The defining computing challenge reaches beyond adding more processors. Human programmers need one parallel body of source code whose causality they can understand and support across generations.
Brian Greenforest frames that challenge at civilizational scale: trillions of lines of future software will span workers, devices, and machines, and people must still own the result.
Parallel Hardware Needs Parallel Language
Execution spreads across locations, so the source model must preserve timing, locality, ownership, communication, and failure. A library call or orchestration layer cannot carry the entire mental model.
One coherent language can make processes, routes, state, and resource boundaries visible without forcing every developer to become a cluster specialist.
Design for the Humans Who Maintain the Future
AI tools can assist implementation, yet human programmers still choose architectures, meanings, responsibilities, and long-lived relationships. The platform must make those decisions legible.
Language researchers, accelerator builders, and distributed-systems teams can bring their strongest models to this problem. The future belongs to the system that makes parallel source code as natural as sequential code feels today.
Study the Parallel-Programming Session
The linked GTC session puts the source-language problem beside a concrete industry account of programming parallel hardware.
LinkedIn status when archived: Visible to anyone on or off LinkedIn.
Have you tried to use this for general-purpose compute? I wonder how easy it will be to write parallel SOURCE CODE for it, one orchestrated parallel easy-to-reason-about source code.
I mean, AIs are not the future. The future is more trillions of lines of code that need to be written by HUMAN programmers, and SUPPORTED by their CHILDREN in the future.
Does ANYBODY address THAT PROBLEM today???
Comments added by Brian Greenforest on LinkedIn
This comment was also preserved verbatim from Brian Greenforest’s LinkedIn data export or the public post page.
Comment 1 · (2022-05-12 01:12:25 UTC)
View the LinkedIn post