# Crack distributed training and we crack AGI much sooner

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2024-07-17
- **Theme:** Decentralized AI — https://www.gregosuri.com/predictions/decentralized-ai/
- **Source:** Evolution to AGI Panel at Web3\_AI Day \| Encode Club \| Linera, Ritual, Nillion, Akash Network (Linera: Real-Time Blockchain)
- **Watch at:** 00:13:45 — https://www.youtube.com/watch?v=-XfPrBtBiKA&t=825s
- **Canonical:** https://www.gregosuri.com/predictions/decentralized-ai/2024-07-17-crack-distributed-training-and-we-crack-agi-much-s-kxug/
- **Cite as:** Greg Osuri, "Crack distributed training and we crack AGI much sooner," Evolution to AGI Panel at Web3\_AI Day \| Encode Club \| Linera, Ritual, Nillion, Akash Network (Linera: Real-Time Blockchain), 17 Jul 2024, 00:13:45. https://www.youtube.com/watch?v=-XfPrBtBiKA&t=825s

## Quote — verbatim

> It begins with compute, right? Access to compute is fundamental — I mean, the spice must flow... We saw promising work by Google DeepMind, the paper called DiLoCo... heterogeneous training, federated training — a lot of work happening where you can leverage GPUs that are globally distributed and optimized for latency. So when we crack that, I think we can crack AGI much sooner.

## Context — site annotation, not spoken

Answering what web3 adds to the pursuit of AGI; argues crypto incentivization "can go places where traditional, closed-source software cannot."