I admit I am a blockchain and web3 enthusiast, not an AI expert. Yet, the idea of AI and blockchain synergizing the technologies into something transformative is exciting. But I have seen many people pose the question of whether blockchain intervention is needed at all, or if the traditional AI system is working fine by itself and is better than any decentralized perspective. Hear me out.
It is known that the inception and development of AI have been happening over a long time, although the trend of AI applications is fairly recent. For the average user, ChatGPT or Gemini are tools imbuing excellence while the developers work with Large Language Models (LLMs) and machine learning (ML) in the backend. But who is talking about responsible AI? In an interview last year, the Head of Enterprise Solutions of Oasis Labs, Vishwa Raman, explained the need, scope, and impact of decentralized AI (DeAI).
Provenance for data authenticity, transparency for eliminating bias in favor of fairness, democratization through decentralization, and breaking data silos are some of the immediate effects of DeAI that will go a long way in establishing a new paradigm of AI.
An aspect of developing and training AI models that sometimes gets sidetracked is ML verification. Now, fascinated by the capabilities of AI agents and how easy they are to use, users are not always bothered to think about digital footprints and trusting AI while sharing and exposing private and sensitive information. DeAI has a potential fix.
The zero-knowledge, optimistic, and trusted execution environment (TEEs) methods seem to be top choices; however, other options like Oracle Networks or Fully Homomorphic Encryption ML can also be explored in verifying before trusting AI.
Now, a developer, testing the waters of DeAI, will need to work with privacy-preserving techniques to make the handling and processing of humongous datasets by AI safe and secure. While DYOR has no alternative, Oasis has been working with TEEs forever and offers their insights. The most critical benefits of TEEs can be summarized easily:
Being isolated hardware environments that securely run code, the autonomy and verifiability capabilities are also extremely tamper-proof, as access to outside parties, even the developers and operators, is shut down.
System integrity and authenticity are ensured with remote attestations
Ultimate solution for the issue of private key custodianship
Another reason for trusting TEEs in the quest to build next-gen trustless AI applications by taking the DeAI route is the synergy of on-chain confidentiality and off-chain verifiability. The ROFL (runtime off-chain logic) framework developed by Oasis has a crucial role to play here.
A lot to ponder over, right? Maybe this recent conversation between the Director of Engineering at Oasis Labs, Peter Gilbert, and the Head of AI at Oasis, Marko Stokic, will help you with some perspective.
Let me know in the comments section how intrigued you are with the decentralized perspective of taking AI to the next level, where innovation meets transparency and privacy.