If you have ever used DeFi (Decentralized Finance), then you probably unknowingly utilized a pricing source called an oracle. All lending protocols rely on an oracle as well as all liquidations; in both cases, a pricing oracle provides the price of ETH to the contract based on a number from one of multiple potential pricing feeds. The reason why it works is because a price is a number. Therefore, many pricing feeds could potentially provide the same number and when averaged together, will likely produce a very reliable number. Try another question: “Was this provider able to fulfill its obligations under the agreement?” This is not a number; it’s a subjective opinion. In order for a person to determine whether the provider fulfilled the obligations, that person will need to review the booking, review the policy and compare each document. Unfortunately, standard smart contracts cannot perform such tasks; therefore, they require some form of human input or a trusted third-party to inform them of the decision. This is where “trustless” begins to quietly come to an end. I think this aspect of @GenLayer is very interesting. The intelligent contract can be written in Python and can access real-time data off the internet and utilize a language model to read evidence and provide a proposed conclusion. However, using only one model to make decisions would simply be providing yet another intermediary. Other validators would also run the same check utilizing their own models. Once they reach a consensus regarding the results of the checks, the validation process will continue. However, if any validator does not agree with the proposed conclusions, there will be additional rounds and possibly an appeals process should anyone dispute the final results. It may not guarantee that the correct decision was made, however, it does create a method for verifying the accuracy of those decisions rather than relying solely upon trust.