Equilibrium Infra Bulletin #44: Reintroducing trust on the Internet, State of Proving, Blob Aggregation, and more...
Equilibrium designs, builds, and invests in core infrastructure for the decentralized web. We are a global team of ~30 people who tackle challenges around security, privacy, and scaling.
🔍 Can Blockchains And Cryptography Solve The Authenticity Challenge?
⚡️ Topic Summary
As generative AI improves, it becomes more difficult to distinguish between AI (“fake”) and human-generated (“real”) content (the authenticity challenge). This accelerates the deterioration in trust we’re already experiencing on the Internet. More specifically, the main problems relate to crime and fraud, political influence, and the loss of human connection.
For content verification in a post-AI world, there are two main paths:
Verifiably AI = Proving content is AI-generated. Either through watermarking (invisible fingerprint in all AI-generated content) or statistical methods (often AI models themself which are trained to detect AI-generated content with a certain probability).
Verifiably Human = proving content is NOT AI-generated. Either through digital signatures (requires hardware-software integration for stronger guarantees) or some proof of humanity (e.g. zk-passport or biometric identity).
Both approaches have their unique benefits and drawbacks:
In solving the authenticity challenge, cryptography does the heavy lifting while blockchains can help with stronger guarantees around tamper resistance and easier verification. Specific use cases include privacy-preserving watermarking, attested hardware, verifiable editing software, or leveraging the blockchain for provenance tracking.
🤔 Our Thoughts
Gen-AI has accelerated the deterioration of trust on the Internet, but cryptography and blockchains can help restore it. While blockchains alone don’t guarantee authenticity, they can provide stronger guarantees around tamper resistance and easier verification.
When it comes to the question of which approach we should pursue to verify authenticity - provably human or AI - the answer is not so much one or the other. Instead, we should pursue both.
💡 Research, Articles & Other Things of Interest
📚 State of the Proving Infrastructure Landscape - 2024 Q4: Latest updates for the proving landscape include Aztec’s proving & sequencing testnet, Gevulot's rebranding to ZK Cloud, and more. Also some analysis on economic security in the context of proof generation (proof mining vs Bitcoin mining, staking vs re-staking, etc).
📚 Introducing Seismic: By leveraging secure hardware (more specifically, Intel TDX), Seismic aims to bring baselayer privacy to blockchains. While we believe MPC-based infra gives the strongest privacy guarantees, modern TEEs can provide an interesting tradeoff between privacy and performance overhead.
📚 Blob Aggregation - Step Towards More Efficient Blobs: By combining multiple rollup’s blobs into one, the limited blobspace is used more efficiently and smaller rollups only pay for what they use (rather than being forced to pay for full blobs).
📚 Chains Will Be Abstracted, Whether You Like It or Not: What differentiates chains in a world where most users don’t need to care about which chain they use (app-centric vs chain-centric users)? Devex and performance, rather than factors like existing userbase or liquidity.
🤌 Personal Recommendations From Our Team
📚 Reading: The Gig Economy - Zero HP Lovecraft: Starts off feeling like a random blog rant before turning into Lovecraftian horror.
🎧 Listening: Hot Chocolate - Heavens In The Back Seat Of My Cadillac
💡 Other: Jack Dorsey = Satoshi Nakamoto?