We interviewed Tanisha Katara, an expert in blockchain governance and the Governance Innovation Specialist at Polygon Technology.
In the last two years, she has also been an independent consultant to 6+ clients including Blockchain protocols, Decentralized Applications (dApps), and Decentralized Autonomous Organizations (DAOs).
As an independent consultant, she helps her clients create strategies and products related to go-to-market, token economics, validator reputation management, governance, and decentralization.
We asked Tanisha to explain concepts around blockchain and decentralization, and how these can democratize and improve AI system governance.
Q: Could you tell us a bit about your background and your work with Polygon?
Tanisha Katara: “I am an independent blockchain consultant who has worked with over 6 clients in the past 1.5 years. My clients include DAO tooling applications, protocols, and data availability layer.
I help them in designing governance models, reputation, crypto-economics, product strategies, and go-to-market plans. In fact, recently, I, along with my co-authors, published a first-of-its-kind research paper that dives deep into conviction voting and its impact on voter behavior.
As the Governance Innovation Specialist at Polygon, my responsibility is to find the most optimal governance solution that balances decentralization, scalability, and security.
My current project includes building a unified Governance Hub that will provide the most seamless community decision-making experience and designing the system smart contracts governance. In the past, we also introduced a decentralized process for validator admissions and created a framework on an on-chain validator reputation model.”
Q: What is a DAO (Decentralized Autonomous Organization) and what role can they play in democratizing AI?
Tanisha Katara: “So, a DAO is a type of organization facilitated by consensus and code. Let’s break that down, shall we?
In terms of consensus, DAOs typically don’t have a central authority making decisions; instead, decisions are made collectively by members of a DAO or by chosen representatives of a DAO. These members own tokens representing their stake in the organization. One fascinating aspect is that while you may not know who these members are in the real world, you can still observe the DAO’s treasury funds and its decisions.
This transparency is made possible by the blockchain, where transactions are immutable—meaning once recorded, they can’t be changed. This feature builds a remarkable level of trust, as the blockchain serves as an ultimate source of truth.
In terms of code, most rules in the DAO are in the form of smart contracts. Smart contracts are self-executing contracts that require no intermediaries and operate in full transparency.
Moving to the 2nd part of the question, AIs are fundamentally opaque and have strong data dependencies. On the contrary, DAOs are open and transparent systems. If this contrast is leveraged well, there’s a very promising future where DAOs can democratize technology.
DAOs can be the “value layer” for AIs. DAOs can facilitate decision-making such as who submits the training data, what kind of attestations are required, who is allowed to query, how many times, and -, compensate contributors for submitting critical inputs.
However, tread carefully. DAOs and AIs are a reflection of the processes that create them. The system needs to be designed critically to fight potential data poisoning attacks or collusion attacks.”
Q: What are some risks at the intersection of AI and cryptography?
Tanisha Katara: “Great question! Essentially, if an AI model has black boxes, you cannot verify its inner workings. It’s as good as a centralized body making decisions without any form of openness.
On the other hand, if an AI model is fully open, attackers can simulate attacks locally and design optimized attacks or identify loopholes in the model.
In such dilemmas, when there is a need to balance the verification and correctness of a decision while keeping the details hidden, blockchain has seen fascinating innovations like Zero Knowledge proving (ZK-snarks).
However, there is a massive amount of cryptographic overhead when you embed a ZK computation inside an AI. AIs, themselves, are computationally heavy! Even if you do figure a computational way out, the training data itself could be corrupt or biased.
Furthermore, ZK Snarks or MPCs can solve for matrix multiplication, which makes up a great deal of AI computation. It’s the non-linear layers that are a challenge and new techniques like lookup arguments are being explored.”
Q: Governance in AI: How can decentralized governance models improve the decision-making processes in AI systems?
Tanisha Katara: “Governance models for AI-based systems have to be extremely well-designed and critically thought out.
Interestingly and recently, I had a request to consult for a large client who wanted to decentralize his proprietary arbitrage bots, meaning, there was a need to involve the community to suggest improvement features, determine rules for querying the system, or even optimize for gas and revenue.
Upon further exploration, I found that the system was dependent on an AI oracle. It is important to be careful with oracles. If the oracle is attackable, that’s a huge amount of money that could disappear in an instant.
The client is now rethinking the oracle and will replace such attack vectors with a robust governance model that can help users participate in conviction voting, receive staking rewards, and slash malicious actors. Governance, if designed well, can both secure the AI system and create opportunities for value accrual.”
Q: Intersection of AI and Blockchain: Can you share some specific use cases that demonstrate the potential of combining AI and blockchain to solve real-world problems effectively?
Tanisha Katara: “Thank you for this question. I am particularly fascinated by how AI helps with Information Defense and Scam Detection.
We can now identify the threats of interacting with a decentralization application, genuinity of a token or an NFT, receive alerts on the consequences of signing a transaction and proactively detect suspicious behavior in a DAO.
In fact, Metamask, Rabby Wallet, and Cyvers are already incorporating robust AI use cases to safeguard protocols and users.
There’s also a use case for the other way around, where while AI protects the community, the community provides value to AI. There is a strong, growing need to align AI goals with the values of humanity. For instance, Open AI launched a grant program to fund ideas and tools that collectively govern AI.
This initiative is just the beginning of how DAOs can be the trust and value layer of AI, as stated earlier.”
DAOs, tokenization of real-world assets, and solutions combining AI and blockchain are set to play a big role in how companies, communities, and individuals conduct business and collaborate.
If you would like to learn more about Tanisha Katara, her research in blockchain governance, or her consulting services, you can reach out to her via her LinkedIn page.
