"We saw that during the FTX collapse, not to get too cryptographically religious here," Dykstra stated. "We had a company with some brand credibility; for example, I invested my whole life assets in FTX. I had faith in them as a company.

However, investors were being misled by the now-closed cryptocurrency exchange FTX, which was internally altering its records. According to Dykstra, it would be similar to changing financial records within their own database through a query.

Furthermore, this applies to other industries outside FTX. The desire to alter their records is incentivized for financial firms. Thus, it gets increasingly problematic when we witness it frequently, according to Dykstra.

But how may this be resolved most effectively? According to Dykstra, the solution lies in data verification and zero-knowledge proofs (ZK proofs), which are cryptographic techniques that demonstrate a point about a piece of data without disclosing the original data.

According to Dykstra, "a lot depends on whether there's a motivation for bad actors to want to manipulate things." ZK proofs can be used to obtain and validate data whenever there is a greater motivation for people to falsify pricing, data, books, money, or other areas.

In essence, ZK proofs involve two parties—the prover and the verifier—confirming the truth of a statement while providing no more information beyond the statement's correctness. For instance, if a ZK proof—a prover—is in place, it may verify to the verifier—without really revealing the number—whether or not someone has a credit score higher than 700.

By indexing data both off-chain and on-chain, Space and Time seeks to provide that verified computing layer for web3, but Dykstra sees it growing outside of the business and into others.Currently, the firm is adding support for new chains to enable the future of AI and blockchain technology. It has indexed from key blockchains like as Ethereum, Bitcoin, Polygon, Sui, Avalanche, Sei, and Aptos.

The latest worry Dykstra has expressed is that AI data isn't truly verified. "I worry that we won't be able to effectively confirm that an LLM was carried out correctly in the future."

Large language models (LLMs) and machine learning teams are now working on developing ZK proofs to address this problem, although Dykstra noted that the process can take years. This implies that the model operator has the ability to manipulate the system or LLM in order to cause issues.

According to Dykstra, a "decentralized, but globally, always available database" is required, and blockchain technology can provide it. "Access to it must be available to all; it cannot be a monopoly."

For instance, Dykstra stated that in a fictitious situation, OpenAI cannot own a journal database on which journalists are producing material. It must, instead, be something that is publicly accessible and uncensorable that is run and controlled by the community. There is no getting around the fact that everything must be decentralized and on-chain, according to Dykstra.