Why GPU network can't be mainstream shortly although it is cheap and gets lots of attention?
- 1 reply
- 0 recasts
- 0 reactions
AI in Web3 mirrors Web2 trends: Hybrid systems In Web3, frontend (LLM) inference for speed and low cost, and on-chain inference for critical tasks needed variability and transparency. This balance will optimize for both cost and verifiability in decentralized AI applications.
- 1 reply
- 0 recasts
- 1 reaction
FHE brings new opportunities to privacy use cases. Composing with other cryptography tools like ZK and MPC, one could achieve decentralized privacy, the ultimate goal. Privacy is not a large sector because of the computation/user experience overhead. Will FHE change this game? There are two 100x use cases for FHE. 1. FHE for DeFi. With privacy & DeFi, we can build a true dark pool. This will start large institutions' blockchain journey. 2. FHE for consensus. Not all computation results could be verified, but some are based on majority votes. Like voting in governance, some nodes will follow leaders' results. You will never be slashed if you follow the majority vote even if it is incorrect. FHE can hide individual nodes' votes. Nodes must do the computation independently and can't copy the leader's result. (Hence the network can achieve a more accurate result. This would be helpful for AI blockchain or data labeling, evaluation cases.
- 0 replies
- 0 recasts
- 0 reactions