The scale of mega cluster buildouts for AI is unprecedented, with data center power consumption expected to rise significantly, driven by the demands of AI training and inference.

Lex Fridman Cast

02/03/25

@ Lex Fridman

The nature of data center tasks is shifting from traditional database access to inference and training, which require thousands of GPUs to be connected in high-speed networks for efficiency.

Mark Zuckerberg

07/30/24

@ NVIDIA

In the future, we will see a proliferation of different AI models, with companies using the largest models for performance optimization.

Jeff Dean

05/12/25

@ Sequoia Capital

The future of inference hardware may include analog computing due to its power efficiency, alongside advancements in specialized digital hardware.

Jeff Dean

05/12/25

@ Sequoia Capital

Scaling and training very large neural networks has proven effective, as demonstrated by our training of a neural network 60 times larger than previous models using 16,000 CPU cores.

Jeff Dean

05/12/25

@ Sequoia Capital

The adaptation of compute platforms is essential for running capable models in low power environments, such as smartphones and autonomous vehicles, as well as in large-scale data centers.

Lin Qiao

05/19/25

@ Sequoia Capital

The future of scaling inference will focus on heavy customization tailored to specific applications, transforming it into a multi-dimensional optimization problem.

Jeff Dean

08/08/17

@ Y Combinator

We designed a system for both training and inference, not just a chip for training, to handle large problems effectively.