How Close Are We To Self-Replicating Robots Conquering Space?

I'm fascinated by the idea of a self-replicating robot probe the the Von noyman probe where a spacecraft goes to a star system finds the local resources builds copies of itself those copies go off to other solar systems and within a couple of million years you have filled the entire galaxy with self-replicating robot probes and you've explored every Nuuk and cranny of your entire galaxy and it makes me wonder where are all the robot probes why don't we see them and you are we alone in the universe is this like technology not possible I would I've been sort of going down a rabbit hole and I've been surprised to find out there are many groups who are working on various ideas of building self-replicating robot probes that you could for example build a factory on the moon that will build a copy of itself and it'll build a copy of itself and then you could build all kinds of stuff from these self-replicating robot factories and the technology is surprisingly close so my guest today is Professor Alex allery he is the Cada research professor at the center for self-replicating research did you know there was a center for self-replicating research at the mechanical and aerospace engineering department at Carlton University and he is working on this idea of building a self-replicating robotic Factory on the moon eventually you could see them on Mars and it's changing the kinds of technologies that you would use to adapt them to this local situation where maybe you can't take an entire chip Fab with you and instead you've got to work with making your own vacuum tubes and 3D printing electrical motors and stuff out of the local lunar regift so it's a fascinating conversation about working with the constraints that you might find yourself with on a place like the moon and yet still being able to build a copy of a factory that could then build a the copy of a factory uh boggles the mind when you think about it all right enjoy this conversation with Alex Aller Alex does the moon have what we need to sustain our presence on the moon uh that's actually a very uh complex question the answer is yes and no yeah yeah my work is actually trying to leverage as much as possible from the Moon um almost like um if you can imagine the native Canadian or you know the idea was to uh live within the environment utilize the environment as much as you could and waste nothing most approaches to doing you know L you know using lunar resources tend to focus on the idea of bringing earth-based technology to the moon and then cobbling some kind of uh you know locally made stuff um whereas my Approach has been basically trying to look trying to push the boundaries as far as we can to see what we can do on the moon and how much we can leverage it and sometimes adapting our technology to suit the moon the resources that are there so I'll give you an example uh an example would be uh if we want to build Electronics on the moon it's not going to be feasible using current current techniques for Solid State Technology but we can build Computing capabilities on the moon if we readdress how readdress how we do it and one potential approach would be to say maybe uh one that I've been looking at is to utilize vacuum tube technology because the basic materials are available on the moon and that means we have to readdress how we architect our Computing uh we don't want to have um CPUs the size of blocker Flats um so we want to try and look at alternative architectures and one of the architectures I've been looking at has been analog neural networks so it's developing a new approach to utilize the resources that are there and adapt to the local environment so it does mean bringing up new creating new technologies new techniques uh for utilizing the resources there are some limitations however and one is that there are some materials which are very scarce on the moon uh things like sodium and chlorine uh which are extremely useful as reagents and in fact one of the things one of the relaxations I've put in the way we approach things is yes okay we're going to have to bring reagents from Earth to the moon but we can recycle them so we don't actually consume them as such so an ask you the question there yes and no aspects to it right well let's imagine like right now today as we've seen a couple of Landers successfully and unsuccessfully reach the surface of the Moon and they are made 100% out of Earth Resources by Engineers purchasing gear and and putting it all together and and machining things and and and so no part of these are made on the moon but if we imagine as we transition to going from everything being made here on Earth and being delivered to the surface of the Moon to some future as you may be imagining it where suddenly or eventually every component of the spacecraft Lander Rover uh research station whatever is built using local lunar resources what does that transition look like do you see what's the lwh hanging fruit to start with and then shift into more more complex parts of the task well it's interesting that because the there have been two main approaches to employing things like self-replication on the moon one which is supported by some some guys at Nasa at Kennedy who've been suggesting that once we reached a certain stage of industrialization on the moon we can then start to retrofit self-replication onto pre pre-existing uh structures and and um capabilities I would argue the other way I would say part of the problem is that uh when you try to create a self-replicating machine you have to ensure that the the self-replication loop is closed what that means is that the uh everything that's in the self-replicator has to be replicated and that's actually very difficult to do and you have to employ a certain amount of Ingenuity to do that so my Approach is is essentially to approach it almost like the way the origin of Life way which is we if we we don't really know how life really originated but we know it must have evolved from the ground up with up you know from the local resources that were available uh so we have to essentially emulate the same thing so we have to work with the local resources that are available to us and utilize only those resources um we also have to ensure that things are simple enough that they can be replicated so it the idea of retrofitting doesn't really work because what you're doing is you're taking complex system and then you're replacing with something simpler which is has reduced functionality so that's that's not really the approach I would I would take basically this is an All or Nothing thing so when we build a self-replicator we have to figure out how to build the self-replicator build one on earth and then send it to the moon and then get it to utilize the resources that are available on the moon but is that even possible here on Earth it it's funny I um I issued a challenge a self-replicating challenge a couple years ago uh to see if anybody could could you know I I said I want to watch a robot make another robot um and and you know right now at our current level of just knowledge industrial capability even if you make the most you know you're here on Earth and you can shovel dirt and material towards the the robot as needed you can plug it in and it can get its electricity you know you can give it all kinds of of jump starts do would we be able to have a robot make a copy of itself here on Earth uh it doesn't make a lot it doesn't make a lot of sense to do it here on Earth uh for two reasons uh the first reason is economic uh it would have to compete with we have a self-re our entire economy and and technology system is self-replicating in in essence in itself so to replace it with a smaller machine with competing with a current capability doesn't make a lot of sense the second reason is that ultimately a self-replicator has to use local resources the Earth is a very complex tectonic environment uh which is basically weathering processes redistributed um resources all around the earth and concentrated them in certain locations so a self-replicator manufactured for for one location can't be translated to a new location because it have to be you know redesigned to to utilize the new location but but I guess asking like if you put it in the ideal situation if you put it in a um Best Buy or you had it in a pile of Lego bricks or you had it in a uh you know junk pile or or whatever is the most ideal situation possible you can possibly imagine for a robot to try and build a copy do you think that we could do that here on Earth yes um essentially if you had uh let's say a um let's take your junk pile let's say so piles of metal of various different types that you need um uh yes you you can and and I I can envisage that that would be possible and we need to develop something like that which is adapted to the geology of the Moon to apply it to the moon and so I've been focused on on the moon which has a limited resource availability in terms of materials compared to Earth but this actually turns out to be an advantage because if you can imagine you you got your mobile phone for example there are about I think I remember reading there are something like about 60 to 80 different materials in this phone each one of them has to be mined and then processed and so there's this huge fan in to making this phone in order to close the self-replication we need to narrow that fan in as much as possible to minimize miniz the amount of processing that has to be wrapped into that self-replication Loop then minimize the amount of mining minimize the amount of materials processing and even Min minimizing the types of manufacturing machines you need and assembly and so on so forth so what this means is that we want to minimize the the number of materials that we use we want to minimize the number of parts that we manufacture uh and those and also we want to minimize the amount of energy we need and minimize the amount minimize the amount of information to specify all this now you talked about energy uh a few moments ago um the way on Visage going to the Moon is we send a seed machine uh maybe about 10 tons and then it's it grows by growing its energy capability it's energy capacity by building more and more uh energy uh solar solar stations and then it grows to its full extent and then it has enough power to to replicate itself and it will replicate a seed another seed and then that will grow and so on so forth so the idea is that we're build it's self-replicating in its entirety but it has a maturation stage you got seed and then it matures and so on so that way you can handle the the power aspect relatively straightforwardly uh in terms of like Earth yes you can we can build something like this um and I see no reason why and in fact my goal is to build a demonstrator on Earth um and then you know basically send it to the moon for um you know and then adapt it yeah I mean it makes a ton of sense that you put it in a sandbox with all of the local resources that it might require and then find out how it goes wrong and then fix that problem either by giving it the better resources or or changing the the plan so I want to talk about like broad Strokes the different components you know we talked about power I'm sure you talked about chips briefly so so what are sort of the major components that you have to think about and I'd be interested to sort of hear what are the lunar analogues to those things so let's start with power for example so how do you make power on the moon okay let me ask your first question first and then lead into that sure so uh what I've uh the key to making it work I remember I mentioned about the types of Parts you need minimizing the types of Parts you need um we want to get a lesson from biology and this is something called exaptation exaptation is where uh feathers first evolved in dinosaurs to for thermal insulation and then they were later adapted in B in the you know offshoot we call birds for flight so it's a case of where um one capability was had been had evolved and then was adapted to a new capability so I've taken very much the same approach and so going to power for example one of the uh one of the constraints I've had is how do we minimize the number of numbers and types of parts that we need so I've looked at various types of Power Systems and what I've tried to do is start off with what do we need to create a robotic machine so the two main well two main types of components we need to create a robotic machine are an electric motor which basically provides actuation and that can be applied to any kind of actuation and we also need something to imp Implement some some kind of electronic amplifier now transistors we can't use but vacuum tubes we can but a vacuum tube itself is a thermionic device so that provides us with the capability to convert heat into electrical energy so if we combine let's say a frel lens which generates heat from solar energy Focus that onto a thermionic converter we have electrical energy so that's the approach I've taken with gener with raising power we utilize the same components that we utilize for other things similarly for example another piece of piece of another piece of the puzzle that we need for robotic machine is sensors and the most important sensory modality is Vision how do we create Vision Vision systems and the approach I've taken is to look at things like um uh photo multiplier tubes which are basically vacuum tubes which are sensitive to light so what we're doing is we're taking a particular type of device particular type of art the vacuum tube and adapting it for computing for energy generation and for vision so it's basically adapting the same type of device uh for multiple applications the same with the motor for example you talk about energy well during the Luna night we're going to have to store energy how do we do that most people think about batteries but one other possibility is flywheels and flywheels are basically motor driven rims so again we're adapting a single type of component for multiple different applications and that to me is the really key to actually closing this self-replication link so so does it like if you just keep coming up with ways to combine components together are you saying you just end up with vacuum tubes and and some kind of electric motor is that uh yes we also need VAR we'll also need various materials so if you look at if you take the breakdown of a spacecraft you need structure you need uh uh thermal insulation you also need uh electrical conductivity so on Earth we use copper but we can replace that with aluminium for example which is actually there's quite a lot of aluminum on so we also look at it from a materials point of view in terms of the M the functionality of of specific materials but also in as components but the two the two things together where we're looking at the individual materials and the specific components we can combine them to to create with just a few other little piece bits and pieces I think to create this full self-replicating system now when I think about vacuum tubes I mean that is however many technology Generations back from what we currently use with with transistors um what is the I guess like how much data do you need to do this work and and how much you know what kind of I don't know processing speed could you get from arrays of of vacuum tubes well remember we using vacuum tubes in a slightly different way to the way you would use them in the CPU so the a neural net an analog Hardware neural net actually has the program actually in its Hardware so it's very fast um the second aspect is that one of the things we're looking at you've presumed familiar with the way AI has been going over the past 20 years they've been using these deep learning neural Nets which are huge and they require enormous amounts of data one of the thing now there's a limit to how far we can go with that technique and I I I actually think and I'm not alone in this that actually this is a dead end and we need to readdress reassess how we approach Ai and one of the the techniques that I'm interested in looking at is taking neuron Nets and building them in Hardware uh perhaps we might use spiking neural network neurons for example to try and uh enrich the Dynamics of neurons um but we want to minimize the amount of information that we need to to to store now there are two ways to do this one is to have dedicated programs which are have their own individual neuron Nets essentially um another technique which we're exploring is to attempt to implement symbol processing in neuron Nets now this is actually a fairly well it's not a new research area it's been around for a while but it went through a big dip for quite a few years and because of the problems encountered in these very very large neural Nets we have now there now now there are now some research groups who are actually interested in seeing how we can compactify uh neural net representations so that's an area of research I think which will which we're forced we're forced to go down this road uh in terms of like a self-replicating machine right if we put I mean if we ran a gp4 powered robot that had the appropriate training and put it on the moon it seems reasonably straightforward that it could be it could reach the point that could it could carry out these kinds of tasks autonomously without without failure and so I think we're at that level technologically with our current and this is only this is brand new I mean we've we're only seeing this kind of dramatic jump in in robots being able to carry out these kinds of tasks put the blue box in the in the red drawer and open up the bag of chips and deliver it to the to the red pony whatever right um you know this stuff is all brand new but if you sort of think about minimizing like once you've decided okay this is the absolute bare minimum functionality that's required we're going to use a neural network we're going to embed it into vacuum tubes and then we're going to just cut away all the fat until it is the absolute bare minimum to bootstrap up this process is there a amount of memory that you sort of Envision that that people could could sort of wrap their minds around yeah okay let let me give you an example um of uh some of the difficulties in in terms of like the amount of information you need to specify uh a self-replicator or any kind of structure or machine if you look at uh a 3D printer uh a 3D printer you got your s like a your CAD model and what the cad model is basically a set of cartisian points and you slice it through the middle and then you feed a sequence of cartisian points to create your layered structure um you know one on top of the other that's a very information dense approach so one of the things we're starting to look at is instead of taking your CAD model um and slicing it into slices um you can do that but let's change the way we specify it so what we start off with is a set of initial coordinates and then we have growing rules and growing rules are actually very simple so so we can grow the structure so we don't have to specify the caresan coordinates at each level so we grow the structure using these rules and then we simply store the rules sounds very familiar to life doesn't it so well yes it's basically a form of growth so it's basically growing uh is essentially how how how how Anything Grows an embryo grows in in essence um and so it's basically forming it's actually we're actually implementing a form of uh information compression uh so that way we can reduce the amount of information need to record to store the specification of the entire system how far we can go down that road I don't know but a lot of lot of the inspiration I get is actually from biology looking at biology how biology does things and then seeing if we can emulate that in some fashion obviously we know biology works because we we exist yeah I mean there are these I know that that in certain genes will define various attributes of an of an animal body in many cases it's just expressed in an extremely compressed thing that you just change one just one gene is defining how long the legs are or or in Cas it's not it's not it's also the control genes as well so uh for example a lot of growth is basically through timing signals so you you grow and then you you grow for a certain time then this Gene switches off and then it stops growing so it's Bas it's B you just need a representation of time uh in order to create the the shape of growth that you need so these are all things that we can explore to try and minimize the amount of information we need to store to actually replicate something replicate a well we want to replicate an entire machine so you know we've spent a lot of time on the vacuum tubes um so let's talk about the motor yeah that sounds challenging okay well that's that we haven't 3D printed a vacuum tube we have 3D printed an electric motor uh so we we used several different techniques we uh printed the magnets we printed all the other parts we assembled it by hand obviously uh the hardest part was the the wire wiring the wire coils but we've demonstrated we can in principle 3D print an entire electric motor which we which we feel was or I feel is is a significant step towards uh demonstrating that 3D printing either by itself or supplemented with a small number of other types of manufacturing machines essentially can construct virtually anything uh and when I say anything what I mean is that any kind of if you construct any type of machine uh all machines are basically different kinematic configurations of electric motors so robot arm is one type you've got a lathe is another one uh a Milling head is is is another type of uh kinematic machine so once we have electric motors we can then create new types of machines uh to do what whatever we whatever we we want to achieve but the the core of it is the 3D printer which is essentially a cartisian robot it it has a moving head which moves up and down so it's a car and robot in essence so the idea is to try and again minimize the types of machines we require and we've demonstrated that we can print electric motors and eventually I'd like to print uh um some kind of acum tube type device um but that will come later right right it's interesting um uh when you think about like this going to sound like a really weird tangent when you think about say topology that that you have things are either spheres or Donuts or donuts with multiple holes and and so when you really take the essence the pure essence of all of these different maches as you describe them in the end it's a motor it's a motor with stuff connected to it that is making the motor work in different angles different directions using belts pulleys um size scale speed all of that and so if you can make one motor then theoretically you can make every flavor of a motor just by adding extra things to the outside of it or changing the size the shape the speed the torque whatever and then on the flip side you've got like and these are like the IR rued ible components that you have to conf consider and then the other one was the you know you need some way to to have information flow around are there any more like just base components that that there's no way to get to get away from do you think yeah I'd say the other one which we haven't we have them in our Labs but in our lab but um we haven't actually tried to manufacture them and that would be forel lenses um they will be I could see them being quite challenging to to manufacture because of the etching on the surface F Fel lens is basically a flat lens where you replace the the thi you know the shape the refracting shape with a series of etches which emulate basically like Focus the beam to as as a lens would do that's potentially very challenging um I we haven't really addressed that yet we use Fel lenses in our labs to use solar energy to melt metal and stuff um but we haven't actually tried to manufacture one potentially we could potentially replace them with mirrors which would be a lot easier to do some kind of parabolic mirror um but we haven't really gone down that road yet so ID say that's one thing we haven't tackled yet I mean there was a I interviewed somebody a couple of years ago they were looking to build a Luna Rover and they were going to manufacture solar panels on the surface of the Moon just by scooping up regolith mixing in various ingredients and then laying down a film as the Rover goes and I guess like a version of like perovskite or some kind of solar cell material but I guess you're limited by what your self-replicating robot is capable of producing not necessarily what a sophisticated manufacturing robot could produce out of those same lunar resources well I would argue that this solar cell printing robot I'm I'm pretty skeptical I'll tell you the reason why uh yes there's a lot of silicon on the moon and yes you could use you you could basically use G create um you can certainly create glass quite easily on the moon the question is whether it's actually transparent or not because there's a lot of iron in it and iron darkens glass uh you could also potentially extract silica to create fused silica glass which would be transparent the problem and then you'd have to reduce that silica into silicon itself uh to create the the problem becomes with with the dopant and not just the dopant uh it would be um you have to to to create um a solo cell you need something called a PN Junction um and that's where you get two pieces well you get a single piece of Sil silicon they got two parts to it one part is doped with um some kind of what they call a p type dop dop dopent the other one is is doped with an N type dop dopent so you have to control the diffusion of those dopant through the silicon and that takes uh a lot of precision otherwise the solar cell you're going to end up with is going to have a conversion efficiency which is so poor it's not going to be worthwhile um i' would say you'd be lucky if you get one two 3% out of it conversion efficiency which is you need huge carpet of these things to get anything useful out of it whereas with thermionic conversion uh you can get 10 15% and potentially even as high as 30 or perhaps even 40% depending on how you Cascade it with other techniques so I would argue that uh I I I I believe that the thermionic conversion approach is actually a better one um and I don't I I think that printing solar cell solar cells on the moon is a lot harder than people think it will be and actually the resultant solar cell will not be very very efficient right right but I guess if you're looking the bootstrap then you're not incredibly concerned about the efficiency it would be great but if you can even just get to the beginning that's a big step uh yes I think it I think it assumes that we can uh print a large carpet there are also other things of course all these cells have to be connected uh so for that you do need to use metal um now aluminium you can do that with and aluminium we can extract from the Moon yeah I mean I yeah I I I can see how it might be done but I'm still questioning the um I still question the the efficiency um yeah but which is the reason why I went for thermionic conversion because we we need to have a highly efficient system right and I guess the point being that you've you've already got these photo receptive vacuum tubes that you're already having to build and they can do they can serve double duty so yeah um so then so let's let's sort of put this all together and take me through what would be the mission like you've you've done it you've you've made the Prototype work here on Earth you're able to produce a copy of it able to produce a copy of itself you feel comfortable you buy a a a ticket to the Moon uh what goes to the moon and then what happens next okay you'd have this self-replicating seed so basically it' be a full manufacturing unit without all the power parts required you'd have some power but not not not the full capacity so that then get plunked onto the lunar surface and like how big would it be sorry how big uh I imagine it rather like a a fairly large Lorry articulated lry uh so t 10 t 10 20 tons so fairly significant um but we would need about 100 tons to create all the kind of like uh the additional uh Power raising stuff that we would need to go with it um It's Complicated by by the fact that we also have two weeks of night on the moon so we need to store power uh for the nighttime as well all right so you've landed this 10 ton thing onto the moon what what does it look like okay so it it has several pieces to it so you can imagine if you if you can imagine like a shipping container I suppose it's like inside it will have various pieces of unit so it'll have a set of RS which will scoop up reg lith and then that will put those onto a um a set of process physical processes and the first set of processes would be some kind of Bull Mill uh we've done some B experiments in in in my lab and then we grind it so that each piece is about is the size of a single single Crystal single mineral and then what we do is we separate out into part different pieces so we use magnetic separation for separate out some of the iron and stuff like that and then we'd also use electrostatic separation to separate out various components to separate things like felds you know from anorthite from ilite and so on and so forth and then we would have a set a set of chemical proc in fact just one two three sets of chemical processes so one of them would be uh HCL leeching which we've demonstrated in the lab we've taken Highland simulant uh and extracted silica out of it and extracted alumin so that would be the first stage of processing and so we'd keep things like silica because we want to create glass so we can get pure glasses out of that and glass has many many different types of uses um where I did my PhD the whole library was made out of glass including the steps and everything so you can do quite a a lot with glass um the the alumina we alumin itself is also extremely useful it's it's uh has properties second only to Diamond it's very hard it's it's has a lot of very useful properties but we can take some of that alumina and then we can reduce that into pure aluminium so and that's actually something we're currently building as an electrolytic unit to actually reduce alumina directly into aluminum uh one of the other things we also want to be able to do is take this aluminum we want to be able to mix it with other additives so we might want a little bit of iron in there or we well I I know what we want we want some uh nickel and and Cobalt to add into the aluminium create Alo magnets uh we also want to create um conduct electrical conducting wire we want to create structures so these all require different Alloys and so what we're looking at at the moment we're working with NRC to see what kind of Alloys we can generate from lunar resources and uh and what their properties will be so for example quite often we add chromium to metals to make them resistant to corrosion but there isn't we don't have corrosion the moon so that's something we don't need so we can adapt what we currently use and again this goes back to this idea of adapting to the local resources we have available to us and the conditions we have on the moon and then we can 3D print this aluminium into either structures or or or utilize it as part of a multimaterial um object like like an electric motor so the idea is to to in this package you have you know the final bit is basically the 3D printing printing with an assembling system in there so that's the core of it the rest rest it also have some power raising uh pieces in there as well U and you'd also have all your electronics and stuff like that as well and that would be landing on the moon and the first thing it would build would start to build its capacity its energy capacity and then it would build a copy of itself piece by piece using Motors as components and vacuum tubes as components and then St and just basically build it and then build it right next to it that infite him effect right right well I mean I mean that sounds awesome and also pointless if it's just going to build copies of itself I mean apart from boy that's cool it can build copies of itself where does this become uh practical for what Humanity might want to do in space right okay so uh human uh animal any kind of creature their genes are basically can um can build any kind of creature depending on the configuration of the genes the same thing is true with a self-replicator now going back to the John Von neyman who is the original architect of self-replicating machines one of his one of the uh properties of a self-replicating machine is that is based on something called a universal Constructor which means that it can build anything within reason so if we can build a self-replicating machine we simply change the program that's stored to build anything we like and pretty much anything we like we can build any type of machine because we've got Motors we got uh control electronics and anything so we have the capacity to convert what is in effect a general purpose for each of these self-replicating machines is a general purpose Factory it's a question of how how we change the program to uh build new products so in effect when we're growing this um this capacity what we're doing is we're growing productive General general purpose productive capacity um now you know you've been doing your thinking about the moon how would this be adapted to other worlds what about Mars yeah uh in some ways Mars is easier because Mars uh has materials that we can't we don't have on the moon uh one of the problems with the Moon is that there's very little carbon which means we can't it's very difficult to manufacture things like Plastics on on Mars there's loads of carbon it's in the atmosphere so we can simply extract that convert it into methane in fact through the sabatia process and and that actually was demonstrated you know a few year couple years ago uh on on Mars so once we have the feed stock we can then create other types of plastics and so on so we we're on Mars is actually a little bit easier where it becomes a little bit harder is that aluminium is less easy to get access to so but again we can substitute there's lots of iron uh as opposed to aluminium so we can we can make this trade um relatively straightforwardly I think for Mars um and then other places in the solar system I mean like the classic I think is going to the asteroids yeah asteroids um yes one of the one of the problems we have is on the moon we do need me metals that are plentiful in asteroids but not on the moon uh it's reckoned there are some potential or bodies located under the surface of the Moon where nickeline meteorites have basically landed on the moon uh without breaking up uh and potentially we could access those um those asteroid material materials um so you could in theory practice asteroid mining on the moon without all the zero g difficulties that you're going to have I've recently actually in fact I'll be presenting actually in a in a month or two a paper on well if we can't find these old bodies what do we do because they completely completely flaws us if if uh if this is the problem so one of the things I've been looking at is how to move as an asteroid a nickol an asteroid and land it onto the surface of the Moon uh and what we do is we use uh coil guns which basically scoop up the material and eject it as as you know to get Rocket thrust and basically uh we can you use up about 2third of the material moving it to the moon and landing on the the moon but we still get about a third of it landed on the moon a soft Landing so yeah um so in in a way part of that lunar concept is to look at asteroid mining uh and to develop asteroid mining the most challenging part of asteroid mining is the fact that there's very little gravity and so virtually everything you want to do from drilling to moving around requires you to have some kind of hold down mechanism uh because you got no no gravity field at all uh and that presents a real challenge but it is easier to escape the gravity if you want to send it somewhere else but to me that's trivial uh that that is I would rather have a heavier gravity field and have to use fuel to get out of it than this whole idea the whole infrastructure on around an asteroid has to deal with this very very challenging problem of staying on the asteroid um and that is to me is harder than going to a gravitating body and utilizing the gravity that's there and having to launch off um I mean the Moon is 16g um that's a relatively modest gravity field I'd much rather have that than doing things on asteroids because it's asteroids it's just so difficult to stay on it uh and all the machines that you need need to have some kind of reaction they have they exert a reaction effect on the machine that's interesting though because I think when we wonder about where the alien factories are in the solar system where the Von nman probe factories are the instinct is to look to the asteroid belt but if I understand correctly you're saying the best place would be on the moon we should be looking for a gigantic self replicating robot facility on the moon or on Mars uh well I would well okay let I would say that if if some kind of extraterrestrial intelligence was capable of interstellar flight they may have ways of dealing with as asteroids that either we can't do or we find too difficult or we just haven't got the experience uh a lot of it is basically we learn by by by making mistakes we learn through through experience so once we've actually gone to asteroids then we'll probably develop the experience and and the techniques and so on so I wouldn't necessarily say that's wrong but um yeah was if a Von nyman probe came here or was here I would probably still expect it to be be parked in the asteroid belt um because that provides you access to a whole host of materials that you can't just get on the moon um you know you got things you got carbon as well you know carbonation Condes you got water you got ices and things like that which uh um as as well as all the metals um so you've got pretty much everything you need yeah you can go shopping in the asteroid belt to find the asteroids that have the the bits and pieces that you require yes yes indeed and they they're all mixed as well so you got like some which have some of one some of the other and so on so yeah you could shop around quite nice quite nicely I mean what I like about this what you're working on is that it's very practical I mean what is the first possible what is the minimum viable self-replicating machine that we can possibly make and have it be off Planet but this is this theoretically will fall under an exponential growth curve with miniaturization with new technology coming through where does this where do you think this logically proceeds over the decades and maybe even the centuries and Millennia into the future uh well there are two two um two things um first of all uh my goal uh pursuing this what I want these machines to build once they've replicated themselves I want them to build solar power satellites uh so we can build solar power satellites on the moon put them into orbit around the Earth and provide clean energy to the Earth we could do this once we have self-replication technology and we do it on the moon what we're effectively doing is we're doing two things firstly we're making space exploration space colonization directly relevant to every single person on Earth uh and affecting their lives on Earth providing clean energy the second aspect is that it actually effectively offloads the entire energy generation business or industry onto the moon and thereby providing relieving the Earth biosphere of that particular industry beyond that I think that once you have self-replication capability then it opens up the entire solar system and eventually the Stars uh there's no doubt I actually believe that the only way we are going to go out I mean you look at history we landed on the moon in 1969 and never went back there maybe we might be to get there for the first time for the second time next year space flight is is only has not been made relevant to people and we have to make it relevant to people to actually get people to go out there to make it important for people if we put self-replicating machines on the moon we have productive capacity we have assets once you put assets somewhere now you've effectively lowered the barrier to Commerce and so PE I'm sure people will invent ways and find ways to make money out of out of this kind of capability once people figure out how to make money then the people will follow and I think again we can use the same approach by going to Mars and to and be on to other other moons and and and planets and whatever um to become you know a solar systemwide uh species so I I I do think it's really the only way because it's the only way we could do it at low cost or or minimal cost should I say uh and then of course eventually we're going to start thinking about well if we could do this in our own solar system uh maybe we could send out V probes to other other stars within our galaxy and effectively again if you crunch the numbers you could cover the entire galaxy with self-replicating machines in about a million years which is a very short space of time astronomically speaking right and I think actually is a powerful argument against the existence of extraterrestrial intelligence because we we haven't found any I I I do as well that the fact like the question is not why haven't we heard signals from Alien radio transmitters the question is why isn't the solar system filled with Von noyman probes sent by various species throughout the the Milky Way I mean I guess I've got sort of like two timelines so I guess the first time is like when do you think based on what you're working on that first robot Factory creates a a copy of itself and when do you think we send out that first one to another star system or when did when will it arrive when do we get a copy of a robot Factory in another star system uh okay so uh there are two questions there um the first one if I had the funding that I I needed I reckon we could get the first first prototype self-replicator working within about five or six years wow on earth right then it would take a few more years to get the first prototype to the moon but the advantage of a prototype is if it works if it is truly self-replicating um it basically can be reprogrammed so if it goes wrong then we simply set set up a new program to change whatever it's making so as long as the first one works or can we can get it to work sufficiently well then we can correct it without sending up new machines so that again is very cheap way of doing it's much cheaper than sending up new machines and trying them out and testing them out so I would say that within I I think we once we've created a population of these machines on the moon that would be relatively quick uh because once we've got one or two on the moon and then you got this geometric expansion rate um then you could basically have have a million units uh within 13 generations and if each generation takes an hour a year we got takes 13 years so very quickly we could have industrialized the moon once we've developed the technology from the Von nyman Pro perspective um I would imagine that once we have this capability and it makes effectively opens up people's minds to interstella flight I think because we've got we've got the payload already sorted out um I don't think it would be that long because once once we have the the the um the capital uh what call it it's like the machine C Machine capital or on in extraterrestial environments is relatively cheap to try and build things uh from local resources so we could build start building Starships uh relatively cheaply I think think uh once that happens and then well it depends on how successful you are in developing some kind of propulsion system right right that's that's someone else's problem yeah yeah yeah um so Alex I think one concern that we have is you know here on Earth if you just talk about exponential industrial progression this runs directly into our environment about about how we treat planet Earth and so when someone thinks oh within 13 Generations you've industrialized the entire Moon that starts to sound awful so how do you think about sort of preservation of the Wilderness of the environment of the solar system in this sort of exponential growth of of industrialization yeah um okay there are a couple of uh thing a couple of points in there and I think first of all we look at sustainability now yes you're quite right the exponential growth can happen but within one of the requirements for self-replication is is sustainability because you have to close this physical Loop and the energy loop as well so that forces us to do things sustainably uh or sustainably as sustainably as we can the second thing is that what I'm I'm not I'm not mining water which is a scarce resource on the moon um when in fact I'm actually not in favor of doing this of mining water and then sep and then burning as propellant there are plenty of other uses for water where we can recycle it so burning water on the moon as propellent I I I don't think is a good way to go down that road the second so what we've what we're we're working with is common rock forming minerals they're everywhere and to even a million units is nowhere near exhausting that resource it's it's it's a drop in the bucket as it were um so from that perspective as long as we control the growth or limit the number of generations we we we uh we build um then there's there's no problem there now if we that of course requires us to control that growth that that replication uh so I actually wrote a paper on this uh suggesting that are various things we can do what one is to implement a two-l kill switch so oh I think I reported on this right that we yeah that we need to shut these things down at some point yes so the again what's appeared to be a disadvantage that we had to take salt from from the Earth to the Moon that is essentially a salt contingency so if we deny salt it no longer can replicate the second uh uh kill switch is actually on the moon which is located where the asteroids are either the ores or where we put asteroids uh so that we need all the nickel and Cobalt from there and selenium and stuff like that so that also acts as a kill switch because we can shut that off that supply chain off that's how we deal with it locally uh in addition we also want to we also want to implement something like a hay flick limit uh an a haick limit is a is a biological limit to the number of replication Cycles a cell goes and it's based on telr these are like the ends of uh DNA and they get chopped off they act like a counter a program counter um and we can Implement a similar kind of approach for a self-replicator so we actually Implement so that with the within the coding system uh we have a te set of telr essentially and they get chopped at every replication cycle you can arrange that physically so that when your reader your reader goes to the first point that isn't replicated and so what it does it's it's only reading stuff Downstream um and so every replication cycle that gets shorter and shorter and sorter until it starts to you know chop the code itself uh up and it can't replicate anymore so that's another approach and then of course there's the the approach which is used in uh Communications satellite Communications spacecraft Communications which is error detection correction coding uh we use this for memories and things like that in space and so on so this is a a very well established technology um so the implementation of all these different techniques allows us to provide a multi-layered defense against gray goo or run yeah yeah turning the entire Milky Way into self-replicating robot factories yeah yeah that's really interesting um I mean exponential growth has a way of getting away from you and we see this with pandemics with you know all kind bacteria infections all kinds of things that that yeah something this is an engineered machine so this I I would argue it's hard enough to get the damn thing to work uh let alone you know having having it going running a marck and everything I think that's uh for me from my perspective that appears to be very unlikely but we can Implement all these safeguards uh to ensure that that doesn't happen so what about self assembly how so this the self assembly thing is basically to is to I want to create a module a self assembl module and most people have created these modules with self- assemblable into structures and so on and so forth what one I want to look at is the actual module itself so I want to incorporate a I want to 3D print an entire module so the and this would demonstrate again the 3D printing of a modular robot essentially um so we already 3D printed the motor aspect the next bit we want to printed some of the electronics and some of the sensors so and and some of the latches latching techniques as well so the idea is from a going from it's it's a step to a Step Beyond the motor to prove that we can 3D print other aspects and package it into a single module and create a reconfigurable robot type of thing so it's it's basically a variation that on that same theme right right well Alex it's been a pleasure to talk to you I look forward to the first self-replicating robot Factory on the surface of the Moon and and hopefully a limited number of them uh filling the universe with greu uh but it's a it's fascinating to think like I think this is the inflection point that if we can crack this problem then we can bring the resources of the solar system to us in a way that enables us to continue our existence here and not have to draw upon the resources of planet Earth anymore in the way that we do like this is the way that we become more sustainable to to planet Earth and I'm interested to sort of see how this all plays out uh good luck thank you very much it's been a pleasure I hope you enjoyed this conversation with Professor Alex Aller I'm going to give you some more of my thoughts in a second but first I'd like to thank our patrons thanks to Steven fer Munley Paul roorbach Abe Kingston hey Twilight douge Stewart Steven krosaki David Richards Mark antis Joel yansy Antonio lilara Dustin cable Vlad chiplin moo George David Gilton on Andre gross Jeremy murn Josh Schultz and Jordan Young Who support us at the master of the universe level and all of our other supporters on patreon I think this comes to the heart of for me the most powerful answer to the firmy Paradox that we are first or we are alone in the universe because we are getting to the point now within a couple of decades that we can start building self-replicating robot probes here in the solar system and over time as the technology follows this exponential curve we will build smaller devices they will replicate faster and eventually we will send these things on a journey to the nearest star systems and when they get there they will make more copies of themselves and then they will move on to the next star systems and so if this is a thing that we are right on the cusp of being able to do then it stands for reason that other Advanced civilizations across the Milky Way have done the same and no matter where they start out within a couple of million years they will have essentially visited every single star system in the entire galaxy and where are they and so the better we get at building self-replicating robot probes the more troubling and weird the firmy Paradox becomes and I am excited about the technology I'm looking forward to what we can possibly accomplish as we expand out into the solar system and become a true solar system civilization all right I've done some interesting interviews with the ways that you might move to other star systems maybe not at the speed of light but as fast as possible to make the journey within a single human lifetime so check those interviews out all right we'll see you next time

Channel: Fraser Cain

Views: 38,303

Keywords: universe today, fraser cain, space, astronomy, exoplanets, James Webb, jwst, James Webb space telescope, tess, Ariel space telescope

Id: vo863_0SI_o

Length: 61min 46sec (3706 seconds)

Published: Mon Mar 04 2024