Although the development of AI focuses on data, as once the data is tampered with, all model training and inference results become invalid. To address this issue, a Web3 startup called Space and Time, backed by Microsoft's M12, is attempting to solve this problem.

The company decodes and organizes data from mainstream blockchains through a verifiable data indexer and ensures that both the data itself and the query process have not been tampered with using zero-knowledge proof technology, thus providing analysts and developers with reliable on-chain data support.

The ZK co-processor Proof of SQL can turn any database into a verifiable database, ensuring that AI model training is not subjected to 'data pollution.' To reduce the usage threshold, Space and Time has also created an AI chatbot named Houston based on GPT-4, allowing users to directly query data without the need for complex SQL scripts, and the query results can automatically generate dashboards or be used for model training.

Additionally, Space and Time has partnered with the globally renowned auditing firm FTI Technology to test applications in two scenarios:

First, using machine learning models to assist in identifying anomalous behaviors on-chain.

Second, verifying whether the entire data analysis process is traceable and trustworthy.

FTI clearly pointed out that if the training data itself is inaccurate, it will lead to erroneous judgments or missed reporting of suspicious transactions, which is exactly the shortcoming of traditional AI in compliance scenarios. Space and Time's architecture is entirely built on Microsoft Azure, utilizing not only the OpenAI API but also Azure’s virtual machines, security components, networks, storage, and other full-stack services. It has already launched on Azure Marketplace, where customers can deploy and use it directly.

In summary, what Space and Time offers is not just a blockchain data platform, but an infrastructure that provides a 'verifiable data root' for AI, especially suitable for scenarios such as financial risk control, auditing, and on-chain analysis that have extremely high demands for data credibility.