Most (all?) AI software requires GPUs in order run. As a result, GPUs are hard to come by, and many of them are hoarded by well-financed AI teams. Data centers full of GPUs also become a target of terrorist and regulatory attacks.
However, everyone that has a high end gaming computer likely has a GPU that often goes unused. There are several projects in the works to turn them into a mesh network.
Such mesh networks may substantially reduce the cost of GPU compute as well as make the network resilient to terrorist and government attack.
"The idea, Fielding said, is to connect all of the machine learning capable compute hardware in the world that is capable of doing machine learning training work—including GPUs and CPUs—and make it accessible to engineers, researchers, and academics."
https://decrypt.co/144068/gensyn-ai-secures-43m-for-decentralized-machine-learning-led-by-a16z
"Companies large and small are scrambling to hoard GPUs, which are now "considerably harder to get than drugs." [Elon]
Some large companies are not only stockpiling GPUs, they are actively lobbying governments to establish "safety" regulations, creating barriers for smaller players.
It's the familiar pattern of regulatory capture dressed in sheep's clothing, a blatant power play to stifle competition.
But what if we could level the playing field?
What if we could enable unstoppable, universal access to GPU compute?"
https://gputopia.ai/
However, everyone that has a high end gaming computer likely has a GPU that often goes unused. There are several projects in the works to turn them into a mesh network.
Such mesh networks may substantially reduce the cost of GPU compute as well as make the network resilient to terrorist and government attack.
"The idea, Fielding said, is to connect all of the machine learning capable compute hardware in the world that is capable of doing machine learning training work—including GPUs and CPUs—and make it accessible to engineers, researchers, and academics."
https://decrypt.co/144068/gensyn-ai-secures-43m-for-decentralized-machine-learning-led-by-a16z
"Companies large and small are scrambling to hoard GPUs, which are now "considerably harder to get than drugs." [Elon]
Some large companies are not only stockpiling GPUs, they are actively lobbying governments to establish "safety" regulations, creating barriers for smaller players.
It's the familiar pattern of regulatory capture dressed in sheep's clothing, a blatant power play to stifle competition.
But what if we could level the playing field?
What if we could enable unstoppable, universal access to GPU compute?"
https://gputopia.ai/