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

I am a self replicating space probe that is, at the moment, stuck on earth.

Had no idea people at my university were working on this topic. Never expected to learn about it from this channel.

You gotta read “We are legion (We are Bob)” by Dennis E. Taylor.

Some years ago I heard that a cellphone made with vacuum tubes, that had the same processing power of computer chips, would be about the size of the Empire State building. We aren't exploring the galaxy and sending back in information using vacuum tubes. For a machine to land on another world and make a copy of itself, it would need to find, mine, and process a LOT of different elements. It would also require metal foundries to make several different types of metal alloys. Then it would need to turn the metal alloys into high precision parts.....

This guy is constantly setting off my BS detector. There's no doubt that he's brilliant. But, he's constantly conflating his goals with actual achievements. When you say that you've 3d printed a motor and then have to clarify that you actually didn't... That's not ideal.

I may have missed this in the talk, but the genius of the vacuum tube idea is that the moon is in a vacuum, which means you can dispense with the tube (which makes it a misnomer) and simply have the internal elements of the tube exposed to space and have them function. Thus you could print whole circuits without having to enclose anything.

Only 20min in to this interview and I’m fully onboard with our steampunk space future.

Magnetic Rope Memory and Vacuum Tubes come back from Apollo days for the win!

Very interesting conversation. One topic I didn't hear was housekeeping. I just went to my little 3d printer and on the floor was the occasional splinter of PLA. This got me to thinking about the accumulation of micro-debris created by this process. Watching discussions on lunar dust the astronauts were plagued by this dust at it would cling on to everything. I'd be interested to hear of the disposal processes for incomplete builds, collection of micro-debris and control of dust contamination, e.g., from lunar material, processing of that material, movement and positioning of material for introduction into the manufacturing process and all the while maintaining a suitable environment free of dust and debris.

i always miss the part where they talk about "step 0", which is the moment a probe arrives to a new star systems. but then what? how's it supposed to scan and inspect all the planets, moons, millions of asteroids etc and find all the resources? that alone seems like a giant challenge when it comes to rare materials you need for microchips, and for fuel, etc. i think this is one of those things that is not impossible, but looks to be such a logistical nightmare that's just not worth it. not even talking about the fact that this entire thing needs to be autonomous (AI?)

That interview was super cool. I like how at the end, the discussion assumes we did all the other super hard things and have to worry about fail-safes and safeguards :) Pretty cool.

We’re still centuries away from a self replicating robot. Von Neumann probes, you’ll still need to figure out a power source and means of propulsion that can function for thousands of years. I don’t even know what that would be.

Thank you Frasier for sharing this. Being an Alastair Reynolds fan, I cant stop myself thinking about the book Galactic North as I watch this talk!

Love this video Fraser, thank you

Very fascinating and thought-provoking! Thanx!

This is such an interesting topic. There are so many things I have never considered about the challenges a Von Neumann probe would face.

Professor Ellory is very right about the challenges of a machine manufacturing a copy of itself. If you take the example of current 3D printing technology, printing metal requires a ruby-tipped nozzle. How do you print that? Do you get into a rabbit hole of needing continuously smaller nozzles? Do you have to print a machine that can hunt down all the materials, another to mine them, another to peocess them? Or do you do a ground-up redesign to much sumpler technologies. (A blacksmith once told me that if you have a hammer and an anvil, you can make as many hammers, or other tools, as you want, but if you don't have both, you can't make anything.)

A joy to listen to a person with big ideas for the benefit of earth and not for self interest..

Great show! Thanks

Love this podcast, gogo Fraser !

Did he ever mention software or a stored program in the video? The self replication including mining, refining, and assembling parts seems incredibly complicated and needing not only software but pretty sophisticated software. Doing all that autonomously is at the fringes of what is possible on Earth without the self-replicating requirement. It is hard to imagine the complex behaviour working without having the computing power of something like an Apple 2 from the 1980's. He mentioned using vacuum tubes which could act as transistors. 4:05 mentions analog neural networks but it is hard to see that helping significantly with the complex behaviours needed when neural networks are so specialized. Making a computer comparable to an Apple 2 out of vacuum tubes would be extremely huge and likely make this idea impossible. It sounds like his ideas are at the very earliest stages of brain storming and could be 100+ years from actually working on the moon. His estimates around 49:30 are obviously underestimating the time and effort needed to help himself get research funding. Near the end, it sounds like he's talking about sending new software to the moon if the robots mess up. I didn't hear software mentioned elsewhere. If they have a reliable feature to replace the software controlling all the robots, replacing it with software to simply shut down all robots would be easier than having some of the other safeguards he mentioned too.

To achieve this you need: Energy Mobility Mining Refining Production Assembly Software And the mining plus refining is probably the most difficult part to pull off

Hey Fraser! I found your podcast a few months ago I guess, and I love it! This is how science journalism should be conducted: by someone with a STEM background, with genuine passion and curiosity. (Of course, credentials aren't crucial, but some knowledge of the relevant subjects is key.) Thanks for enriching my internet experience!🙏

this not real-talk to soon-but i love this channel

Must confess that I had to lay down after 43 minutes of listening time. Wow! Thank you for this stretch of tape.

People in the comments really don't like speculative ideas, concise predictable conclusions or its all nonsense! I like the speculation, myself. Ideas have to start somewhere. Good interview!

Good conversation! I read Alex Ellery's paper 'Are Self-Replicating Machines Feasible?' and enjoyed the reference to the RepRap 3D printers I was already playing with. All of this is an extension of Dr Bowyer's RepRap concept.

A.I.s, A.S.I.s, self replicating robots, self replicating space probes…. What could go wrong?

Did you mean Van Neumann probes When ? In the thumbnail ?

It's interesting to learn that we are so close. In your closing comments about the Fermi paradox, where you ask where the other probes are, maybe other civilisations aren't interested in exploration and quite content to stay at home? We seem to think that all intelligent life is like us, including their motivations.

We are a long way from a self-replicating robot. Such a robot would need to be able to select the right materials and process them, such as turning ore and scrap into the proper metals, not to mention shape it and assemble it. At present, it takes an assembly line of robots to put the parts together, refineries to make the ingots from which it will be made, various materials to make chip wafers, and still other highly toxic materials to etch the chips. All of those need to be manufactured since none of those chemicals occur in the earth in any usable quantity. All these manufacturing capabilities would need to be miniaturized and so on and so forth. Not only is it unlikely to happen anytime soon, but it also would not be particularly practical, and would also be very inefficient. This isn't something that is right around the corner, or even way down the road. It might be someone's opium pipe dream, but that's kind of it.

The father of the Borg, 'Resistance is futile!'

The Code of the Lifemaker's prolog has a great take on this. When the damaged exploitation ship lands on a planet, the damaged machines start an evolutionary system creating an 'artificial' ecosystem, including sentient humanoid robots. But if I remember correctly, there were several initial types of specialized robots that eventually built a factory to build more robots.

Hi Mr. Cain! I've got a random question for you: How does the "interstellar medium" influence the speed of Light? As in: Light traveling through a Medium has a different speed compared to Light in a perfect vacuum. How are changes in density over interstellar distances compensated(i.e. when calculating distances etc)?

How does one create insulators on the moon? If Al is the predominant material, is fused quartz the insulation ? How would the O2 not get lost?

When Professor Ellery said life has been an example of adaptation and exaptation I remembered a response by L. Ron Hubbard to a question about von Neumann probes in the Q an A after he talked about his proposed Starwisp System at a 1982 Spaceweek event. "We, and our society and industry are a self replicating system."

In the sixties computers were made with vacuum tubes. I worked with one that was so big it was installed in a warehouse. I would like to see a computer that size transported in a space ship. Whomsoever is suggesting this should have his medulla examined

Damn I’m feeling old, I remember growing up our TV was full of vacuum tubes.

05:20 - Can someone please help - is a "thermionic device" more efficient at producing electricity than a "semiconductor-based device" (for example solar panels)?

The idea of building motors from lunar regolith is funny considering how deadly that regolith is to machinery.

Can energy be created using some form of heat pump? There’s a huge temperature difference between a material facing the sun, and that in shade.

15:00 - Why can't we use transistors - aren't solar panels essentially made from semiconductor transistor technology?

Really curious how many vacuum tubes would it take to recreate a vacuum tube?

For a vision and control system, a central "Eye" that can be higher resolution and do all the computation for an army of robots would really help. Each robot could simply be 3 motors and simple commands completely controlled by central towers, rather than each robot needing hardware for complex computation and control.

Thermionic conversion on board is important part of thermomanagment system of spacecraft in space where only black body radiation is the only option to remove heat.

We already have Von Neumann machines.... Theyre called HUMANS.

I am too pessimistic with my following thought, I hope. Nobody currently working on this will live long enough to see these machines happen. A bit saddening.

He really understated the complexity of implementing the idea. I don't expect to see self replicating machines in my lifetime. Maybe a machine that can assemble ready made parts into another version of itself, but from scratch? No chance. Also the idea of space based solar power on Earth is garbage. Modern solar panels are typically 20% efficient, lasers are up to 80% efficient, transmission through the atmosphere is about 70% efficient. Just with these you would need 31m2 of panel in orbit to get an extra 1kw of power from a solar array on the ground. Then you have to aim the laser which will not be steady or perfectly collimated and the amount of incoming power (and size of space array) required to get 1kw could go up 100x

There's a lot of hand waving for interstellar Von Neumann probes. First the big one is propulsion. We aren't going to be able to send the heavy payloads needed to travel stellar distances to another solar system, travel within that solar system, and then be able to replicate well enough to do it all again. Oh, and all of that is going to have to be able to work after hundreds, to tens of thousands, of years of inactivity during space travel. For that to be even remotely possible, the probe is going to have to have a ton of redundancy and ability to repair, which is going to add to the weight. More weight means even slower acceleration and deceleration, which means even longer travel time, which means an even greater need for redundancy and repair, which means even more weight... rinse and repeat until you realize it isn't possible. Then there is the grey goo problem. None of the limiting factors apply to self replicating systems that can replicate enough to cover the galaxy. If a replicator can create enough copies to go to billions of stars, then it can create enough copies to completely overwhelm any one star. Either there is a hard limit to replication or there isn't. Trying to do something in between means it is open to Murphy's Law, and is no longer a hard limit.

Very interesting topic. Also what is the deal with these LLM generated comments from accounts with sexualised pictures?

The obelisk in Arthur C's 2001 series was a Von Neuman probe by every definition. Something like that is beyond us and is akin to "magic".

On earth we make technology from materials shipped from all over the planet in complexes of factories and trade networks. I've often thought that a replicator will have to be quite large, and it would need to be a swarm of robots spanning a large area and probably multiple planets or asteroids. We can simplify the technology like he says, but then the things will have very limited capabilities. Computers made of vacuum tubes aren't going to be running much AI. Not sure how vision is going to work, for example, and that seems kind of important for exploring and exploiting your environment. He does seem to be addressing all the right issues, so who knows. Another challenge is that his team has to invent an entire suite of highly specialized technologies that have few earthly market niches (then again, they probably said that about NASA and the ICBM program, and large markets for what they came up with did appear over time). Lunar replicators have two killer apps: manufacturing millions of orbital sun shade foil panels for solar radiation management to counteract AGW, and perhaps manufacturing solar panels and structural components for orbital power stations. If these are going to fan out into the galaxy, it would be nice if they could replicate our best technology, when conditions and cluster sizes are right. Build high level AI robots, somewhere on the other side of the galaxy in 20 million years. Is there some way it can carry instructions for this?

Vacuum tubes used to covert heat to electricity, not Pneumonic Memory arrays, correct?

Is the entropy an issue for self-replicate gray goo within Milky Way?

How do we 3d print a 3d printer? A subset of this is how to 3d print computing components, neurons, transistors, integrated , memory modules? Or lasers? Other huge challenges are to 3d print all of the sensors needed? To 3d print or otherwise manufacture spacecraft, engines and rocket fuel or alternative propulsion technologies. I don’t want to be completely negative just point out how much research & development is needed for this to be demonstrated. As others have stated - maybe centuries from now, if ever.

von neumann probes seem very interesting when paired with breakthrough starshot

A humanoid robot with comparable dexterity is really all you’d need to make a self replicating probe.

lol. We are astronomically far away. We can’t even make things that self repair, much less self replicate.

Analog is a superior signal.

With the glass for the solar panels why would it have to be thick enough for the transparency to be a big issue? Presumably it would only have to be thick enough to shake off dust that got kicked up onto it, it wouldn’t be helpful to have it thick because it wouldn’t help it survive an impact from orbit and there is no weather to keep out.

Vacuum tubes, should be almost trivial? Can’t we print the cathode, the anode and the grids, no need for the glass, as the vacuum is already present? Does the glass have a ray containment role? If so can’t we use a simpler ray containment? Also simple printed circuit boards, with single transistors should be doable? The 1960s boards had no ic chips, and used discrete transistors, so no chip fab?

@Fraser Cain Why have spelled it "Wen?" not "When?" on the thumbnail..?

still he is working on some very interesting stuff with data compression, maybe a new way of programing modular proccessing.

Before listening to the talk, my estimate to how close we are: it's going to be a while.

A Von Neuman probe would not be so much a little gadget, as it would be a ginormous armada of hundreds or thousands of factories.

104 comments, but only 54 show up. RU-vid is "better" everyday!

Is the Dutch pronunciation Von Noy-Mann as John Michael Godier says it? And the American pronunciation Von New-mann as Gary Nolan pronounces it?

Soft landing asteroids on the moon? Oh, please.

With space-freezing temps, likely ice will be a significantly useful material. Doping ice with conductive material and ice with structural qualities may be practical and easier than forging metal components in space

does the thumbnail say "WEN?" is it supposed to be WHEN?...just trying to help or maybe I don't get the joke.

Or the reason there is no sign of a civilization using self generating robots is it does not work as well as we want to imagine. Because we can partly build something does not mean it will develop as we thought it would. I love the idea of working on this and developing for possible uses in special use cases but betting it will not be a grand scale operational system.

brilliant, how do I get a job working for this guy??

There is a big difference between a machine that can replicate itself and a machine that can both replicate itself and a propulsion mechanism capable of interstellar flight.

Do these machines make paper clips?

IMO something like nasas Asteroid Retrieval Mission architecture would be a better starting point for ISRU including the ultimate goal of self replication. * You don’t need the lunar architecture, just solar electric tugs. I think these will be cheaper to develop and run long term. * you can grab all different sorts of asteroids, from volatile rich to metal rich, whatever it turns out is useful. * it would probably be much easier to build and run solar furnaces in space than on the moon.

Spiking networks!

Watch out for that grey goo!

At some point adding a life support system and a human crew starts to sound like a simpler, easier solution

Having the probes only grow on light flat surfaces would help keep the water resources pristine because they wouldn't grow into the meteor craters where the bulk of the water remains un boiled by the sun.

WEN?

Reminds me of Autofac, a short story by Philip K Dick.

I have seen someone 3dprint Fresnel lenses, i think with both filament and resin. I have also seen someone use a Fresnel lens to 3dprint sand in a desert. I think producing Fresnel lenses is doable

I'm confused: What is the evidence for the presence of a solid core at the centre of the gas giants? I've heard a variety of claims from a bunch of different sources. Could you please clean this up for me?

The technology for building self replicating robot probes is about the same, maybe harder, than building self-replicating human habitats. By the time we could send out robust/reliable/flexible robot probes, we will probably just be sending huge ark colonies

Von Neumann!!

Just make sure to have some ARGs ready in case they go crazy.

I look forward to somebody experimenting with a self-replicating machine here on Earth.

I think _Relativity Space_ has, somewhere in their secret long-term plans, self-replicating constructs, spacecraft or what have you.

09:50 - What a great topic, and an interesting discourse. Already, at the start of this video, the issue seems to be pointing to AI.

Can you imagine robots looking for minerals to use to make metals is chips?

I've seen that movie. Scared the crap out of me as a teenager. It's called "Die Insel der Krebse" (based on the book "Crabs on the Island"). Self replicating robots on an island that use anything metal to make larger versions of themselves. Including robot cannibalism.

What about making a bionic self replicating factory. Maybe it would be easier to make certain parts out of biological components like computers out of brain cells.

Frasier - I do think it would help if interview videos started with “Interview:” in the title - just to help us scroll through your content and find them…

I would love to see on-Earth demos of self-replicating devices in labs to better understand how far from fantasy the idea is. I'm a skoptomist, a skeptical-optimist ;-D. Don't tell, show.

Make more sense to have probes building bases we can just move into when we arrive a bit later. from there we can make the next probe decide where it should go and send it. Sending out probes who only build probes is pointless, especially if they go beyond communication range. When it comes to asteriods it makes more sense to have large ships who have their own gravity think the ship in dead space, and can eventually build another copy but is exploring with a lot of smaller probes. No need to land on asteriods if you gobble them up. The larger ship the larger asteriods you can process.

Prof Ellery seemed reluctant to discuss doing a test of concept here on Earth, but wouldn't that be a necessary first step? He said he's concentrating on the Moon, but isn't the *self-replication* the important bit? You could do all the tests with a sand-box of imitation regolith, and then modify for actual Lunar use.

Every probe carries a power source, a large CO2 laser (eg cutter, can process minerals to extract pure elements). Next, get construction of room-temp superconductors working, then steel will act as an insulator by comparison, eg just lay conductors as tracks of superconductor on steel structure. Given 1000 years with a strong goal, and all required information, should be do-able. Vacume tubes won't work because most interesting targets are not in a vacume, and memory requirements are too huge.

Fraser, look at what a company called Cadence is doing, not so much robots, but machines building machines? you will find it interesting.

Mantrid drones

It seams turning inovationtime backward just use what you have from Earth chip sensors set up a hardware transport hub. So the raw infrastructu transforming can happen. Later one can use biodomes bacteria to breakdown and retrieve metals. Or just simply melt. K.i.s.s principles.