How Bots, Deepfakes and AI Agents Are Forcing a New Internet Identity Layer | Alex Blania on a16z

The concept of "proof of human" addresses the increasingly complex challenge of distinguishing human users from AI agents or bots online. This problem is escalating rapidly, with current bot activity being less than 1% of what is anticipated in the next year or two. The core issue is verifying that an individual interacting on a platform is a unique human, ideally associated with only one account, and remains in control of that account. This differs from simply authenticating a user (like Face ID on a phone) because proof of human requires distinguishing a new individual from all previous individuals in a network, a "one-to-n" problem rather than a "one-to-one" authentication. Three main approaches to proof of human were considered and largely dismissed. The first was a "web of trust" model, relying on past online behavior and attestations from known individuals. This was rejected because AI agents can easily mimic human behavior, create multiple accounts, and even attest to other AI agents. The second approach involved using government IDs. This was deemed problematic due to concerns about free speech, loss of anonymity, and the unsuitability of existing government identity systems for a global, internet-scale problem. The third approach, biometrics, initially faced skepticism due to privacy concerns and the difficulty of ensuring uniqueness at a large scale. The speaker's company, Worldcoin, pursued a biometric solution focusing on iris recognition due to its high entropy, making it sufficiently unique for a large number of users. To address the historical vulnerability of biometrics to replay attacks, Worldcoin developed a device called the "Orb." The Orb uses multiple sensors across the electromagnetic spectrum to prevent fraudulent inputs, such as showing a display instead of a real eye. Privacy in a biometric system is a critical challenge, especially when verifying uniqueness against a global database. Unlike one-to-one authentication where biometric data can remain on a personal device, checking uniqueness requires comparing an individual's biometric data against all previous registrants. Worldcoin tackles this using multi-party computation and zero-knowledge proofs. When an individual verifies with an Orb, their iris image is used to calculate an iris code, which is then split into multiple pieces and distributed among several computers. This ensures that no single entity holds complete biometric information. These distributed pieces are then used in a computation to verify uniqueness without any party ever having access to the full iris code. Additionally, zero-knowledge proofs allow users to prove their uniqueness to platforms without revealing their identity or any personal data to Worldcoin or the platform itself. This approach aims to provide strong privacy guarantees while still enabling anonymous verification of human uniqueness. The need for proof of human extends beyond social media to various online interactions. Dating apps, for example, are already using World ID to verify that users are human, and eventually, that they are the person they claim to be. Video conferencing is another area where deepfakes pose a significant threat, making it difficult to ascertain if one is interacting with a real person or an AI. Gaming is also impacted, as players increasingly want to ensure they are competing against other humans rather than superhuman AI. Even the content creation economy, particularly on platforms like YouTube and TikTok, faces challenges as AI-generated content becomes indistinguishable from human-created content, raising questions about authenticity and monetization. Advertisers, for instance, would want to know if their ads are being viewed by humans or AI. The broader implication is that any online interaction primarily involving human-to-human connection is vulnerable to AI impersonation. The current state of Worldcoin's product shows 18 million verified users and 40 million total app users. A significant strategic shift for the company is to focus heavily on the US market over the next year, aiming to deploy around 50,000 Orbs to ensure accessibility within 15 minutes for most people. This involves large-scale distribution partnerships, like with major retail chains, as well as more localized deployments. To address the challenge of reaching everyone, Worldcoin plans to introduce "Orb on Demand," where an Orb can be dispatched to a user's location. The conversation also touched on the broader economic and societal implications of AI, particularly the need for robust identity infrastructure for government services and democratic processes. The speaker highlighted the massive fraud in government stimulus programs and social security systems, arguing that the ability to identify unique humans is crucial for efficient and secure distribution of funds. Without a cryptographically strong way to identify citizens, the integrity of democracy itself, including voting processes, could be compromised by large-scale AI impersonation. While the Orb provides the highest level of human verification, Worldcoin also offers less accurate methods like "Face Check," which uses phone cameras and multi-party computation for anonymity, but is considered a temporary solution given the advancements in deepfakes. Government ID verification with NFC chips is also supported, but platforms have shown less interest due to the associated stigma. The overall sentiment is that as AI capabilities advance, the need for a reliable proof of human solution becomes paramount, and the alternative—an internet saturated with undetectable AI agents—will be so undesirable that people will readily embrace verification methods.