In the early 22nd century, sending humans to other planetary systems is still too costly to be realized, so the first interstellar missions will be piloted by AIs. An AI piloted probe will be sent to the Alpha Centauri system, and its mission is to establish a permanent outpost around Alpha Centauri A.

The question: At a minimum mass, what equipments must the probe carry to accomplish its mission? The lower its mass, the better. To answer this question, you don't need to consider the mass of the propulsion system that transports the probe to Alpha Centauri, and you only need to consider the payload.

Mission parameters:

  • Due to the distance, our probe can not receive timely remote control from the Solar system. And you must not use any human (or human brain in a jar) to pilot this mission.
  • The AI that pilots this mission is an AGI that is at least as intelligent as humans in all aspects. It constantly uses the vast majority of its computational power on maintaining human-level self-awareness, but the rest of its computational power still allows it to solve mathematical problems way faster than humans do. Our AI correctly understands its mission and is willing to do it. (I.e. it does not betray its mission.) It also has the engineering abilities and blueprints required to conduct the mission.
  • The hardware our AI runs on is ten thousand times more energetically efficient than today's hardware, so each instance of it runs on only tens of kilowatts of power and 100 cubic meters of hardware (radiation hardening and redundancy included). We further assume that the process of manufacturing this hardware is as complicated as manufacturing CPUs today.
  • The role of our AI is a commander, which means that it controls many non-general lesser AIs. The non-general AIs can operate robots, vehicles and factories, and they use much less computational power.
  • Our probe can bring resources, robots, additional tools, pieces of machinery, spare parts and manufactured computer hardware with it. The probe and its equipment also need a power source to operate. Fission, fusion and solar technologies are some possible choices for a power source, but antimatter is not available.

Requirements for the answer:

  • The outpost our probe establishes must be fully self-sustaining. It must be able to extract and refine resources in the Alpha Centauri system to manufacture computer hardware, robots and new parts and construct a base without receiving any further supply from the Solar system. (Manufacturing computer parts is a must.)

  • In my story, Alpha Centauri A and B have no large planet, so our probe can only exploit asteroids. We further assume that the types of asteroids in the Alpha Centauri system are similar to the types of asteroids in the solar system.

  • There is no nano-assembler, so our probe can only carry conventional (macroscopic) manufacturing tools, like laser machining and photolithography devices. Edit: Additionally, every technology we use in our probe should work based on the physics laws we know today.

  • $\begingroup$ "The [...] in many ways, it does not surpass humans by too much": Except that it can perform computations about a billion times faster, it has perfect memory, it can solve optimization problems without even spinning its fans, it can run simulations of discrete and continuous systems at will, and so on. You know, it is a computer, after all. $\endgroup$
    – AlexP
    Apr 25, 2022 at 20:32
  • 1
    $\begingroup$ @AlexP In my story, I assumed that it takes many exaFLOPS to run an AI that has all human mental functionality, and that they work as black-box neural networks. This computational power can not simply be diverted to solving any math problems without severely slowing down the AI itself. On the other hand, AIs can have coprocessors that can solve problems as easily as you said, but the AI that pilots the interstellar mission does not have too much of such a luxury as its peers, because mass is the most important limitation in this mission. $\endgroup$ Apr 25, 2022 at 20:43
  • 1
    $\begingroup$ @AlexP But I do feel that you have a point. If the AI can use 1 petaFLOPS in its many exaFLOPS on solving math problems, it still surpasses humans by much. $\endgroup$ Apr 25, 2022 at 20:53
  • $\begingroup$ Related min industrial base to colonize mars min tech for probe to colonize $\endgroup$ Apr 25, 2022 at 23:01
  • $\begingroup$ Just going to say 100 cubic meters for a computer is kinda... extreme. Modern mainframes are on the order of 4 cubic meters. Typical humans are ~100Kg or 1/10 of a cubic meter. At the size your talking I start thinking maybe Iain Banks 'Culture Great mind' kind of size. $\endgroup$ Apr 25, 2022 at 23:16

4 Answers 4


A few hundred tons

Your AI is too massive and too useless in the initial colonisation phase to be directly involved. You mention that specialist systems exist. These will suffice.

Another thing to consider is that the bootstrapping process you need will be something that has been perfected in the interplanetary age. Von Neumann replicators, which are the technology you are ready asking about, are immensely useful as the basis of any economy. This is what any group building space stations or colonies wants. Cooperation between extraction operations and high tech imports are nice force multipliers, but could be done away with if more time is invested. So keep in mind that you aren't using some experimental setup, you are simply using a sports car type Von Neumann maschine relative to the usual economy class ones.

I will make a few assumptions about the spacecraft, as this is the best and only source of refined materials the probe has. I'll assume a magnetic sailor, which was pushed up to speed using a stream of relativistic laserpushed lightsails (Kare Sailbeam concept, used a laser for point defense along side a magnetic shield and decelerated using the magnetic sail against the solar wind at first and and then antimatter pulse propulsion. This gives us: structural frames (mostly advanced carbon nanomaterials), radiation and debris shielding (plastics or ice and carbon respectively), a powerful laser system, a telescope system (the point defense setup), kilometers of nanotube re-enforced high temperature superconductors (magnetic sail) and last but not least a handy quantity of antimatter (you could use He3-De (or Tr-De) fusion as well, but this complicates the setup and requires a drive and/or reactor)

With the above in mind we should consider the local resources (which won't be a surprise as analysing the composition of astroids from a few lightyears away is trivial for a space bound telescope (or an Oort Cloud spanning array of them), these telescopes are enabled by the Von Neumann maschines we are discussing). There are solar energy and metallic, rocky, carbon-rich, and volitile-rich asteroids. Furthermore we got the advantage of working in space: no gravity, three dimensional working environments, solar energy, free storage, easy access to very low temperatures, ... Noticeably absent is industry and a good source of fusion fuel.

With all that in mind I'll now consider how a light advanced unit can set up the required infrastructure to slow down the bigger and slower spacecraft carrying the AI that comes in behind it. As you insist on bringing a probably hundreds of tons heavy computer along with you the splitting of the mission in the preparation and the establishment phase is only reasonable. The second vessel, which could just as well be the first in of a large fleet of colonial vessels, can carry a few tons of hardware to make industry components, which will speed up our progress towards an advanced economy considerably. The advanced probe will do all the heavy lifting, so that the AI can arrive, plug in and do the delicate lifting. The mission of the advanced unit is:

  • reconnaissance
  • resource acquisition
  • basic infrastructure establishment
  • slowdown of the bigger second probe

The last part of that mission statement will alongside the magnetic sails cut the fuel budget on boared any of the probes and following ones to only the little bit aboard the advanced probe. Use antimatter for best results.

As the first three goals must be achieved to achieve the last one, we'll consider that last one primarily. I'll walk through the process and show what is needed along the way. The last goal is achieved by building a laser array that can propel mass in the path of the incoming second probe. It will vaporise these masses and use it resulting plasma to slow down alongside it's magnetic sail. Note that this array can be much smaller than the array that is needed in Sol to accelerate the probe. The vessel itself supplies all the kinetic energy, we just need to position the masses. This system is additionally the best propulsion system available to the infrastructure project. It outperforms the chemical and solar-thermal rockets as well as the momentum exchange tethers we'll use alongside it by orders of magnitude, even while it is being build up (a beamed power torch drive system) Additionally it can be used for power generation in solar thermal powerplants. This is the kind of multi purpose system that is very attractive for such a mission.

So, after the slowdown burn completes we want to find an asteroid with a good balance of resources. I suspect that a group of close ones might be the best we can do. Volatiles and carbon are most important ressources in the beginning. A suitable astroid should be found near the systems frost-line. Our goal is to do reconnaissance. While initially the antimatter drive is our best propulsion system, we want many smaller probes that can do things on their own. For this we need AI cores, manipulation probes capable of gathering small quantities of materials, an electrolysis plant for fuel production, a general purpose resources separation centrifuge, a 3D printer which can build useful stuff using the initial feedstocks, probes capable of assembling what the printer makes (initially the same ones that gather resources) and probably a more specialised device for creating carbon based materials (nanotubes are excellent conductors, plastics are very versatile and so on). These inital systems build a fuel production system and light inital scouts (initially impactors so that interesting asteroids can be analysed spectrometrically by the telescope, then explorers, gathers, assemblers, ...) All of this is still powered using the probes remaining antimatter.

As the fuel economy is secured, the initial exploration is done and industrial feedstocks are coming in, it is time to scale the production. The printer and the carbon lab should have as few complex components as possible, as this is the bottleneck for scaling. It might be worth considering that both systems could be used to build better versions of themselves using a few stored components and local resources. However this works in detail, the production devices will scale up their numbers alongside their complexity using as many local resources as possible while minimising the number of imported components. Think RepRap on steroids. At this stage replicators should be sent to other promising asteroids where different resources are available. Especially important in this phase will be power production. Solar-thermal is the obvious candidate. Alternatively some form of void adapted and electricity producing plant could be used. In the first case we need mirrors (plastics), a working fluid, a turbine, a generator and a radiator array (consider liquid droplet radiators if vacuum oils can be produced). The latter options could probably be produced using carbon based materials from the first asteroid alone. The latter option is an utter wildcard, I guess that you need to supply the seed with an incubator and then with the elements required for organic chemistry. For power transmission we'll use the venerable powerline. Carbon nanotubes can be layed freely between facilities in space and double as transport aids.

With production well on its way and power in place, we'll follow my bank's advice and diversify our assets. As power and replicators arrive at the new resources extraction sides, specialised hubs emerge. Propulsion systems for transport between the hubs include chemical rockets like we used initially but also other options. Our plastic mirrors enable solar thermal rockets as well. For high priority shipments we'll use the sailbeam concept. The old point defense laser pushes projectiles (carbon based dielectric sails, maybe diamond) (the forge for this was build by the expanding industry) to several thousand kilometers per second. The transport probe releases a target mass behind itself. The mass is vaporised and the superconductors we have left over or an Orion style pushing plate transfer the momentum to the transport craft. Note that this works well from the manufacturing hub with the laser to outlying hubs, but not in reverse this is meant to speed important things up, like the delivery of critical components. At this point you will start to complain how specialised dumb systems do all of this. Well, they don't. Important decisions are delegated to the incoming AI, which remotely adwises from maybe even a light-year away. In case of uncertainty, the systems work on creating resource stockpiles (of paperclips for example). As the complexity of the construction project scales up, the lag time to the AI decreases.

With the diversified economy we'll now focus on the laser array. Here we use classical lasers or the solar pumped variety. I can't say which option is better. Anyway, they are produced and start deployment of the deceleration masses. Most likely these are dielectric sails with a water payload and potentially some sort of steering thruster pack. After the AI's vessel has been slowed down, new and previously unavailable components enter the industrial bootstrapping process. These components will allow us to go to full mature economy status. This means semiconductors, but also artificial wombs and space habitats...

To summarize, I think the AI is pretty much the largest contributor to the mass budget. Structural materials can serve double duty as refined industry feedstocks, other systems should be repurposed if at all possible. The antimatter powering the initial phase of the operation is a perfect light weight solution. Concerning the initial probes and replicators, redundancy pays here. Not only in case of damage, but also for the development process. One could take more or less components. Their mass is balanced by the time you are willing to invest and the risk you want to take vs the cost of accelerating the whole thing.

Mass budget

Big probe:

  • AI 120t (assuming a server like density of 1.2 t/m3)
  • components 10 to 100t
  • structural 50t (mostly shielding and the magsail)
  • equipment 10t (antennas, point defense, telescopes and thrusters)
  • power 1t (antimatter) 25t (fusion)

191 to 295t

Small probe:

  • structural 20t
  • equipment 5t
  • power 1t (antimatter) 25t (fusion)
  • computers 1t
  • fabrication equipment 20t
  • components 1 to 25t

38 to 91t

Mass total: 229 to 386t

Keep in mind that this is mass budget is pure fiction. Let's say a slight informed and conservative guesstimate. Despite that, I'm confident that I'm only off by an order of magnitude at most. In either direction.


Just to spite all conventional estimates:

A von Neumann automaton with ansible

The probe is a single spore of a von Neumann automaton. Upon arrival it uses the surrounding to build own copies quantum satis. The ansible (FTL communication device, think: quantum entanglement radio) is needed to be a part of the swarm, so that cramming in all the hardware for AI in the probe itself becomes optional.

So, basically, it could be a single nanobot.

  • $\begingroup$ Sorry. I should make it clear that the whole thing should be based on today's scientific understanding. $\endgroup$ Apr 25, 2022 at 23:08
  • $\begingroup$ No problems. But some kind of input from your side on the size of AGI hardware would be nice, since we could also take the approach of @Amadeus and scale up. $\endgroup$ Apr 26, 2022 at 15:37
  • $\begingroup$ I wrote in the mission parameters part that the AI needs 100 cubic meters of hardware to run, not far from a supercomputer today. $\endgroup$ Apr 26, 2022 at 17:54

Well obviously, you are going to have to bootstrap once you get there; you need a manufacturing facility, and all you have is asteroids. At a minimum, you need a lot of steel, and lots of refined chemicals, so you begin with focused solar power as your energy source. You will also need gravity for many things, so some sort of centrifugal facility.

In space, you can focus (even with just super-thin mylar mirrors) enough energy to melt small asteroids; capture the gases and the molten rocks, and separated these into constituent parts with a small centrifuge. That can separate into distinct layers the various compounds and elements; e.g. you can get pure refined iron, or pure silicon, silver, gold, copper, etc. You can store much of these elementals until they are needed, the cold of space will allow you to cool much of it to a solid for use later. The point is to build and form a much larger centrifuge, one (small) component at a time.

Initially, the AGI will be the brains behind all of this, even operating the large centrifuge(s), so it can construct a fabrication plant for chips, wiring, parts, batteries, and other parts for new robots that are independently intelligent.

Once you have more robots, they can build more fabrication plants and build whatever you want, including more robots. Habitats, cities, process asteroids into farmland within hydroponic enclosures and grow the seeds you carried with you, whatever. Even artificial wombs and incubators if you carried along frozen fertilized blastocysts of animal life, for populating a centrifugal gravity habitat.

I can't say what the mass is, but it shouldn't be too crazy. The Summit Supercomputer, the fastest in the world a year ago, is about 10,000 square feet and maybe 8 feet tall, 2.41 million processors. It weighs in at about 340 tons. Suppose your AGI need 1000 times that area, 10M square feet, 8 feet tall, and 2.4 billion processors. That is 80M cubic feet: but that is only 431 feet, say 150 yards per side for a cube. And it would weigh in at 340,000 tons. (And less, really, since it would not require much of a cooling system in space.)

Of course that is just the brains, along with manipulators, centrifuges, solar panels, rare ingredients, etc. So I will just say one million tons, altogether, using existent technology (except the AGI doesn't exist). That would be a plausible boot package. (excluding fuel, as requested).

  • $\begingroup$ Just for me, for a better understanding: 10000 × square(foot) = 929.0304 m²; 8 × foot = 2.4384 m; (10 × 1000000) × square(foot) = 0.9290304 km²; (80 × 1000000) × (foot^3) ≈ 2265347.727 m³; root(2265347.727 × (meters^3), 3) ≈ 131.3343387 m; 150 × yard = 137.16 m. $\endgroup$ Apr 26, 2022 at 15:42
  • $\begingroup$ @OlegLobachev Yeah, that looks right. A metric ton (1000 kg) is 2205 lb, so 1 metric ton = 1.1025 tons. 340,000 tons = 308,390 metric tons. I rounded up to a million tons, = 907,029 metric tons. Or roughly triple just the brain, call it 900,000 metric tons. But I also assume this AGI can maintain itself; even on Summit, which is composed of cards in racks, particular cards go bad fairly regularly. Power supplies go out, fans break, the cooling system fails or something. Cards, memory, cables get replaced. But we manage our own minor injuries all the time, I assume the AGI can do that too. $\endgroup$
    – Amadeus
    Apr 26, 2022 at 16:52

Like the intro of @TheDyingOfLigth answer, especially this part

  • Another thing to consider is that the bootstrapping process you need will be something that has been perfected in the interplanetary age. Von Neumann replicators, which are the technology you are ready asking about, are immensely useful as the basis of any economy. This is what any group building space stations or colonies wants. ... So keep in mind that you aren't using some experimental setup, you are simply using a sports car type Von Neumann maschine relative to the usual economy class ones a typical SEAM setup every colony has, it just that your mass requirements are higher than typicaly(meaning starting from least mass possible).

So yes, it exactly a way to look at the problem. You may even be starting with just regular setup, without any more strict requirements, but one which is typically used to eat out an asteroid.

And here we somewhat get back to that K2 question of yours Lift 10 Billion Tons of Material From a Planet (Every Second) and associated questions/answers, including bootstraping in general (there are some interesting ones Semiconductor foundries are a thing of the past. Rebuild the computer industry if you can and there is more of those bootstraping q's on WB in different settings)

This paper is exactly what you are looking for https://arxiv.org/abs/1612.03238 (Affordable, Rapid Bootstrapping of the Space Industry and Solar System Civilization) as it has exactly the constrains you mention - least amount of mass of that boostraping complex. Unfortunatly it more like concept, which needs more numbers and details and thougths, and it did undervalue the meand of regular production in favor of 3d printing, but it has it both of those considered which makes the work better than many public consideration of replacing everything with 3d printing - great tech, but no need to make it all more complex than it should be, conventional tech works great for conventional stuff. I do recomend to read the work, and maybe even talk with the guy he was available on twitter if I recal correctly. The problem is the same, as constraints are the same. And there is a lot to say on the problem, but it a deept rabbit hole.

100t and mass of your AI is sufficiently reasonable number. 1t may be also a reasonable number, it just harder to validate it, but considering that intro of @TheDyingOfLigth it could be expected. In an extreame - you need a pair of hands, and few tools to make it happen, and some energy source, 10kW is enough, to start the process - so it more like how much a pair of robotic arms mass is plus maybe additional 100kg of equipment and materials to make life easier. Difference with 10kW or 1GW starting conditions may be like a year or two, so it not so essencial.(in a paradigm you accepted, exponential growth of dyson swarm setup)

That potencial of exponential growth is applicable here as well, it just scale is much smaller, but basics are the same - cover your energy needs first with simpliest means posible which grow the fastest, then start then growth of complexity, reshaping available materials for more complex stuff which may be more efficient and opens technology branches, allowing you to clumb to the level of thech you need.

As source of material there is no need to be picky, any sufficiently big rock will do, no need to search gold asteroids or carbon one or those which have enough water, or much iron. Any sufficently big one will do - so you picking 1km diameter one or bigger.

Build your crude energy extractors with traces of iron, you may have in hands or electrolyse from regolith there. Just regular rocks have suprisingly high iron conent - percents - that's enough, so as there Na, K, Ca, Al which can be as good for reflective stuff - so any rock has enough of metals to make colelctors. Many structural parts of every equipment can be that melted regolith - it does not have be that specifically strong in microgravity and basalt and diabase compositions are strong enough for many mechanical purposes. Including heat to mechanical work and to electricity conversion units.

  • so in nonconventional environment, caring about EROEI - conventional materials are not the first goal, and using available materials as is can bring one quite far in that bootsraping process. There may be significant differences depending on average composition, is there water (or somewhat bound hydrogen) or not - all those probably can be handwaved for maturity of that bootstraping technology.

  • I do mean that any rock will do, they are all more or less the same. Also I mean water in abundant quantities is not so important and not required and trace quantities, like on the surface of the moon, it is sufficient. So as other elements even if it traces of them it may be sufficient. But it a big topic, and no need to dive in it too much so say that at least 90% or more of the rocks will do.

Splitting the whole rock, 1km3 one, even if the composition of it isn't best(unlucky one) it should get you to all the modern technologies. So it minimal requirements for a target, and avoid thick water gas shells - they create more problems than solutions(like Ceres one)

Working with plasma and with abundant energy, and we are talking potencially about 400'000 GW here, may be good way to separate all the material in its elements, extracting trace elements etc, which you may need as doping for andvanced CPU's etc. And reaching that does not require current level of computing power, vacuum tubes can do, early transistors can do, early cpu's can do, basic analog electronics can do - so there is an assortment of things to pick from and use as appropriate to comply with available bill of materials, to extend that bill of available materials.

  • just note - insulators can be replaced with SiO2 coating. And making coatings of all kinds is what vacuum good for.

So a workhorse is 1920-1970 tech stack, and when few GW's are reached it may start complication process, for a fraction of that energy, 1-10%, the rest is your workhorse to convert the whole rock into a powerplant, while using 1-10% of energy output for more complex tech preparations, and when done 100% of it for raising tech sofistication to required level.

Then send 100-1'000'000 t payloads to conquer other asteroids you can easily get. Move to a bigger body and start K1,K2 processe and make stations to catch human payloads or to transfer/send resources - energy materials back home. Or whatever.


  • You said: "Difference with 10kW or 1GW starting conditions may be like a year or two, so it not so essential." But sorry, I haven't mentioned that, any bit of time would matter in my setting, because there may be chasing enemies behind. Also, basalt fibre is a very good material too. – @Vegetable New Man

I did mention that as characteristic of the process, not that one has to start with 10kW, it just that one can have very little to begin with. But it need to understand it all is a tradeoff - more energy more mass of the setup. You may have a reactor whatever GW's it is, as part of the ship for free, but to convert those GW's to work and parts and production - it requires corresponding amount of equipment. GW's alone is not enough, I would say for every kW you have it 10-100kg of equipment - it also a range, not a fixed number and there will be some difference in time and such.

So it may be more of a question how big are those interstelar ships you have, and how much they can take, rather than how small a setup can be. More you take, faster things unroll, that's for sure, but it maybe have a slower speed of travel and acceleration and as result bigger jorney time and loss of time here.

It more or less all up to you, and your story - as for any problem there is a solution. With a growth rate like a double in a week - the difference between 10kW setup and 1GW one is about 4 month. And if ship has less mass it maybe can be accelerated faster to a higher velocity, and wih 2% difference in resulting speede near 0.1c you may win a year as an example(for a 4ly jorney). And if you consider that most of the mass (which aspect is skipped, by your request, again, I do not assume anything about your story, it up to you, I just consider a typical scenario without magic, and it is up to you which side you shift the knobs) may be one which is needed to deaccelerate at arriving system, and it can be a huge mass - difference can be(or not) very significant, more significant than 4 month.

So in a sense I meant it may be irrelevant for the process - you win in one place lose in another, or vice versa. But I would say smaller is preferale, but not necessarly(as an example when one can't take advantage of that, guarantee safe jorney for that small setup, safety margins are not satisfactory etc) - smaller stuff can be kicked to a higher velocity and it is the main loss/gain of time for interstellar expansion.

Yes, basalt is good, so as gabbro-diabase - both can be cast as metalls, some random paper, just a title not sure what the content of it is, but this direction Evaluation and modification of the initial composition of gabbro-basalt rocks for mineral-fiber fabrication and stone casting

That casting part is interesting, it can replace pure metalls, especially in microgravity, and do so for all kinds of applications. It may have certain disadvantages, but for a rapid bootstrapping those disadvantages may be irrelevant, but advantages what is top importance.

  • $\begingroup$ You said: "Difference with 10kW or 1GW starting conditions may be like a year or two, so it not so essential." But sorry, I haven't mentioned that, any bit of time would matter in my setting, because there may be chasing enemies behind. Also, basalt fibre is a very good material too. $\endgroup$ May 5, 2022 at 16:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .