Gregory Laughlin in Nautilus:
The size of things in our universe runs all the way from the tiny 10-19 meter scale that characterizes quark interactions, to the cosmic horizon 1026 meters away. In these 45 possible orders of magnitude, life, as far as we know it, is confined to a relatively tiny bracket of just over nine orders of magnitude, roughly in the middle of the universal range: Bacteria and viruses can measure less than a micron, or 10-6 meters, and the height of the largest trees reaches roughly 100 meters. The honey fungus that lives under the Blue Mountains in Oregon, and is arguably a single living organism, is about 4 kilometers across. When it comes to known sentient life, the range in scale is even smaller, at about three orders of magnitude.
Could things be any different?
Progress in the theory of computation suggests that sentience and intelligence likely require quadrillions of primitive “circuit” elements. Given that our brains are composed of neurons, which are themselves, in essence, specialized cooperative single-cell organisms, we can conclude that biological computers need to be about the physical size of our own brains in order to exhibit the capabilities that we have.
We can imagine building neurons that are smaller than our own, in artificially intelligent systems. Electronic circuit elements, for example, are now substantially smaller than neurons. But they are also simpler in their behavior, and require a superstructure of support (energy, cooling, intercommunication) that takes up a substantial volume. It’s likely that the first true artificial intelligences will occupy volumes that are not so different from the size of our own bodies, despite being based on fundamentally different materials and architectures, again suggesting that there is something special about the meter scale.
More here.