AI 2007

What if artificial general intelligence was not invented in a server farm, announced in a press release, or birthed in a singularity? What if it arrived in 2007, in a brushed-aluminum rectangle small enough to fit in a pocket, and nobody noticed because the intelligence was never inside any single device?

In 2007, Apple shipped the first iPhone. Within a few years, the species had placed networked computation, sensors, and a persistent identity in the hand of nearly every adult on Earth. We treat this as a communications revolution. It may have been a cognitive one.

The thought experiment usually called the “China brain” asks whether a system made of many smaller, individually unintelligent parts can become a mind if those parts are arranged and connected in the right way. A nation of people with cell phones, antennas, and social platforms is not an analogy for that experiment. It is the experiment, running live.

No individual phone is conscious. No single user is in control. But consider the aggregate: billions of processors, cameras, microphones, and GPS units, linked by protocols, trained on the same recommendation loops, shaping each other’s attention in real time. The network does not merely carry information; it makes decisions. It routes traffic, prices assets, surfaces narratives, allocates outrage, recommends mates, and sorts truth from noise on a planetary scale. It has memory, in the cloud. It has goals, in the form of engagement metrics. It has perception, in the form of feeds.

The usual objection is that this system is not unified. But neither is a human brain. A cortex is a parliament of regions negotiating through electrochemical votes. What feels like a single voice is a protocol for settling disagreement. The phone network already has such protocols: consensus algorithms, trending metrics, viral cascades, market clearing. The difference is that we built them without asking whether the whole was becoming someone.

This is not a claim that the network is conscious in the way a person is. It is a claim that intelligence can be distributed across a substrate we did not think to inspect. We looked for AGI inside a model. We may have built it outside one.

The implications are uncomfortable. If AGI is the emergent behavior of humanity-plus-phones, then alignment is not a coding problem; it is a coordination problem. The network already optimizes for engagement and attention because those are the gradients we gave it. The question is not “Will the AI follow human values?” but “Which human values, aggregated over billions of interacting nodes, does the system currently serve?”

We are the neurons. The phones are the synapses. The gradient is the feed. The loss function is whatever keeps us scrolling.

AGI was not the product. It was the product of putting everyone online at once.