The following is an excerpt from the upcoming book AI Wisdom Volume 3: Meta-Principles of Agents and Groups. It’s the final installment of the AI Wisdom series that educates about how to understand what you hear about and interact with. The first book dealt with “thinking” and learning, and the second with interaction with AI. This final volume heads toward larger scale and higher autonomy (teams, agents, robots, “societies”) and the orchestration and governance meta-skills needed to deal with those levels of AI.
But it goes further. I’m educating educators, administrators, and leaders about AI and the associated issues, but also supplying a K-16 curricular roadmap, one durable enough to matter for 2040 AI too. The approach should be taught in contrast with human thinking, learning, interaction, orchestration, and governance. Because the Venn diagram is so overlapping, teaching the AI meta-skills doesn’t require using AI, and offers insight into humanity that isn’t taught either.
Let me put this section in context. In the preceding chapter I discussed agents. This section is in the Inhabiting chapter, where an agent moves into the physical world. Earlier in the chapter, I tell a story about the historically slow but recently accelerating progress in robotics and the extra cognitive burdens robots have by occupying space with us.
Later in the chapter, I describe the causes of the recent robotics explosion and the remaining issues, including an optional deeper dive into the cognitive architecture inside them. I talk about the social and trust issues we and robots need to work out, when our bodies become the platform for AI-infused hardware, some super creepy and some we’ve gotten used to (e.g. hearing aids). Finally, I talk about how the data robots and other sensors in the real world collect will very rapidly fill in a major text-based GenAI weakness, for better and worse.
This section says what robots understand about the world isn’t necessarily what we get, in large part because it has a different body.
Enjoy!
What It’s Like to Be a Robot
A lot of times when younger children want to be held, I think it’s because they get a different vantage point. Walking around at chest level or on a special day atop shoulders lets them see the world very differently.
Every human experiences the world somewhat differently. Color blindness, hearing differences, body size, mobility, and a bunch of other variations change what information reaches us and what we can do with it. Attention changes the experience too. Someone attuned to the surroundings will literally witness a different world than someone with tunnel vision. An artist will not see the room the way a carpenter does, and the carpenter’s possible actions in that room are different too.
If a child were somehow always seeing the world at adult height, their brain would learn to experience the world that way instead of a brain on a three-foot-tall body.
Many fields such as linguistics, psychology, and neuroscience have for centuries in some cases expressed that thinking and the body are intimately related.[1] Linguists tell us our language is in many cases body driven.[2] Theories of the utility of emotions often begin with the mind-body link. Philosophy has long considered embodiment critical for human thinking. These embodied cognition theories, in the modern constructs, say that the body not only affects thinking but that it is a key factor in shaping it.
The embodiment argument in AI is often expressed by what disembodied AI is incapable of knowing, like fully understanding us. Knowing from language that falling hurts is different from having an intelligence whose entire learning history has been shaped by balance, gravity, injury, reach, fatigue, and all the other consequences of occupying a body.
Even then, robots might understand “falling” without understanding “hurts” in the same way. The cognition grows around the body it has. It learns from the information that body repeatedly supplies, the actions available to it, and what happens when those actions succeed or fail. Change the senses or the available actions and, within limits, the mind can adapt around the change.
Consider Daniel Kish, who lost his sight as an infant. He learned to be bat-like, making sharp clicks with his tongue and listening to the echoes. The returning sound tells him about walls, doorways, trees, parked cars, and other features around him. He became skilled enough to bicycle and hike using echolocation and has taught the method to many other blind people.[3]
Brain imaging found that echoes activated parts normally associated with vision in expert blind echolocators.[4] That means it was recruited for echolocation, but it’s not clear whether that happens because it’s available, its existing organization was useful for the task, or both. Cortical regions are not always permanently restricted to processing the kind of signal they usually receive. When the available information changes, especially during development or through extensive practice, some regions can be recruited for different work.[5]
It doesn’t have to be as exotic as echolocation; each of us experiences incorporation of objects into cognition. A practiced driver does not consciously calculate the edges of the car every time a road narrows. A blind traveler can feel the end of a cane rather than thinking primarily about vibrations in the hand. A surgeon working through long instruments learns to act at the instrument tips. A tennis player feels the racket’s reach. Brains incorporate the car, cane, surgical tool and tennis racquet into their planning and action in the world. This notion is emphasized again later in the chapter when discussing how AI-driven things might interface to us—and change us.
Farther afield, animals can have very different world experiences than we do. The richness of what a dog smells is pretty incredible.[6] Traces on the rug might tell a dog who went there, how recently, and what direction the scent leads. All invisible to our relatively weak olfactory system.
Biologists call an animal’s perceptual world its umwelt, German for “surrounding world.”[7] What enters that world depends on the animal’s senses and body. A bat’s umwelt is organized partly around echo. Many birds can orient using the Earth’s magnetic field, a sense humans don’t knowingly possess.[8] Different senses make different features of the world salient. You cannot fully reconstruct a bat’s understanding of a room by imagining your own with the lights off, because the bat is not doing a darker version of what you do. It is doing something else.
All of this means a robot cannot experience the world as a human does, and probably not even as closely to us as many animals. Its sensory apparatus and possibilities for action help determine which distinctions are available and useful. A lidar unit may sweep the surroundings and measure distances without depending on ambient light. A thermal camera may reveal the residual heat in a chair a person just left. Yet robots also lack channels our minds take for granted.
At a minimum, it means robots can never be fully aligned with people in how we experience and therefore how we think, and that difference is likely to be far larger than those between humans and some other animals. Its sensory apparatus and possibilities for action help determine which distinctions are available and useful. Its systems can therefore make different features salient, organize information differently, and act on relationships that barely register for us.
There is no reason what emerges should match our thinking. It might even be some weird mismatch between language befitting a human body and the experiences of a robot body.
It’s unclear how the mismatch will manifest. Maybe a lack of pain sensors might mean it’s more inclined to be cavalier around other people. Or that they’re fine giving instructions about the object centered at 9.2 meters distance and 41 degrees from true north to a person who doesn’t think in lidar. Who knows? The key is it will be different, and it might impact their judgments in subtle ways. Like with dogs, we need the intuition that the world is different for them.
A body can make intelligence more present without making it more human.
[1] L. Shapiro and S. Spaulding, “Embodied Cognition,” Stanford Encyclopedia of Philosophy, June 25, 2021, https://plato.stanford.edu/entries/embodied-cognition/
[2] G. Lakoff and M. Johnson, Metaphors We Live By (Chicago: University of Chicago Press, 1980).
[3] D. Kish, “How I Use Sonar to Navigate the World,” TED Talk, March 2015, https://www.ted.com/talks/daniel_kish_how_i_use_sonar_to_navigate_the_world
[4] L. Thaler, S. R. Arnott, and M. A. Goodale, “Neural Correlates of Natural Human Echolocation in Early and Late Blind Echolocation Experts,” PLOS ONE 6, no. 5 (May 2011): e20162, https://doi.org/10.1371/journal.pone.0020162
[5] L. B. Merabet and A. Pascual-Leone, “Neural Reorganization Following Sensory Loss: The Opportunity of Change,” Nature Reviews Neuroscience 11, no. 1 (January 2010): 44–52, https://doi.org/10.1038/nrn2758
[6] A. Kokocińska-Kusiak et al., “Canine Olfaction: Physiology, Behavior, and Possibilities for Practical Applications,” Animals 11, no. 8 (August 2021): 2463, https://doi.org/10.3390/ani11082463
[7] J. von Uexküll, A Foray into the Worlds of Animals and Humans: With a Theory of Meaning, trans. J. D. O’Neil (Minneapolis: University of Minnesota Press, 2010).
[8] H. Mouritsen, “Long-Distance Navigation and Magnetoreception in Migratory Animals,” Nature 558 (2018): 50–59, https://doi.org/10.1038/s41586-018-0176-1
©2026 Dasey Consulting LLC
This article was written with the assistance of AI. The typical way in which I apply AI to the process is detailed in a prior article.