# Bittensor

https://bittensor.org/

Relevant participants:

### Miners:
- register themselves
serve models (queries routed to them by network via Proof-of-Intelligence (PoI)
- PoI is basically a testing phase of the ML training, increasing likelihood of miner being selected to serve data when request is sent to network

### Validators
- query miners, "keeping them honest"

### Decentralized Mixture of Experts (MoE)
- Leverage multiple miners (i.e. models) to fulfill a specific query

## Questions:
- How do you stop a dishonest miner from serving wrong inference results?
- You don't. The miner has to pass 1) regular PoI tests by network for increased likelihood of fulfilling requests and 2) inference sampling by verifiers so that you are producing valid inferences. However, targeted "wrong" inferences can still be delivered and miner is not slashed for that specific activity.