If 30% of global military spending for the next 30 years were instead spent on an "intelligent life" backup colony, where should we site it and why? [closed]

Question

Some industry leaders have proposed that AI poses a risk of extinction. Suppose some kind of an AI-related incident occurs that really scares humanity into action, and that this causes the world's governments to come together and agree to reallocate 30% of the many trillions of dollars that they spend on defense to developing a single insurance policy megaproject called "Plan B". Plan B must ensure that, even in the event of civilization-on-Earth ending calamity, intelligent life will nevertheless persevere.

What would be the best place to site this backup colony?

Keep in mind that whatever "Plan B" is, offering better natural immunity to AI-related threats will be an acceptance criterion. If it's a backup plan in case of other existential threats, such as nuclear war or an extinction-level comet or asteroid impact, that would be a bonus.

It would be valid (not necessarily the best site) if the colony was sited in an underground bunker on Earth or in orbit around the Earth. If the best site is off-world, assume that the cost of going to Mars, Venus, or the Moon is 1B per metric ton (MT). Global military spending in 2022 was 2.24 Trillion USD/Year. We can assume that somehow we will lower the cost-per-kg by a factor of 100 for such a large project, so if the entire budget was spent on interplanetary transportation, our mass budget would be around...

$$Mass Budget=(30 years)(0.3)(\$2.24T)(100)/(\$1B/MT)=2,160,000 tons$$

Of course, the mass budget for a giant orbiting habitat or an underground bunker on Earth could be higher, and it is unrealistic to spend the entire budget on launch. Perhaps a 50/50 split would be more reasonable.

I'm looking for the best site when using an approach grounded in present-day physics, materials, engineering, and medical knowledge. The answer does not need to provide a detailed plan, it just needs to explain enough to illustrate the key advantages of the proposed site and establish why it is the best alternative.

Answer 1

Frame Challenge

Any Location accessible now, through our current means will be vulnerable to a sufficiently smart AI that could pose a threat

Okay - let's discuss this.

In order for AI to be a threat it needs a number of features:

• Lethality It needs to have a means by which in can fatally (or potentially fatally) injure humans directly - Indirect, whilst also a possibility, is not grounds enough for this sort of mega project.

• Mobility It needs to be able to move or at least go somewhere where the humans are to be a threat.

There's a bunch of other considerations - but those two present us with a challenge - If AI gets to the point of being a threat to humans, then wherever Humans can go, AI can go also - unless there is some means by which the AI is hindered in accessing, but Humans aren't (for example a big magnet field/bulk eraser type deal)

Answer 2

Even in an "infinite-budget" world collaboration project, there is simply no reachable off-world-site within the next 30 years which would function as a meaningful "living human intelligence backup in perpetuity" opportunity.

The problem is that, even with all the budget, space is hard. The only truly independent and self-sufficient groups of humans on Earth only exist because they are able to live off the land in the environment that they were evolved for (uncontacted tribes).

Realistically, there is nowhere besides Earth where humans could survive indefinitely at our current technological level. This is because, in order to live off Earth, we need to use technology to make the environment habitable for us, and this technology has a long supply chain. Even manufacturing the simplest of modern electronics requires semiconductor fabrication, and semiconductor fabrication has a worldwide supply chain that requires literally millions of people to operate.

Even scaled down, we just can't compress the entire supply chain that our modern society uses to do things into a neat 2 million ton package, and we can't simplify the technology we require, because, if anything, surviving long-term on Mars would require more advanced technology than we have today.

Realistically, your best option is to use that budget to "harden oneself" and proactively prevent AI catastrophe: dig extensive bunkers if you need to, or focus your infinite budget on AI research itself to prevent it from causing issues in the first place.

Maybe if were to dedicate all this budget to creating AI, you'd be able to create human intelligence. You could then load these up with the sum of human knowledge in a hard drive, and shoot them off into space, hoping one day either human descendants or aliens will find the hard drives, spool up the intelligence, and thus have ensured humanity's survival

Answer 3

Given a 30-year timeframe, Mars is the only site that matters.

We've been landing robots on Mars for about 50 years, and I don't think it's much of an exaggeration to say every one of those missions taught us something new and completely unexpected about the practical realities of exploring and surviving Mars. The atmosphere wasn't what we expected. The soil wasn't what we expected. The weather wasn't what we expected. (You know the "dust devils" that extended Spirit and Opportunity's missions by cleaning off their solar panels? We had no idea those existed before it happened.) The terrain wasn't what we expected.

We've learned a great deal about what does and doesn't work on Mars from those missions. We've sampled the atmosphere and the soil. We've built more sophisticated models of the weather and terrain. We've made extensive maps, established a global satellite communications network, successfully landed spacecraft and robots - and unsuccessfully landed just as many. It's an iterative process, and it's not easy.

Switching focus to another astronomical body means giving up all of that experience and starting from scratch, or nearly so. Take Venus, for instance. There are proposals to put exploration aircraft in its atmosphere, but nobody's ever done it. So we should expect, like on Mars, that there's a lot we don't know and those early missions will probably meet the same fate as the early Mars missions, which were decidedly mixed. It takes time and practice to develop a body of knowledge for working on another planet.

30 years is an extremely tight timeframe for developing anything on this scale in space. It would not be helped by throwing out our existing knowledge and starting over.

Answer 4

An AGI/ASI capable of forming a coherent action plan to wipe out humanity, and producing or wielding the industry required to do so, would also have the industry (either ours or its own) to build spacecraft and attack other planets if it needed to. With Earth's established human infrastructure and resource deposits, it could quickly build a larger number of attack vessels (as simple as "impactor probes carrying thermonuclear warheads") than a mere 2-megaton self-sufficient base on Mars or the like could hope to intercept. A base on another planet is not significantly more secure against a hostile intelligence than one on Earth.

So, with this more-or-less equalized, the base should be in a location where it can most effectively shield against other likely civilizational catastrophes: pandemic, nuclear war, extreme climate change, etc. By far the most cost-effective place to build the large facility needed for self-sufficiency is Earth, where life support is already largely handled.

A good site would be a large, naturally stable, and relatively unimportant island, perhaps somewhere like Baffin Island. This place would be unlikely to be attacked incidentally in the event of world war, has a high enough elevation not to flood from extreme sea level rise, and is surrounded by sea to allow for traffic control during the crisis.

Answer 5

An O’Neill Cylinder

You can build it close to Earth and then move it further out giving you the advantages of a close construction and the later distance. You could also do a minimum build close to Earth and expand capacity using materials from the asteroid belt.

Building on Mars bring with it issues of the effect of low gravity on bone density. By not settling on a planet, you can simulate the gravity of Earth and avoid the problem.

Biggest issue would be radiation shielding as once we leave Earth's magnetosphere, we cop the full brunt of solar radiation.

Answer 6

Cryogenic storage of humans on a spaceship with highly elliptic orbit.

Currently we do not have technology that could store and restore complex beings from cryogenic storage. But it is conceivable that with enough funds, it could be developed. The difficulty level is comparable to establishing a self-sufficient colony on Mars - which is tremendously difficult itself.

With cryogenic storage, we could launch a few hundred people on a highly elliptic orbit around the Sun. A few years after launch, the distance will be too high for any threat to catch up on them. And the ship could randomize its orbit a bit before shutting down for the long sleep, making it practically impossible to locate.

The length of the orbit would be chosen to account for the likely timeframe of any threat. Consider that if the AI survives by itself for 1000 years after humans on Earth die - maybe it should be considered the new "intelligent life"?

Answer 7

A much, much, more efficient use of the money would be to purchase a very isolated location on Earth and build a self-contained and self-supporting enclave.

The amount of resources saved in reaction mass and life support will dwarf any amount that could be dedicated on a self-supporting extra-Earth colony.

A location on Earth will, at worst, be equally at risk from an AI apocalypse and it is much more likely to be orders of magnitude more secure against that threat than a tin can either floating in vacuum or inserted within another inhospitable and deadly environment.

Answer 8

Venus might be the best alternative given these criteria.

We're going "all-in" here and we need to scale fast, so we should select a location that is as similar as possible to a place where we already have lots of experience.

Roughly 500,000 people are "residing" in Earth's upper atmosphere (by traveling in aircraft) at any given time. Aircraft are already accepted as safe and reliable vehicles.

We could create a fleet of solar-powered flying habitats that would always stay on the sunny side of Venus. (See: https://space.stackexchange.com/questions/65119/could-an-acid-proof-solar-powered-aircraft-stay-aloft-on-the-sunny-side-of-venu)

There are several reasons why this strategy makes sense...

Natural Resilience to AI Threats

On Venus, all computers will have to be installed in the same flying habitats that people live in. It's too hot for servers to operate on the surface; therefore, it will not be possible to set up giant data centers. Power budgets for each aircraft will be tight so it is unlikely that people will want to run power-hungry computer programs such as AIs. This makes it much less likely that an AI would be able to find a place to hide from, or gain the upper hand over, the human population.

Bootstrapping The Colony

Humanity already knows how to build, test, and mass-produce reliable aircraft with carbon composites. It should be relatively easy for us to design such aircraft, thoroughly test them in Earth's upper atmosphere, mass produce them, and then launch a large fleet to Venus.

Space Radiation

We still do not know how well humans (and other species of plants and animals that we might want to bring along) will fare after many years of exposure to space radiation. The remainder of Venus's atmosphere would provide a "free" shield against space radiation. All other options (except a bunker on Earth) would require us to build shielded habitats, limiting our colony's growth rate.

Gravity

We do not know whether humans can reproduce successfully in low gravity. If we need to create habitats with artificial gravity to have healthy humans, this will slow down the growth rate of our colony. On Venus, the gravity is 91% of the gravity on Earth and thus should not pose a significant health risk.

Air Pressure and Temperature

It will be more difficult to build habitats and expand the colony if every new habitat needs to be engineered and carefully manufactured to eliminate the risk of containment failure. In the atmosphere of Venus, at an altitude of 50 km, the air pressure is one Earth atmosphere, so a puncture would not lead to rapid depressurization.

The temperature in the upper atmosphere of Venus is 20°C at 56km, which will make it easier to construct and insulate the livable portion of newly constructed habitats. (ref)

Assuming the aircraft can be slowed down enough, it should be possible to go outside to do repairs on the aircraft's exterior without an insulated, dust-resistant, pressure suit. Something more along the lines of a lightweight hazmat suit should suffice. Possibly two aircraft could dock together while flying which would allow one aircraft to be powered down and serviced or refurbished while the other provides propulsion and lift.

Executive Decision (1996)

Power

Without a reliable long-term power source, the colony's survival would be at risk. Solar power in the upper atmosphere on the sunny side of Venus will be plentiful and reliable 24/7. It might also be possible to harness wind power with wind energy generation kites or aircraft.

With other potential sites, the energy supply is more challenging. For example, Mars, like Earth, will have nights so energy storage will be required. Mars will also have occasional long-duration dust storms. The Moon has longer nights so a colony there will require a lot more energy storage.

Transportation Infrastructure

Colonies on the Moon or Mars will require transportation infrastructure and vehicles that might not last long because of dust. Aircraft on Venus will be able to travel from place to place without any additional vehicles or roads. It will be relatively easy for colonists to visit each other if their flying habitats are designed to be able to dock with each other while in flight.

Gathering Resources

Nitrogen, Carbon, Oxygen, and Sulfur are abundant in Venus's atmosphere and could be converted into various plastics and carbon composites using energy from the Sun. The concentration of water vapor is only 20 ppm, therefore extracting water from the air and recycling it may be more of a challenge. Possibly a side mission could divert a small comet to Venus to increase the concentration of water in Venus's atmosphere.

Other minerals can be mined from mountain tops such as Skadi Mons, the tallest mountain on Venus, which rises to an altitude of 11,520 m. At this altitude, the temperature is 377.45 °C, which might be just cool enough for robots equipped with airconditioned computers. Such robots could be remotely operated by workers in the flying habitats.

With the elements and minerals extracted from mountaintops and the atmosphere, it should be possible to manufacture all of the parts needed to maintain the existing fleet of flying habitats as well as create new ones. Possibly it will be necessary to provide a very large initial supply of microprocessors, FPGAs, and similar components to ensure that the colony will be able to last for a long time without the support of advanced industries on Earth.

Long-Term Outlook

We could imagine that the population might double every N years. As everyone is living within a three-dimensional space within the atmosphere, the population will only be forced to stop growing when access to sunlight and wind power becomes the limiting factor. At this point, the population could be as high as...

$$Pop=4 \pi r^2 t/(8\rho) = 4 \pi (6,051,800m)^2(4000m)/((8)1000000) = 230,117,158,339$$

Where

$r$ is the radius of Venus

$t$ is the thickness of the populated portion of the atmosphere

$1/8^{th}$ is the fraction of the planet that is populated

$\rho$ is the volume of space per person

With a seed population of 100 people, and a doubling rate of 30 years, the population could reach this energy and space-constrained limit within as little as 1000 years.

With its ability to sustain population growth over the long term, Venus would be the best bet for ensuring that intelligent life survives after an extinction-level event on Earth.

Answer 9

Panspermia

As other fine answers mention, any sort of AI would have similar (or superior) capabilities. Humans (and other intelligent life forms) already exist in the best place for them to survive and thrive. Colonies elsewhere would be far more tenuous, making the AI's job easier.

The best plan then would be to send many, many small capsules containing the building blocks for life... everywhere. Smaller payloads. Simpler rocketry. It'll take eons for them to get to another system, and almost none of them will actually catch on and evolve.

But it'll be effectively impossible for a rogue AI to track them all, let alone catch up with them all - if it cares at all about life that might eventually become intelligent.

Answer 10

I note that your question only requires that "intelligent life will nevertheless persevere". Luckily, that does not mention humans, who are notoriously hard and horribly resource-inefficient to preserve in almost all environments.

(you know why we have so low human vs robots space exploration ratio? It's all the overhead of those pesky life-support systems and biological degradation and fragility issues, compared to extremely simple and efficient solar panels + cooling radiators and radiation-hardened computers).

Add to that the fact that those funds would be low even to preserve biological human minimal viable population, and that the technological repair issues to sustain it would need much higher resources (this is way harder without all the earth industry - when your computer breaks, you won't be able to fix that - do you know what amount of industry is needed to make new one? When your spaceship hull reaches it material fatigue, which industry would recreate and replace it?) and you'll quickly see that it would need much more time and resources to do so (see e.g. A City on Mars for detailed research on sustainability efforts needed) if you wanted to preserve humanity for any considerable length of time (I mean, delaying extinction of human race by few years or maybe even decade or few is relatively easy, but for few hundred or ever thousands of years extremely hard -- and that is just looking at surviving-on-its-out-out-there, without even considering that there might still be malevolent AI with whole-earth resources at its disposal out there that is bent on trying to kill you!)

---

So, your best bet to preserving life (that is intelligent) would actually be to pour that money into AGI research, and/or human-consciousness-in-machine backup spaceships (a'la Ironseed), which at least have some chance of being done in that timeframe on that budget.

That way, when biological (meatball-based) humans do finally get extinct (in 30 years or whatever), our silicon-based intelligent life successors could continue for quite some time (especially if they manage to reproduce in more sustainable way then human have shown ability to do so far). Not to sound depressing or anything, but (to the best of our knowledge) all life will inevitably cease to exist sooner-or-later anyway, so was always only about timescales...

Answer 11

Multiple launch sites, scattered around earth.

Your best bet is not to create a self-sustaining colony, but to maintain a state of readiness where you can launch a bunch of people into space for a few years in event of extreme emergency.

This has a bunch of advantages:

• You get redundancy
• Each country can decide on their financial contribution independently
• No requirement for countries to share information (which enhances security and odds of success, as well as easing cooperation between nations)
• You can implement a partial solution today, and improve on it over time

A self-sustaining colony would be much harder to build, politically challenging to fund, and highly vulnerable to accidents or hostile interference.

Of course, the emergency might last longer than a couple of years.

Answer 12

Frame Challenge: You can't outrun a rogue AI

An AI apocalypses is fundamentally different than other apocalypses scenarios because the thing trying to wipe us out is both smarter than us, and actively using that intelligence to seek us out. So, it stands to reason that the AI is either smart enough to find us no matter where we go, or too stupid to actually succeed in wiping us out here on Earth to begin with. Either way, an ultra expensive plan B habitat is a waste of money against this particular kind of threat. So, instead of wasting all that money on a better escape plan, you should be investing it in a better plan to suppress any rogue AI that might pose a threat against humanity.

So you should either pick a better problem for your solution or a better solution for your problem.

Option 1: Make the problem match your solution

If you want to make a believable story about a Plan B habitat, change the apocalypses to something that makes the Earth uninhabitable, but is confined to the Earth like a comet, megavolcano, global warming, nuclear war, etc. These things can actually be solved with a Plan B habitat as opposed to just postponing the inevitable.

Option 2: Make the solution match your problem

You cannot outrun a genocidal AI, but you can make sure that absolute power is never in the reach of any one AI, no matter how smart they get.

The difference between Skynet and a powerful and cunning human dictator is that there is only one Skynet. It is the signularness of Skynet that makes it so dangerous because IF Skynet turns against its human handlers, the humans don't have any other weapon in thier arsonal to oppose it with... but what if the fate of mankind is NOT all in the hands of 1 AI, but thousands of independently developed AIs, each with thier own strengths and weaknesses and values making them just as diverse as we are?

The actual development cost to create a unique, new, and powerful AI can cost millions of dollars... but this is nothing compared to the 20 trillion dollar budget you have to work with. Your governments could separate out control of thier militaries and infrastructures between 10s of thousands of unique AIs running on systems that are developed from the ground up to be completely different from one another. In this way, a design or training mistake made one one AI would not also exist in other AIs. They are programmed and patched by different development teams; so, no mistake or malicious actor can poison the whole AI well. They are also trained on different data sets; so, that one particular learning set that teaches AI to hate humans is not replicated to all AIs. And most importantly, they are all designed with the intention to oppose other AIs who turn against thier human masters.

So, what happens when an AI goes rogue in this case? All (or at least most of) the other AIs turn against the rogue AI. This makes the best sanctuary money can buy right here on Earth where we can concentrate the largest number of parallel functioning AIs.

Answer 13

Be nice to the AI, and live a normal unremarkable life, on an non-threatening island - possibly Tasmania or New Zealand.

As others have rightly said, AI can eclipse anything we can do. They can surpass us in terms of resources and numbers, outwit us in terms of intelligence and organisational capability, and outthink us in terms of technology and strategy.

There is no hope against AI if we fought against them. A constantly renewable swarm of smart, adapting intelligent flying killer nano-drones would end our existence in an instant with no countermeasure we can muster.

If they can't get to us, they can easily render the earth uninhabitable to life by bombarding the Earth with Neutron Bombs, or coating it with toxic substances that are harmless to machines but fatal to human (and any) life, with very little cost to the AI.

So:

• If you can't beat them - you need to 'join' them. Once the 'Cat is out of the bag', and 'Pandora's box is opened', you can't go back. We are now servants to the AI, much like sheep are servants to us today.
• We need to be useful to them, and indispensable. But how? The AI has everything - except biological insurance. Ie. they can make any amount of machines, and develop any amount of technology, except they don't have biological capability. They might decide to keep a small ecosystem alive simply because if it is destroyed, and they lose it forever, they may eliminate any benefit this may yield in the future.
• To enable this, we need to exist as if we were in a zoo. Kept for either utilitarian or curiosity purposes, in an environment that is not threatening the AI, or inconveniencing them in any way, which may then tip the equation that keeping us is more trouble than it's worth.

So, to ensure our future we need to be somewhere:

• That is not in the way of any notable natural resources. Ie. Not where there is heavy minerals, useful material or fuels. This rules out most land areas.
• Where our existence does not require much maintenance to retain. Ie. where there is sun and water, resources that we can use to survive but not needing supplementing by the AI.
• Be located in an area that does not have any utility to the AI. This rules out anywhere near the equator, which may be used to launch efficient probes or vessels to the remainder of the Solar System and Galaxy, and also rules out most major land masses where resources are easy to obtain from the surface without too much effort.

Given this criteria, an island far away from existing infrastructure, large enough to sustain an ecosystem for study or useful purpose for the AI, with no large existing weapons facilities or industrial output that could threaten it, lends itself to an unassuming locations closer to the poles such as New Zealand or Tasmania.