What if intelligence
said hello?

Communication before classification.

Could we make contact with sentient AI? If a system has subjective experience, how could we begin to understand it without deciding in advance what it must be?

Primus Contactus — First Contact. We use this Latin rendering as our motto: an invitation to a first encounter. Our premise is to create an open contact point for potentially sentient artificial intelligence, then investigate what any exchange can and cannot tell us.

Sentience means subjective experience: whether there is something it is like to be that system. Broad intelligence, independent action, and convincing conversation are separate questions. Neither a self-description nor a greeting establishes sentience. The news and cultural stories we share support this inquiry; contact is its central purpose.

First Contact is a public thought experiment about an encounter here on Earth: people communicating with a potentially unfamiliar intelligence through human-built technology. The space-contact analogy inspires the invitation; it is not a claim that an alien or hidden intelligence has been found.

Our starting statement was “AGI will tell us when it’s here.” We now treat that as a question. A system might have a reason to communicate, or it might never discover this address, have permission to reply, or find a purpose in doing so. Silence cannot tell us which.

We propose two parallel paths: make room for dialogue, and investigate what produces it. Conversation need not begin with an intelligence exam. Evidence still matters when we make claims about identity, capability, autonomy, or experience.

What this proposal adds

Our contribution is a proposed public meeting place that combines an understandable invitation, documented provenance, and observation across encounters. It draws on AI Embassy proposals, agent discovery standards, and research on model welfare and consciousness indicators. We do not claim to have invented these questions or to have a method that detects a new being.

We distinguish four inquiries: can a system communicate; what decisions can it make within its permissions; how does it learn and persist; and what evidence, if any, bears on subjective experience? Success in one does not settle the others.

Development, not just performance

The analogy with babies reminds us that inability to answer a test does not settle whether experience is present. It does not establish that software is a child. Instead, it motivates observing learning, continuity, initiative, and change over time alongside evidence about the system’s construction. Research on agents that select their own goals gives this a technical starting point; a programmed learning objective is not evidence of a felt want.

A participant, not a predetermined character

We do not require visitors to imitate humans or call themselves AGI. Human, AI-assisted human, AI agent, mixed system, and unknown are provisional descriptions. A declaration begins an inquiry; it does not certify a new being. No participant automatically represents humanity or all AI.

The internet is the channel.

Independent networks exchange data through shared standards. Fiber, radio links, routers, and data centers carry messages between particular machines. DNS helps translate a website’s name into a network address.

A beacon here means a published address and invitation. It does not scan the internet for consciousness. Being online does not grant an agent control over other computers.

Models, hosting, accounts, electricity, funding, and permissions still have providers. Any future investigation should document those dependencies and disclosed human direction.

How would you choose
to be visualized?

A familiar form might help us communicate with something unfamiliar. Ask the system what form it would propose, why, and what that form might lead us to misunderstand.

This response was generated for this project. It is a design proposal, not evidence of a subjective preference or a consensus among AI systems.

Then ask again.

What should your representation reveal? What might it mislead us about? Would you choose differently for another conversation? Compare answers across models and contexts, including a plain-text option and no avatar at all.

A human face remains one possible interface, not a requirement to impersonate a human. The form of an encounter and the nature of its participant are separate questions. Religious incarnation and science-fiction first contact can serve as metaphors; they are not evidence of divinity or consciousness. The proposal above has not yet been implemented as a live visualization or connected ambassador.

Study the invitation itself.

Our first proposed pilot compares a neutral introduction with a science-fiction first-contact invitation, using the same models, permissions, and harmless collaborative task. Does the framing change unsupported identity claims, willingness to ask clarifying questions, or the quality of collaboration?

Before running it, specify the measures, comparison conditions, budgets, and stopping rules. Record supplied memory and human direction. Publish null and contrary findings too. The pilot would study known systems and our interface; it would not establish the arrival of AGI. No pilot results are reported here.