Earlier Articles:
Token:
The Hypertensor Blockchain bought out its presale in January 2024 promoting out for a complete of ~$1M in DAI (a stablecoin pegged to the US greenback). Now in Might, the wrapped model of the Hypertensor token that resides on the Ethereum blockchain below the image $TENSOR is reside on Uniswap.
Hypertensor is an AI layer 1 blockchain that decentralizes, incentivizes, and democratizes synthetic intelligence. The blockchain interfaces with subnets which are peer-to-peer decentralized deep studying purposes. These subnets decentralized the internet hosting of open-source machine-learning fashions like LLMs, picture diffusers, synthetic intelligence brokers, and extra AI classes.
Subnets & Incentives
In every subnet, every peer hosts a subset of a mannequin alongside different friends internet hosting different items of the mannequin, collectively internet hosting a whole mannequin. These friends are chargeable for internet hosting fashions and producing consensus amongst friends.
Throughout every epoch, every peer should submit consensus information on every eligible peer within the internet hosting mannequin. This consensus information ensures that friends interfaced with the blockchain are verified to be internet hosting a mannequin, and are additionally submitted with a rating based mostly on computational contribution in direction of the mannequin. As soon as scores are submitted to the blockchain, all friends should agree on the scores given to every peer based mostly on their computational contributions as a server internet hosting a mannequin. The scores given to a peer by one another peer are then put by way of an interquartile algorithm that’s adopted by the averaging of the ensuing information. Friends who could also be dishonest with scoring friends and exceed the allowed delta from the typical rating are vulnerable to being eliminated as a peer and having their stake slashed.
To learn extra about Hypertensor, the Whitepaper is a good place to start out!
References
Hypertensor Testnet v1.0 Whitepaper: https://hypertensor.org/whitepaper.pdf
