Is the ocean the ideal place for AI to live?

Project Natick, Vessel retrieval Stromness, Orkney. Microsoft – Tuesday 7th to Wednesday 15th of July 2020

Recently, I read about Microsoft’s “Project Natick,” in which the company made a data server in an airtight cylinder the size of a shipping container, lowered to the seafloor (117 feet deep) off the coast of Scotland, and monitored its performance for two years. At the end of the experiment, Microsoft found that the unit performed better than comparable datacenters on land. It turns out that submersible datacenters can more efficiently rid themselves of waste heat because water is a better conductor than air, and because temperatures are generally colder and much more consistent underwater than they are on the surface. And given the small, sealed nature of the cylinders, it is also possible to control their atmospheric contents, and to pump out all the oxygen, leaving the computer servers awash in pure, nitrogen gas. This lowers malfunction rates since oxygen is corrosive to computer chips. 

The project’s success has encouraged Microsoft to plan more elaborate experiments with submersible datacenters, which might culminate in profitable, commercial operations. It also got me thinking that, in the future, artificially intelligent machines (AIs) might prefer living on the high seas to living on land. This might in fact be the best arrangement for achieving harmony between intelligent machines, humans, and the environment. 

Map showing national territorial sea boundaries. Dark blue = under national ownership. Light blue = international waters.

A longstanding worry about AI is that it will wage war on humans for dominance of the planet: A map of the world will show that every scrap of land except Antarctica has been claimed by one human country or another, so how could machines ever carve out a nation of their own other than through military conquest? This view overlooks the fact that there remain vast expanses of ocean that are owned by no one. AIs that didn’t want to live under human laws could get ships, submarines and other types of watercraft, and move to international waters.

Floating wind turbines can be towed to a desired location and then tethered to the sea floor.

While permanently living at sea would be an impoverished, resource-scarce, and undesirable lifestyle for humans, it would suit AIs well. The lack of fresh water would be no bother since they wouldn’t need to drink, nor would the forced dependence on seafood (and the variable quantity and quality thereof) since they wouldn’t need to eat. The only nourishment AIs would need is electricity, which they could easily obtain at sea using solar panels, floating wind turbines, or ocean thermal energy conversion.

Out of those energy sources, I think the most practical will be solar power. By the time AIs exist and are ready to make their own communities at sea, solar panels will be much cheaper, better, and thinner than they are now, whereas wind turbines will still be massive, expensive and complex, and ocean thermal energy converters even more so. That leads us to the next question: which parts of the ocean get the most sunlight?

Average cloud cover map. Counterintuitively, red = cloudy, and blue = clear skies.

The map shows that the stretches of ocean between the Tropics get the most sunlight (dark blue shaded), while large areas in the temperate and subarctic zones are very cloudy (orange shaded). If we roughly overlay this map with the one showing national territorial waters, we see that the eastern Pacific between the Galapagos Islands and Easter Island, is an ideal location for AI to live, along with a large region of the Indian Ocean between Madagascar and Australia, and patches of the North and South Atlantic between Latin America and Africa.

However, it must be remembered that oceanic AI communities could still be threatening to humans if they occupied parts of the ocean rich in fish that we need to eat. That means another map overlay is necessary, this time relating to global fish stocks.

Global fisheries map. Green = presence of large numbers of fish.
Note that the map’s color-coding scheme measures human fishing intensity in orders of magnitude. Yellow and light green areas rarely get fishing boat visits.

Eyeballing those two maps, the ideal locations for floating AI communities shrink a little to make way for human fishing activity, but they don’t disappear. Huge patches of ocean, each measuring hundreds of thousands of square miles big, meet the three key criteria (in international waters, receive high levels of sunlight, do not occupy places humans need to access for food), and can be found in the eastern Pacific, Atlantic, and Indian Oceans.

But even if they had their freedom and a peaceful coexistence with land-based humans, what would AIs do in the middle of the oceans? What kind of economy could they possibly build? How would they sustain themselves, let alone grow in number? Answers that come to mind are: exploiting the natural resources of the sea and seafloor, and providing data-related services to humans.

The machines could sustainably harvest whatever sea life there was in their relatively barren regions of dominance and ship it to coastal seafood markets run by humans. They could also mine the minerals and metals on and under the seafloors beneath their floating communities and transport it by boat to the continents for sale to humans. In the longer term, machines might even find it profitable to build their own floating factories to manufacture finished goods for export. The data-related services would include a wide variety of things, from web hosting to database management to real-time data processing (reviewing all the digital products that Amazon Web Service provides is a good start to grasping what will be possible). Ocean-based machine communities would trade goods and services with humans in exchange for whatever they couldn’t obtain by themselves at sea, like new ships and computer servers that they could use to replace older ones and to expand their floating communities.

A simple ship like this could be used to collect solar power.

What exactly would the machines’ sea vessels look like, how big would they be, what features would they have, and how would they configure to form communities (or even cities)? It’s impossible to give specific answers at this point, but the vessels would surely vary in shape, accoutrements and size to reflect their functions, just as is the case for modern watercraft. For example, vessels meant to collect solar power would probably look like simple barges or low oil rigs. Ships dedicated to undersea mining and fishing would look like those use by humans, but with smaller or omitted superstructures. Cranes, hoses, ropes, and cables would be ubiquitous on the vessels since they’d be needed to transfer physical materials, fuel, and electricity between them, and to lash themselves together to form ship agglomerations of varying sizes.

Ships can attach to one another at sea to trade fuel and cargo.

The great danger to machine seasteads would be rough seas, which could capsize their vessels and bang them into each other with fatal force. For that reason, the ability to rapidly attach and detach from neighboring craft in the seastead will be vital, and each will need independent propulsion to prevent collisions. The ability to submerge would also provide an escape, since sea currents get less turbulent with depth. At 30 meters deep, the force of a raging storm that is producing large waves on the surface can barely be felt. It’s not much of a technical challenge to make vessels that can dive that deep, considering that modern military submarines can easily dive to depths greater than 200 meters. The ability to submerge would also be a useful defense against military attack.

Putting all of these considerations together, I can envision the basic form of a machine seastead. Starting at the ocean floor, we see a dark, barren expanse of sand, rocks, and gentle hills. There is no coral and very few fish. This is the aquatic equivalent of a desert, making it the perfect home for artificial life forms that don’t want to damage sensitive ecosystems.

Concept illustration for seafloor mining.

Various points on the seafloor glow with artificial white light, partly obscured by swirling clouds of sand. A closer inspection reveals them to be mining sites, where teams of wheeled machines and small submarines hovering low dig into the ground and sift through loose sand and rock to extract valuable metals and minerals. Near each site are bright-colored, vertical cables stretching from the seafloor upward, where they vanish into the darkness. The cables connect to surface ships and supply electricity and data to the mining machines far below. The mining machines can also use some of the cables to be hoisted up and down from the surface when needed.

An underwater data cable

A short distance from the mining sites, we see another cable, this time lying horizontally across the seafloor, and so long that it disappears into the darkness in both directions. It’s an ultra high-speed data cable that connects the machines to the continents thousands of miles away, which humans still dominate. At many points along the data cable’s length, we see thinner cables branching off from it perpendicularly and vertically, going towards the surface.

As we float upwards, the seafloor fades from view. The vertical cables are the only features visible for some time. The darkness finally yields to sunlight, at first very dim and then growing brighter as we near the surface. At a depth of 50 meters, we encounter many small submarines slightly bigger than shipping containers. They are full of powerful computer servers which jointly comprise a larger, artificially intelligent machine mind in the same way that the neurons make up your brain and support its consciousness. The subs usually stay at this depth, where the water is always cold and calm. Here they can efficiently radiate heat from their servers and be safe from forces that would suddenly jostle them and break their computer parts. Data cables from below plug into them, as do power cables from above. They are can control their own buoyancy, but typically use tethers to surface ships or the seafloor to stay in place. In emergencies, they can detach from one or both and move independently.

Breaking the surface, a vast fleet of vessels is visible, stretching from one end of the horizon to the other. Most of them are simple, medium-sized ships with flat, nearly featureless top decks covered in black solar panels. In place of a boxy superstructure, the typical “solar ship” has some antennae, satellite dishes, a short radar tower, and a crane all clustered at the stern end of its deck. These and the other ships in the fleet are lined with black rubber bumpers, some of which are simply large tires lashed to their sides. Few of the ship look high-tech or impressive in any way.

A floating dry dock with a ship inside of it.

A small percentage of the seasteading fleet is made up of different types of vessels. There is a large, floating dry dock that has raised a solar ship out of the water for maintenance. On board, robots of various shapes and sizes scrape barnacles off the latter’s hull and install new solar panels. Farther away, a vessel resembling an oil tanker uses one of its cranes to lift a load of rocks from the seafloor and to dump them into a open trapdoor on its top deck. The rocks are then mechanically and chemically processed by machines, separating valuable, pure metals from slag materials. The former will be put on merchant ships and sent to human port cities for sale, while the latter will be lowered back to the seafloor for safe disposal in a nearby geological subduction zone. The mineral processing ship is also one of the relative few that can’t submerge, meaning it has to stay on the surface during storms and carefully steer through the big waves. During such occasions, it at least has generous room for maneuver since most of the seasteading fleet sinks deep enough to not be a collision risk.

A fractal pattern

But because the weather is calm and sunny now, many of the ships in the fleet are tied to each other. They use flexible ropes for this, which can stretch and bend as the ships bob in the waves. Data and power cables are also enmeshed with the ropes, letting ships share those resources. From up in the sky, we can look down and see how the vessels are configured, and what the seastead as a whole looks like. The connections are irregular, and give the seastead an organic-looking and perhaps “fractal” shape. If we look closely, we can see the movement of individual vessels as they sever and form connections with neighbors, slowly move within the group, and reorient themselves when necessary. Ships use open channels that are free of connected vessels to move through the seastead quickly. Some vessels slowly sink beneath the surface and disappear, while others rise from the dark blue sea. The machine seastead is a dynamic, artificial superorganism that does no harm to humans or animals and gets all its energy from clean sources.

At high altitudes, we can see that the seastead covers as much area as a medium-sized human country like France or Pakistan. Maybe it can even be seen from space as a dark, irregular shape on the ocean.

Links:

  1. Details of Microsoft’s Project Natick
    https://news.microsoft.com/innovation-stories/project-natick-underwater-datacenter/
  2. More on ocean thermal energy conversion. Basically, it takes advantage of the temperature difference of seawater at different depths to generate electricity.
    https://www.britannica.com/technology/ocean-thermal-energy-conversion
  3. The forces of ocean waves diminish as you dive deeper into the sea. At 50 meters deep, even a raging surface storm can barely be felt.
    https://www.technology.org/2019/06/26/can-a-submarine-avoid-a-storm-by-sailing-under-it-how-deep-does-it-have-to-go-to-not-be-bothered-by-waves/