At a recent event in San Francisco, Sam Altman and Alex Blania made five key announcements regarding the future of World—the company formerly known as Worldcoin. Given these announcements amount to mere enhancements of the original operating model for Worldcoin, this PR blitz did little to persuade detractors of World's approach (for details, follow this link to Perplexity for a summary of current critiques of World). Unfazed by such criticisms, World and proponents of the solution continue to build out the network, orb sites, and integrations required to bring billions of unique human beings onboard.
In my view, the crypto, security, identity, legislative, and AI communities already have identified the more controversial implications of World's approach to Proof of Personhood. The Tools for Humanity retina-scanning orb, in particular, causes a spontaneous chorus of industry watchers to cry "dystopian." I'm glad to see the feedback system working so effectively.
Alternative Approaches to Proof of Personhood
There are, of course, alternative approaches that don't rely on special biological scans to establish personhood. For example, Cloudflare offers a service that combines traditional methods like CAPTCHAs with cryptographic attestations to provide a low-friction user experience. Alternatively, a blockchain called Idena offers a decentralized approach to Proof of Personhood use cases.
For reasons covered in a previous post, I find the Cloudflare and Idena approaches stronger, more practical, and more palatable than the retina-scanning orb approach that Tools for Humanity espouses. Emphasis on behavior, group affiliation, and cooperation should be at the root of any attempt to conjure up trust on the internet.
But there remains a deeper philosophical question. And that question is the subject of this post.
On the Internet, What Does It Mean to Be Human?
This time last year, I spoke at KuppingerCole's Cyberevolution conference. I titled my presentation, "Thanks to AI, the future of security is collaborative." Part of the presentation offered the supposition that it's not really possible to be human in an online ecosystem. Here's the slide that introduced this hypothesis:
My assertion then, as now, is that on the internet, nothing can be considered "human"—everything is just an avatar, agent, or bot. And here's the slide that introduced that assertion:
My conclusion remains that "cybersecurity can't rely on physical identity." Instead, I offer a new type of being—one that exists online—a genus I named Homo ex Machina. And yes, that's both a reference to the movie and a turn of the Latin phrase deus ex machina or "a god from the machine." I also offered some fun ways to refer to the species as "GPme" or "me.GPT" (note: still waiting on OpenAI to pick those up!) The following April, I got some welcomed company when DeepMind cofounder and Microsoft AI CEO Mustafa Suleyman took the stage at TED2024 to declare that AI is becoming a new digital species (> 04:52). However, Homo ex Machina is something more than a new species: it's the melding of human and agentic actors to form a fullly networked being.
A Proposal: Proof of Homo ex Machina
It's of course human nature to gravitate towards our instinctual mechanisms for trust; but the online world fully denudes us of the social emotions, chemical processes, and biological signals that our instincts depend on in the natural world. Worse, this digital habitat has very few corollaries for what we call "laws of nature" in the physical world. If we accept that the internet is a novel ecosystem with a new form of inhabitants, then establishing "personhood" is—at best—of tertiary value. What we really need is Proof of Homo ex Machina (should we say "PoHEM" and pronounce it "poem"?)
The founding impetus behind AI for Society Online was to cultivate a movement to establish a social contract for the digital world. The first part of this contract, agent-to-agent interactions, reflects our real-world experiences (such as interpersonal relationships and person-to-organization transactions). But the virtual world has few inescapable "natural laws," and so requires us to invent such laws for society online. Of course, the dark web and pockets of the internet will largely ignore such laws (similar to the story of Tombstone in the Cowboy War); but over time, the overwhelming benefits of trustful interactions will make a compelling case for joining what Martin Luther King called the Beloved Community.
When actors on the internet prove they are of the genus Homo ex Machina, they signal that they ascribe to the laws of relation (symmetry, risk, and projection) and to the laws of digital nature (yet to be defined). And by interacting through a trust protocol, we can accomplish the stated goals of privacy, trust, and security that World and others espouse.
World is Doing Deus' Work
I applaud World for drawing so much media and practitioner attention to the most critical unsolved problems to living in an agentic world. We need a much broader conversation of how to determine which entities—be they gpMe's, agents, bots, or humans—bear the right to perform important functions and transations online.
Unfortunately, the existing corpus of identity and access management (IAM) techniques used in the enterprise don't account for agentic access. Similarly, workload permissioning standards such as SPIFEE don't provide a framework for AI agent access management. What's needed is a new set of standards and protocols informed both by previous efforts in this space and by scientific studies into the nature of cooperation and trust. This is the primary project for AI for Society Online.