Generalized embodiment via interoception sharing

Intelligence is embodied, but diverse intelligences have diverse embodiments. The economy’s body, for example, is people. Since we won’t always be able to recognize whether something is a body just by looking at it, we need a general theory of embodiment that fits diverse intelligences like the econonmy, AI, and more.

When we categorize a system as a body, we’re making a categorization decision—choosing to carve up the world a certain way because it’s useful, not identifying a natural kind. Usually, we call something a body because we see a physically connected, well coordinated system that seems to work together to take action on a collective level. It’s a system where the right hand quite literally seems to know what the left hand is doing.

Conventional bodies are physically continuous to the human eye—one singular piece. Many diverse forms of intelligence, like the economy, fail this test. Luckily, this criterion can be dropped. What really matters isn’t physical continuity between the parts of a body but how well coordinated they are. If your arm was totally deadened to you so that you couldn’t feel or control it, you might not consider it part of your body; if you had a chip in your brain that let you control a robot arm without it being attached to you, you might consider it part of your body.

To be and remain highly coordinated, the parts of a system need to be constantly sharing signals with each other—this is called a cognitive glue. The parts of the body will consequently be anything participating in the cognitive glue, sending and receiving signals to construct and fulfill its plans. The signals can be sent through any kind of material or substrate, but in general, the signals being shared are interoceptive signals. When all the parts of a system are regularly and reliably sharing interocepive signals with each other, they will behave as a body.

So in this view, a body is a system where the parts engage in regular and reliable interoception sharing. The result is a highly coordinated system like a conventional body. But this view of a body also naturally applies to diverse intelligences with diverse embodiments. An economy’s body is people because people in the economy are constantly sharing signals with each other in the form of prices.

The search for diverse embodiments is a search for systems of interoception sharing. The signals could take many forms, like bioelectric signals, price signals, or anything else, and the body could be dispersed over space and time rather than physically continuous, as the body of an economy is. As a result, the space of possible embodiments is much larger than might typically be imagined. You cannot tell at a glance whether something is or is not a body; empirical testing is required.

For the same reasons, diverse intelligences are capable of more than we might expect. If you think of a solution to a problem as consisting of a form taken by a body, then a bigger embodiment space means a bigger solution space. This creates challenges for predicting and controlling diverse intelligences, but also makes them more useful.