As AI proliferates and things on the internet become easier to manipulate, the need to ensure data and brands are verifiable is more important than ever, said CTO and co-founder of Space and Time. Scott Dykstra said on TechCrunch's “Chain Reaction” podcast.
“Not to get too religious about crypto here, but we saw it with the FTX collapse,” Dykstra said. “There was some brand trust in our organization, just like our personal savings were in FTX. We believed in it as a brand.”
However, the now-defunct virtual currency exchange FTX internally manipulated its books and misled investors. Dykstra believes this is similar to querying a database for financial records, but he operates within its own database.
And this extends beyond FTX to other industries. “Financial institutions have an incentive to falsify records, and we see it all the time, making it more problematic,” Dykstra said.
But what is the best solution to this? Dijkstra believes the answer comes through data validation and zero-knowledge proofs (ZK proofs). A ZK proof is a cryptographic action used to prove something about a piece of information without revealing the original data itself.
“It has a lot to do with whether the bad guys have an incentive to manipulate things,” Dykstra said. Whenever there is a higher incentive to manipulate data, prices, books, finances, etc., ZK proofs can be used to validate and retrieve data.
At a high level, ZK proofs work by having two parties, a prover and a verifier, who verify that a statement is true without communicating any more information than whether the statement is true. For example, if you want to know if someone's credit score is above 700, a ZK proof allows the ZK proof (the prover) to confirm that to the verifier without actually disclosing the exact number .
Space and Time aims to be a verifiable compute layer for Web3 by indexing data both off-chain and on-chain, but Dykstra said it will help others across the industry. We expect this to expand into the industry as well. As of now, the startup indexes from major blockchains such as Ethereum, Bitcoin, Polygon, Sui, Avalanche, Sei, Aptos, and many more to drive the future of AI and blockchain technology. Adds support for chains.
Dykstra's latest concern is that AI data is not actually verifiable. He says, “I'm pretty worried that I won't be able to efficiently verify whether the LLM was executed correctly.”
There are teams currently working on solving that problem by building ZK proofs and large-scale language models (LLMs) for machine learning, but they say it could take years to create them. Dijkstra said. This means that a model operator can tamper with the system or his LLM to do problematic things.
Dijkstra said there is a need for a “decentralized but global and always available database” that can be created through blockchain. “Everyone needs access. It shouldn't be a monopoly.”
For example, in a hypothetical scenario, Dijkstra said OpenAI itself would not allow a journalist to become the owner of the journal's database for which they are creating content. Instead, it must be community-owned and operated by the community in a way that is readily available and cannot be censored. “It has to be decentralized, it has to be on-chain. There's no way around that,” Dykstra said.
This story was inspired by an episode of TechCrunch's “Chain Reaction” podcast. Subscribe to Chain Reaction on Apple Podcasts, Spotify, or your favorite pod platform to hear more stories and tips from entrepreneurs who are building the most innovative companies today.
Connect with us:
X, formerly known as Twitter, here. Email: chainreaction@techcrunch.com