Marriage
“Marriage” – a cartoon that illustrates how decentralized computing options might provide an alternative to the centralized cloud computing platforms of Amazon, Microsoft etc.
AI language models benefit from more computing power applied to them. For example, the 10x increase in compute from ChatGPT3.5 reduced the error rate from 15% to 5%.
However, the current approach of clustering GPUs in centralized large data centers is so expensive that it is becoming the preserve of a handful of large tech companies.
The cost has direct consequences in limiting access to high performance compute, at the expense of academia, and creating an oligopoly with ethical misuse risks.
Furthermore, these cloud computing companies already enjoy large margins and are under pressure to recoup their enormous capital investments by passing on the cost to customers.
An alternative approach is applying decentralized cloud architecture to tap into latent global compute capacity that sits across everyday devices, such as Apple Macbook Pros or Sony Playstation 5s; harnessing it to solve large scale compute challenges.
For example, projects like the Berkeley Open Infrastructure for Network Computing (BOINC), which currently lists 30 active projects and nearly 1,000 scientific papers produced using its decentralized compute network.
Applying an enterprise-grade layer to this are new startup companies like Gensyn.ai, which is building a decentralized machine learning compute protocol that acts as the rails upon which AI models can train on any compute device in the word.
As we draw closer to artificial general intelligence, decentralization computing options ensure we do not have to make a trade-off between scale, trust and price.
