Randomness in AI

AI systems use randomness when initializing weights, shuffling data, sampling tokens and exploring actions. That randomness is usually internal and not meant for public raffles.

Do not use model sampling as a public lottery

LLM temperature sampling is not an auditable raffle mechanism. Outputs depend on model version, prompts and hidden seeds. Participants cannot independently verify fairness.

Human decisions still need proof

When a company picks a winner or assigns a scarce resource, cryptographic audit trails beat “the AI chose.” EntropyDraw targets that human accountability layer.