# Decentralized training on heterogeneous compute is promising

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2024-06-13
- **Theme:** Decentralized AI — https://www.gregosuri.com/predictions/decentralized-ai/
- **Source:** Erik Voorhees and Greg Osuri: "The Power of Permissionless" - Akash Accelerate '24 (Akash Network)
- **Watch at:** 00:18:47 — https://www.youtube.com/watch?v=4nRz_dVhE28&t=1127s
- **Canonical:** https://www.gregosuri.com/predictions/decentralized-ai/2024-06-13-decentralized-training-on-heterogeneous-compute-is-c3jp/
- **Cite as:** Greg Osuri, "Decentralized training on heterogeneous compute is promising," Erik Voorhees and Greg Osuri: "The Power of Permissionless" - Akash Accelerate '24 (Akash Network), 13 Jun 2024, 00:18:47. https://www.youtube.com/watch?v=4nRz_dVhE28&t=1127s

## Quote — verbatim

> I think we're seeing a lot of progress in the decentralized training space. I'm very very positive, especially a lot of the research papers that are coming out, like DiLoCo and whatnot, that propose training on heterogeneous, heavily distributed compute — so seems to be promising.

## Context — site annotation, not spoken

Greg responding to Voorhees's claim that efficient decentralized training is the most important unsolved problem in AI. Captions render the paper name as "dialog Co" (DiLoCo).