This episode is sponsored by Brilliant
Every year hundreds of elves are tragically injured in toy-manufacturing related accidents
while reindeer are driven to the brink of exhaustion… but in the future they may be
replaced by robotic factories and delivery drones. So today’s episode comes out in the midst
of the Holiday Season, and of course many winter-time holidays have a bit of focus on
gifts and good cheer, which I’ve often suspected was meant to help counteract the rather glum
mood that comes with the cold and snowy time of year. Of course the most familiar avatar of that
gift giving is Santa Claus, so it’s no surprise that one of the common terms for hypothetical
devices that can make almost anything is a Santa Claus Machine, one of the most well-known
fictional examples of that being the Replicators in Star Trek, which can produce many things
by a simple verbal command, notably tea, Earl Grey, hot. These are also often known as Cornucopia Machines,
for the horn of plenty, and if we’d been doing the episode around Thanksgiving we’d
probably have titled it that, but it’s December so Santa it is. As a young and rather credulous kid I once
asked my grandfather why Santa was at the North Pole and in an impressive display of
ad hoc composition he said it was because there was a large supply of minerals and oil
in the Arctic regions and Santa needs lots of raw materials to make presents. There was no Google or Siri back in those
days so he was fairly used to fielding bizarre questions from me and I suspect breathed a
sigh of relief when my sister finally taught me to read, but it got me thinking on the
logistics of Santa’s annual gift giving operation and I tended to revisit the topic
every Christmas as I learned more math and science and grew increasingly shocked at the
sheer scale that would be necessary. In hindsight I’m not sure why I didn’t
just look up the statistics of the Christmas industry, which if you’re curious, in the
US alone is a 500 billion dollar annual industry estimated to involve almost 5 million jobs
at home and abroad. Now the assumption that this means 5 million
of our proverbial elves working in mega-factories at the North Pole would be incorrect, most
would be in other aspects of the manufacturing and delivery process, tens of thousands of
elves doubtless work on arctic oil rigs or warehouses, and some presumably have the unpleasant
task of cleaning the reindeer stables. How many actually work in the vast polar foundries
and factories would depend on the degree of automation involved, and ideally almost all
these processes could be automated. This of course takes us into our topic for
today. Machines able to produce anything, in fiction,
generally tend to appear to do so almost like a magic wand, it just appears or fades in. Likely simply as time saver for special effects
and props departments who can just pause the recording and stick the item into the shot. However, such a machine need not be arcane
and compact, but more likely a very large automated factory, at least initially. There are many hypothetical and fictional
examples of entirely automated manufacturing, but the most likely is what we call a Clanking
Replicator, an artificial self-replicating system that relies on conventional large-scale
technology and automation rather than nanotechnology or teleportation. Think of a large automated factory that was
able to deploy robots to gather raw materials and could turn those into every device needed
to operate that factory, including those robots, who presumably also do the assembly. Non-artificial examples of a clanking self-replicator
include humans, and really any organism. Indeed all our cells. A typical cell, while microscopic, is less
of a simple machine or factory than a giant city full of warehouses, refineries, factories,
roads, and distribution networks. Evolution has created a dizzying complexity
in each cell by trial and error, without the need for a brain. Each cell is a self-contained self-replicating
survival machine. We can assume it will be possible at some
point for us to make our own artificial ones, even if we have to piggyback on existing biological
cellular structures to do this. For that reason, while many of the devices
we’ll discuss today are more speculative and iffy, the Clanking Replicator itself is
considered an inevitability, a matter of when not if. The other more classic example in sci fi is
tiny little nanorobots able to reproduce and perform tasks, which raises the doomsday scenario
of what’s called Grey Goo, which is when they run loose and turn a whole planet into
more of themselves. This is also likely possible if we’re just
discussing cells, since cells are microscopic and so are viruses, which are even smaller
and essentially a working example of grey goo. Indeed you could argue our planet has already
been grey gooed, or maybe green gooed, we just happen to be the end result of that. Much smaller ones, able to manipulate individual
molecules or even atoms, are called Universal Assemblers, and their feasibility is a bit
less certain. Eric Drexler, the gentleman who got the debate
started in earnest with his 1986 book Engines of Creation, offers us one example called
a Molecular Assembler, ultra-tiny machines able to take individual molecules and build
anything, including more of themselves, from them. About this same time we got our hands on Buckyballs,
a new carbon allotrope that opened the doors to carbon nanotubes, graphene, and serious
discussion of nanotechnology. The inventor there was Richard Smalley, a
chemist who was born just down the road from me in Akron, and would later receive the Nobel
prize in chemistry for the discovery. He and Drexler got into a debate about the
viability of such machines now known as the Drexler-Smalley debate, a series of articles
and open letters the two exchanged in the early 2000s. The debate isn’t about the possibility of
manipulating molecules or even individual atoms. Specialized proteins, called enzymes, in all
biological cells and even viruses manipulate molecules all the time. We’ve also had industrial solutions since
the 1980s that can manipulate individual atoms using scanning tunneling microscopes. The debate rather centered around whether
we can miniaturize these types of machines and make them self-replicating. Both Drexler and Smalley raised many good
points but of course the key one from Drexler is the notion that you should be able to make
something very tiny able to reproduce itself at least a couple times before breaking down,
leading to an exponential growth process same as we’d see from bacteria in a petri dish. We can’t be sure yet what that minimize
size can be but can use biology as a safe upper maximum. One of the points raised by Smalley though
is of special interest to us today, as is the basic concept which can be expanded to
other types of these machines. This is the Sticky Fingers Problem, sometimes
also called the Fat Fingers Problem. Any given machine is going to need some sort
of finger equivalent to interact with whatever media they are in and to build stuff. These have to be made of atoms and thus have
an irreducible size. I can use my hands and fingers to handle large
object easily but need special small tools like tweezers to work on anything smaller. To quote Smalley, “Because the fingers of
a manipulator arm must themselves be made out of atoms, they have a certain irreducible
size. There just isn't enough room in the nanometer-size
reaction region to accommodate all the fingers of all the manipulators necessary to have
complete control of the chemistry.... [Also,] the atoms of the manipulator hands
will adhere to the atom that is being moved. So it will often be impossible to release
this minuscule building block in precisely the right spot. Both these problems are fundamental, and neither
can be avoided. Self-replicating, mechanical nanobots are
simply not possible in our world.” Needless to say, even if this argument is
true, it does not rule out self-replicators, just ultra-tiny ones, again biology offers
billions of different schematics of functioning microscopic ones. Nor do you necessarily need precision placement
on atoms to get them in the right place. Time will tell on those but the problems shows
up when we discuss 3D printing. In 3D printing you are layering a substance,
or a few different substances to create an end product. The higher your resolution, the smaller your
printer heads or nozzles need to be, or the more numerous. As with the fat fingers problem, while we
could assemble a vast two dimensional array of nozzles, those nozzles still must be bigger
than whatever is coming out of them, even if we assumed a single material, not multiple
substance with their own nozzles, so the nozzle array does need to move back and forth or
side to side to print a layer, just not a lot. Way back in the self-replicating machines
episode, I mentioned that there’s a maximum speed you can print at an atomic or molecular
resolution. We probably didn’t spend enough time then
to clarify some objections but we’ve got some speed limits on this. Firstly, you obviously cannot use just a few
nozzles, one for each substance, as trying to print out anything, even microscopic, with
atomic detail on any sane timeline would require the needle to be move so fast it would generate
sonic booms and tear itself apart. Anything Macroscopic would require it to move
at relativistic or even FTL speeds and require the kind of superstrong materials only found
in science fiction. Santa’s toys are going to be blown apart
and end up a useless wreck. Here’s our second speed limit, printing
at this resolution, the atomic or molecular level, means anything macroscopic has on an
order of a billion layers, so even if you have many nozzles, they all have to be moving
in unison, each with its own area of coverage. Only materials don’t actually move in some
perfectly rigid fashion, they compress and wave. A common thought on faster-than-light travel
is why you couldn’t have a very long and rigid stick from Earth to Mars and just push
on it on Earth and poke someone on Mars instantly. This doesn’t work and indeed they’d not
get shoved by the end for years because you didn’t push on the stick, you pushed on
a package of atoms near one end which then push on the next set closest to them, who
do the same and so on, sending a wave that is moving the stick and doing so at typically
a few times the speed of sound, depending on the material. The short form is you can’t just shove on
some big printing nozzle array and expect every last nozzle to move instantly the exact
right distance to the sub-nanometer scale and disgorge their next bits or pixels on
the right trajectory, then a nanosecond later get shoved again and repeat. So you can only go so fast and maintain accurate
resolution, and such a device is pointless if it isn’t accurate, I don’t need to
3D print a concrete house at molecular resolution after all and thus can move faster. Now you could angle the nozzles, like pivoting
cannons, each shooting at their own area, though it’s always worth remembering that
you need some machinery for that kind of thing and that takes up space, and the more area
each has to cover, the faster it needs to spin and stop, and as an example, a kilometer
long rail gun on a spaceship can’t pivot and shoot at a new spot in an instant because
turning that fast would rip the gun right off its mountings or snap the barrel from
the centrifugal force. None of these nozzles are being made out of
unobtainium, they have to obey real laws of materials, acceleration, stress, relaxation
times, inertia, and at this scale, quantum mechanics. If you can get through all of that, you still
have another speed limit, which is how fast your ‘ink’ is moving. If you’re throwing colored balls at the
wall to stick and form a pattern, you can only throw those so fast before they’d start
bouncing off, or smashing themselves and the wall to bits. You can’t throw atoms and molecules around
to hit each other at supersonic speeds, let alone relativistic ones. When those atoms land on the previous layer,
they have to be moving slow enough to get caught and not damage the connections of their
neighboring atoms. Nor can you spit out new volleys right behind
them which leave before the previous one landed or you’ll start getting lots of interactions
between the particles as they approach. And if you’re thinking you could get around
this with extreme precision calculations, don’t forget quantum, because we are doing
this at the scale where something’s momentum and position starts getting uncertain. And remember any extra energy of motion of
those particles is turning into heat. Printing circuit boards at the atomic scale
is probably doable, printing a cup of tea in a few seconds is probably impossible. One hates to imagine the sheer amount of computation
that would be needed to rapidly print any macroscopic object in a few seconds without
everything getting jumbled up, but it would probably slag the device, and possibly the
city it was located in, from all the heat generated running the math. We’ve another example of making tea beside
Star Trek, in the Hitchhiker’s Guide to the Galaxy, Arthur Dent ties up the whole
computer in his quest to synthesize a proper cup of tea and it’s not hard to imagine
why - any macroscopic object contains trillions of trillions of atoms, each of which would
need to be precision placed. Of course the classic Picard Star Trek menu
option of Earl Grey presumably relies on transporters to teleport stuff into place. In theory you might be able to use quantum
entanglement to copy an existing cup of tea on to a pocket of matter, but since the no-cloning
theorem would wreck your template tea cup and entangling trillions of trillions of atoms
together just to make a cup of tea seems fairly ludicrous, we can probably discount that approach
to teleportation replication in the future, and it is our best option for making teleportation
technology under known science, see the Teleportation episode for details. One last major concern, which may be an even
bigger problem, is heat building up in whatever you’re printing. After all, on atomic or molecular scales,
heat and motion are the same thing. This is something that we discussed back in
the Nanomachines episode, but is definitely worth bringing up again since it’s one of
the biggest problems with molecular assemblers. The amount of heat that can be rejected from
a given volume is directly related to the amount of surface area, and how conductive,
that material is. The smaller you go, the less surface area. So, if you try to print too quickly, you’ll
end up melting whatever you’re trying to make and possibly the machine trying to make
it too. This might be an option for making self-cooking
food, but proteins aren’t the strongest of structures, so even there you could destroy
what you’re trying to make before it’s done. Santa’s Christmas mince pies are going to
end up a lump of smoldering charcoal. Come to think of it, isn’t it convenient
that those on the naughty list get a lump of coal? Maybe Santa’s been experimenting with a
Santa Claus machine for some time? Of course all of this only matters if you’re
aiming for resolution at the atomic level. Your feedstock material and nozzles can all
act a lot faster if they’re printing at a lower resolution like the microscopic. A cellular printer, as opposed to an atomic
one, is using as its bits or pixels individual biological cells, and each of those contains
trillions of atoms. The problem is, and it’s what make atomic
printers more tempting, is there’s a lot of different cells and you pretty much need
a nozzle for each type. An atomic printer only needs a maximum of
around a hundred since that’s how many atom types we have and typically you would only
be using a handful of types, a molecular printer is much worse. This brings up the issue of specialized manufacturing
which is exactly what these sorts of universal assemblers are supposed to circumvent, a general
or universal creator of objects from some data template. A specialized food printer for instance could
just have nozzles for protein, fat, and carbohydrates, that’s most of what food is, with just some
trace micronutrient vitamins. Of course it isn’t, most food is mostly
water but that at least is very standardized, there’s only one type of water, ignoring
various isotopes that mostly don’t matter chemically. That’s not true of carbs, fat, and proteins. Even ignoring that proteins are size-wise
to a carbohydrate what a skyscraper is to a hut, they all come in a lot of different
types. Even amino acids, which proteins are made
of, come in many forms, even just limiting ourselves to the main ones in biochemistry. For carbs you’ve got monosaccharides and
disaccharides, the most commonly recognized being glucose, fructose, sucrose and lactose. They are, in turn, built into polysaccharides,
made of a dizzying array of combinations of these. Polysaccharide examples are cellulose, which
paper is made from, and starches, like you find in your potato chips. We also don’t eat most of our food raw,
and it’s kind of hard to print a loaf of bread from a big dough vat. You might be able to make a machine that can
print any food on demand, maybe even on a sane time scale, but someone is bound to point
out that a centralized facility or facilities can make this things better and faster and
deliver them to you by drone, in many cases. You could also employ a swarm of drones, each
with specialized tasks, hanging around till they get a call. The notion that building a ton of drones might
take a lot of effort comes to mind but can be instantly discarded, since you are specifically
talking about machines that make stuff. You’re not building ten million specialized
drones for making or delivering a certain class of item, you are building one machine
that makes those drones, or makes the machines that make those drones. When I was a kid the big new rage was Dominos’
promise to have Pizza at your door in 30 minutes or less or it was free. Nowadays we’re seeing this get expedited
with the notion of devices which bake the pizza en route. It’s an alternative scenario for Santa too,
that instead of having some Doctor Who bigger-on-the-inside Sled that can store vast amounts of matter
in TARDIS or Hammerspace, or reindeer able to move at faster than light speeds to return
to the North Pole for loading whenever the cart runs dry, that Santa’s sleigh is a
matter replicator ramjet that sucks in air as it goes and subjects it to a lot of chemistry
and transmutation to create gifts. This is complete with nanoscopic elves inside
doing the work. Gotta have elves, of course! I’ve no idea where they get the energy for
this, but I have long suspected that Rudolph’s glowing red nose actually indicated a radioactive
atomic power source was mounted there, or perhaps a kugelblitz black hole. While this would explain how they obtain the
necessary speeds required for the trip and possibly the time dilation implied, I still
have not determined what purpose the cookies serve in Santa’s refueling operations, it
would seem to imply thus far unknown properties of chocolate chips as an energy source. Further experimentation is probably advised. Now, assuming you get ones of these replicator
methods working, even just the clanking replicator, we have the issue of unemployed elves and
reindeer to deal with, and of course people, already an ongoing problem as technology improves. Something like this is likely to develop gradually
and as we’ve noted before, a surplus of cheap goods tends to largely address the issue
of getting paid so you can buy them. However, what replicators, self-replicators
in particular, introduce is a new paradigm of manufacture. It’s the complexity more than the size or
quantity that is your bottleneck. It’s weird to think of some vast cylinder
space station hull massing Gigatons being cheaper than a single thumb-sized microchip,
but as long as you’ve got quadrillions of tons of raw material lying around the solar
system, it’s really the design phase that’s expensive, not the production itself, and
a big hollow metal cylinder is not exactly complicated. In theory, you release a single clanking replicator
whose additional product is sheet metal on an asteroid and check back in a few months
later. Job complete, it’s not really going to tie
up hundreds of R&D teams to program and modify it to take in rocks, smelt them, roll them
as sheets, and weld them together, whereas you might need hundreds of people to design
the printing template for a single dessert item, or a single microchip, or just one door
in an airlock on that cylinder hab. We’ll discuss this more next week when we
look at using this kind of technology in place of giant generation spaceships, but it’s
a thing to remember, even nowadays thought and complexity are way more expensive than
raw materials or labor. A freeway is much cheaper, pound for pound,
than microchips, let alone the weight of a thumbdrive holding the schematics, data itself
doesn’t really have mass, whereas it obviously has a vast value. The various templates designed for replicators,
be it 3D printers or nanomachine assemblers or Trek-style ones, are likely to be hugely
valuable and a major industry, indeed maybe one of the few industries left if devices
like these are developed. One can imagine the kinds of funds and feuds
involved in intellectual property rights in this regard, from copyright lawsuits to piracy. Now the piracy issue mostly would be a problem
if everybody had a Santa Claus Machine in their basement, as we’re seeing today it
would most likely instead be a lot of somewhat specialized printers at depots with delivery
by drones or mobile printers that got an order and headed your way while making it. However if we did have them so everyone could
have one in their basement, that intellectual property rights issue is secondary to the
terrorism one. Forget printing a nuclear bomb in your basement
from a pirated template or some anarchist cookbook online, you’d have problems getting
the uranium unless your Santa Claus Machine had cheap transmutation too, in which case
you’d probably make antimatter bombs instead. Rather, you have to worry about any random
lunatic printing up a lot of Ricin or hunter seeker kill drones, or worse, self-replicating
hunter-seeker kill drones. I did mention this was the nightmare before
Christmas, didn’t I? This sometimes gets suggested as a Fermi Paradox
solution, and is one of our examples of Suicide Pact Technologies, ones where everybody inevitably
invents it without realizing it will wipe you out or where it’s just so attractive
or easy that it ends up happening even though you know it’s dangerous. It’s not always in an obvious way either,
we see an interesting example of a civilization collapsing and rebuilding after getting its
hands on a Cornucopia Machine in Charles Stross’s novel “Singularity Sky”, where everything
falls apart simply because nobody needs anyone else at all, and we explored some of the consequences
of that in our Post-Scarcity Civilization series in more detail. Great book by the way, deals with a lot the
issues of extreme and god-like artificial intelligence and technological singularities
too. We often classify Artificial Intelligence
as another possible example of a Suicide Pact Technology and it indicates another possible
weird scenario in this case. We probably don’t have to worry about the
prior example because as mentioned, these are unlikely to be something everyone has
in their basement but rather lots of slightly specialized ones working from central depots
for drone delivery, Amazon dialed up to 11, basically. Such a scenario makes it way easier to control
it against dangerous tech and copyright violations because only certain people can approve new
templates and it won’t print anything not on that template, and people don’t go up
and fiddle with it either, drones fly in and pick up the goods for delivery. That’s not completely secure of course,
but better than any lone random lunatic being able to bring about the Apocalypse in their
basement. However, any given template will need some
customizable parameters. I don’t want a shirt, I want a shirt of
a specific material, color, size, shape, etc. One problem we have now with online shopping
is that you’ve got millions of products and usually at least many thousand that would
show up for any common keyword that you need to sort through. In many cases a lot will be standardized to
the person though. I don’t need to see every type of tire or
seat cover available, just the ones that will fit my car’s make and model, and you can
enter those so whenever you search it only offers you ones that fit that. Similarly, you can obviously select clothing
sizes while shopping but even these are fairly general compared to tailoring and it would
be nice if your phone just scanned you regularly and kept your measurements updated when you
went clothes shopping. However it’s not just if it fits but if
it looks good, and that’s not only more complicated but something we often want a
second opinion on, not just ours, so a computer can’t just take our scan and CGI a model
of us in that garment to see if we like it. I mentioned in the episode “Jobs of the
Future” that consulting would be a big industry in the future and often for surprising things,
but it’s easy to imagine one of the earlier uses of AI would be using it to help pick
out the right clothing fashions for people. We always worry about AI going crazy on us,
either Skynet style defense computers or some grey-goo like the Paperclip optimizer, but
there may be more mundane and less threatening routes. Imagine instead an AI built to advise people
on fashion, maybe one with an Asimovian-prescription about harming people with insults when it
tells them that they really do look fat in that garment. Consider also that people are bound to use
it for some opinions on some fairly bizarre clothing since they can do it in private. That’s a big market requiring endless banks
of supercomputers trying to handle the often illogical fashion tastes of billions of people,
and probably not monitored as closely as a defense computer would be. I could so easily imagine that fashion-computer
going quietly, utterly, and homicidally insane. It reminds me of the Robot Santa Claus from
Futurama, which due to a programming error had much too high standards for who was naughty
and nice and how to deal with that, and went mad and flew around each Christmas attacking
people. Devices like the ones we’ve looked at today
offer us a future where almost anything we could wish for was ours for the asking, just
like a Genie in a Lamp, and just like that, it is always good to remember to be careful
what you wish for… While A Santa Claus machine can make almost
any object you might wish to gift someone, the one thing it can’t give them is knowledge. Knowledge, as we all know, is priceless, and
a love of it is the gift that keeps on giving. Help your loved ones nurture their curiosity
and build problem solving skills in a fun, bite-sized way, with Brilliant. Go to brilliant.org/IsaacArthur and grab a
gift subscription to spark a lifelong love of learning. So we looked at replicating machines today
and next week we will close out our fourth year here at SFIA with a look at how this
technology might help us colonizing the galaxy, in Seeding the Stars, and while that will
finish us out for the year, SFIA will be back the next week to start 2019 up by examining
a different sort of travel to different sorts of worlds, as we take another look at virtual
worlds, simulated realities, and how they might become very real indeed. For alerts when those and other episodes come
out, make sure to subscribe to the channel and hit the notifications bell. And if you enjoyed this episode, and haven’t
bankrupted yourself shopping for presents, we’d appreciate your support on Patreon,
and if you have, you can always hit the like button or share the episode with others. Until next time, thanks for watching, and
Happy Holidays!
