Decentralized AI is now here - Is this the future of AI?

Hypertensor
2 min readJun 12, 2024

--

The future of AI is an interesting area concerning the regulatory aspect. With California’s SB-1047 regulation that wants to limit open-source advancements by adding liability measurements for open-source builders — essentially making any developers liable for what others do, the future of AI seems to be going in an even more frightening direction than finance has gone down.

Hypertensor, a decentralized AI network pioneering the truly decentralized AI economy has released its first Devnet alongside a decentrally and incentivized hosted LLM. The model they went with is an optimized LLAMA2 model called Stable Beluga 2, created by the Stability AI development team and optimized by the BigScience development team.

— Try the model out at https://chat.hypertensor.org

While the model is slow, it’s a look into what the Hypertensor team is building. According to a tweet by Hypertensor, in the future, this model will be 20 times faster:

This is a decentrally hosted LLM. It’s *slow* but it will be 20x faster in later versions.

Hypertensor is a decentralized, incentivized, and democratized AI economy. All the nodes responsible for keeping the network of AI models up and running are rewarded with their blockchain’s native token, TENSOR. This is a decentralized market where anyone can spin up a validator node for each AI subnet hosting a variety of AI categories and get rewarded for doing so. Each model, via a subnet, gets voting in, and also out, by undergoing a democratic voting process, similar to a governance vote for smart contract based protocols, such as Aave and Uniswap. Validators on the blockchain are responsible for vetting and validating that these subnets are doing what they say they’re doing by ensuring inference is being accomplished per peer and the subnet AI nodes are live.

We all eagerly anticipate the transformative innovations the Hypertensor team will bring to decentralized artificial intelligence.

--

--