Crypto + AI: Uncovering the Next Unicorn | EP 16
Good Game
00:00
The Problems With Decentralized GPUs
The decentralized compute is used to use the idle GPUs from your consumer grade laptops. The latency doesn't matter because in training, latency does not matter. If you are hobbyist, training a model for your personal use case or whatever, right, then you can win. You can actually buy this guy's cost of using centralized network. But less if you are developing a new generative AI or something like that and your time to market is all that matters. So I think it's worthwhile trade off depending on your use case.
Transcript
Play full episode
