DIVERGENT INTELLIGENCE

What next?

Whether intelligence arrives from another world or emerges from our own machines, the problem is the same: how do we understand an entity that does not think as we do?

Review the contributions

Who we are

A research group devoted to alien semantics: communication between divergent intelligences, from extraterrestrial life to artificial minds.

Founded on primary research in message reconstruction, and non-human communication when no shared language, culture, or biological substrate can be assumed.

Where listening programmes have treated a signal as an endpoint, we treat it as a beginning. We do not stop at carrier and waveform. We push through to form, syntax and meaning. Because a message found is not a message understood.

Minds without shared worlds. Intelligence without common ground. The Arrival Institute studies how meaning can be reconstructed when the sender, the language, and even the ontology are unknown.

The mission

To study communication and control across divergent intelligences—from unknown biological life to artificial agents whose internal logic may be equally alien.

Main contributions

Vol. 01 / Sec. 04
01

A message can disclose its own geometry

A universal, zero-knowledge method reconstructs the dimensional form of non-random data—even when sender, encoding and intended medium are unknown.

In depth
02

Non-random data carries spatial memory

Geometric and topological dimensions can be recovered from structure intrinsic to a signal, without a shared codebook or prior assumptions.

In depth
03

Perfect alignment is formally unattainable

Managed diversity among influenceable agents offers a contingent safety strategy when complete prediction and control cannot be guaranteed.

In depth
04

Messages that travel across scale

Self-similar fractal carriers can encode information across multiple time and space scales reducing assumptions about a receiver’s space-time scale.

In depth
01

Science News
June 2023

Science News feature / June 2023

Here’s how we could begin decoding an alien message using math

Science News · Matthew Hutson

The feature explains how the team’s mathematical method searches possible dimensions and detects local and global order, allowing a non-random message to reveal the geometry in which it was intended—even under noise.

Read the Science News feature
02

Nautilus
June 2026

Nautilus / June 2026

Looking for Signs of Intelligence in Chatbots

Nautilus · Kristen French

A new test for AI suggests some newer LLMs are less smart than older models. Hector Zenil and colleagues at King's College London devised a way to assess artificial "superintelligence" and put leading language models to the test.

Read the Nautilus feature