Deep in ancient seas, around 500 million years ago, animals first cohered into genuine subjects with points of view, integrated systems for perception and action. During the Cambrian explosion, animals went from being chill blobby things that could float around without needing to perceive much, to mobile predators and prey that were harder, faster, and more attuned to the world.
This newfound mobility kicked off a murderous evolutionary arms race—and, possibly gave rise to subjectivity.
In his magisterial trilogy of books about the evolution of minds (Other Minds, Metazoa, and Living on Earth), Peter Godfrey-Smith centers the concept of a subject with a "point of view". As we face down venerable puzzles about qualia and the redness of red, he writes, the point of view is "a useful bridging concept" between matter and mind—and so "an important resource" for tackling the hard problem of consciousness.
(The quotes in this post come from a paper that the books draw on, Evolving Across the Explanatory Gap).
What’s so special about having a “point of view”? After all, doesn’t a rock have one? It’s situated in space, light hits it from some angles but not others, and it reacts to changes in its environment. But we’re looking for something more robust, something that we and our fellow creatures have but rocks lack. One deficiency to notice about the rock’s point of view is that the rock’s boundaries don't carve out anything interesting; half the rock would make a "subject" just as well as the whole rock.
Not so with animals. Animals are integrated wholes with self-maintaining boundaries. They keep the outside world out with membranes, and all of their internal components have to work together. And mobile animals represent another degree of subjectivity, possessing “points of view in a sense that does not apply to other kinds of life". In particular, Godfrey-Smith writes, they develop "an outward-directed form of subjectivity."
An outward-facing subjectivity that arose because of a growing demand, among ancient sea animals, for murder.
The stage was set for the Cambrian explosion once larger, successful animals became "concentrations of nutrients.” These new pockets of nutrients presented an opportunity for other animals. There was now more to be gained from moving around in search of nutrients instead of just soaking them in. First animals scavenged, and then they started killing. This created a lot of selection pressure for better motion, quicker decisions, sharper senses. (The Cambrian period is also where we start seeing fossils, because this harsh era is when animals first built hard, mineralized body parts like shells and exoskeletons.)
But motion brings its own host of problems, what neuroscientist Björn Merker calls "liabilities of mobility." Every time you move, you create sensory changes that have nothing to do with changes in your environment. Did you cause that shadow, or is an Anomalocaris sneaking up on you? Better decide quick!

A sessile (immobile) organism never has to figure any of this out—to a sea anemone, any sensory flux must come from external events. (And besides, where’s it going to go anyway?) But mobile animals have to constantly untangle the effects of their own actions; you can’t be startling yourself every time your claw moves across your field of vision.
This creates a computational challenge. A mobile animal needs an internal model that tracks its own movements and their expected sensory consequences (“reafferance”), and compares them to sensory changes caused by the world (“exafferance”). The construction of this model amounts to "the organism's own registration of the self/other divide".
Disorders of reafference in humans make vivid how indispensable it is to a well-functioning mind. “Self-suppression”, i.e. the dampening of the signals produced by your actions, is notably impaired in schizophrenics. This leads to the characteristic “passivity” delusion, i.e. external forces are causing one’s actions—without the “I did that” tag, the patient’s actions are attributed to a microchip’s commands, and routine mental chatter becomes sinister external voices. People with schizophrenia perform anomalously on unrelated tasks that require calibrating their actions, like pressing their fingers to match a given pressure. Malfunctioning reafferance is also implicated in alien hand syndrome, and might contribute to phantom limb phenomena and Tourette’s syndrome.
The fragility of our sense of self highlights what an evolutionary achievement it was for ancient animals to start solving these problems in the first place. What began as a liability of mobility—the confusion of self-generated sensory noise—became the foundation for a richer kind of internal life. "Once animals start to accommodate and utilize reafference," Godfrey-Smith writes, "the character of sensing changes. The animal is now not only open to the world, but open to the world as the world, as distinct from self." By learning to subtract out their own contributions to their perceptual worlds, Cambrian creatures carved out space for a self.
My understanding of PGS's view is that he thinks that metabolism and autonomously maintaining and caring for oneself are necessary conditions for a sharp self/other boundary (and therefore for having a point of view and consciousness). That makes sense if we're talking about animals, but artifacts also can be clearly distinct from their environments, unlike rocks. Current language models are already sharply distinct from each other and from other software, but none of them autonomously maintains itself.
Also, you might enjoy this video: https://www.youtube.com/watch?v=6YsNRnZRgg8
The best