Overcoming inertia


The tremendous accelerations involved in the kind of spaceflight seen on Star Trek would instantly turn the crew to chunky salsa unless there was some kind of heavy-duty protection. Hence, the inertial damping field.
— Star Trek: The Next Generation Technical Manual, page 24.

For a space opera RPG setting I am considering adding inertia manipulation technology. But can one make a self-consistent inertia dampener without breaking conservation laws? What are the physical consequences? How many cool explosions, superweapons, and other tropes can we squeeze out of it? How to avoid the worst problems brought up by the SF community?

What inertia is

As Newton put it, inertia is the resistance of an object to a change in its state of motion. Newton’s force law F=ma is a consequence of the definition of momentum, p=mv (which in a way is more fundamental since it directly ties in with conservation laws). The mass in the formula is the inertial mass. Mass is a measure of how much there is of matter, and we normally multiply it with a hidden constant of 1 to get the inertial mass – this constant is what we will want to mess with.

There are relativistic versions of the laws of motion that handles momentum and inertia for high velocities, where the kinetic energy becomes so large that it starts to add mass to the whole system. This makes the total inertia go up, as seen by an outside observer, and looks like a nice case for inertia-manipulating tech being vaguely possible.

However, Einstein threw a spanner into this: gravity also acts on mass and conveniently does so exactly as much as inertia: gravitational mass (the masses in F=Gm_1m_2/r^2) and inertial mass appear to be equal. At least in my old school physics textbook (early 1980s!) this was presented as a cool unsolved mystery, but it is a consequence of the equivalence principle in general relativity (1907): all test particles accelerate the same way in a gravitational field, and this is only possible if their gravitational mass and inertial mass are proportional to one another.

So, an inertia manipulation technology will have to imply some form of gravity manipulation technology. Which may be fine from my standpoint, since what space opera is complete without antigravity? (In fact, I already had decided to have Alcubierre warp bubble FTL anyway, so gravity manipulation is in.)

Playing with inertia

OK, let’s leave relativity to the side for the time being and just consider the classical mechanics of inertia manipulation. Let us posit that there is a magical field that allows us to dial up or down the proportionality constant for inertial mass: the momentum of a particle will be p=\mu m v, the force law F=\mu m a and the formula for kinetic energy K=(1/2) \mu m v^2. \mu is the effect of the magic field, running from 0<\mu<\infty, with 1 corresponding to it being absent.

I throw a 1 g ping-pong ball at 1 m/s into my inertics device and turn on the field. What happens? Let us assume the field is \mu=1000. Now the momentum and kinetic energy jumps by a factor of 1000 if the velocity remains unchanged. Were I to catch the ball I would have gained 999 times its original kinetic energy: this looks like an excellent perpetual motion machine. Since we do not want that to be possible (a space empire powered by throwing ping-pong balls sounds silly) we must demand that energy is conserved.

Velocity shifting to preserve kinetic energy

Radiation shieldingOne way of doing energy conservation is for the velocity to go down for my heavy ping-pong ball. This means that the new velocity will be v/\sqrt{\mu}. Inertia-increasing fields slow down objects, while inertia-decreasing fields speed them up.


One could have a force-field made of super-high inertia that would slow down incoming projectiles. At first this seems pointless, since once they get through to the other side they speed up and will do the same damage. But we could of course put in a bunch of armour in this field, and have it resist the projectile. The kinetic energy will be the same but it will be a lower velocity collision which means that the strength of the armour has a better chance of stopping it (in fact, as we will see below, we can use superdense armour here too). Consider the difference between being shot with a rifle bullet or being slowly but strongly stabbed by it: in the later case the force can be distributed by a good armour to a vast surface. Definitely a good thing for a space opera.


A spacecraft that wants to get somewhere fast could just project a low \mu field around itself and boost its speed by a huge 1/\sqrt{\mu} factor. Sounds very useful. But now an impacting meteorite will both have an high relative speed, and when it enters the field get that boosted by the same factor again: impacts will happen at velocities increased by a factor of 1/\mu as measured by the ship. So boosting your speed with a factor of a 1000 will give you dust hitting you at speeds a million times higher. Since typical interplanetary dust already moves a few km/s, we are talking about hyperrelativistic impactors. The armour above sounds like a good thing to have…

Note that any inertia-reducing technology is going to improve rockets even if there is no reactionless drive or other shenanigans: you just reduce the inertia of the reaction mass. The rocket equation no longer bites: sure, your ship is mostly massive reaction mass in storage, but to accelerate the ship you just take a measure of that mass, restore its inertia, expel it, and enjoy the huge acceleration as the big engine pushes the overall very low-inertia ship. There is just a snag in this particular case: when restoring the inertia you somehow need to give the mass enough kinetic energy to be at rest in relation to the ship…


This kind of inertics does not make for a great cannon. I can certainly make my projectile speed up a lot in the bore by lowering its inertia, but as soon as it leaves it will slow down. If we assume a given amount of force F accelerating it along the length L bore, it will pick up FL Joules of kinetic energy from the work the cannon does – independent of mass or inertia! The difference may be power: if you can only supply a certain energy per second like in a coilgun, having a slower projectile in the bore is better.


Note that entering and leaving an inertics field will induce stresses. A metal rod entering an inertia-increasing field will have the part in the field moving more slowly, pushing back against the not slowed part (yet another plus for the armour!). When leaving the field the lighter part outside will pull away strongly.

Another effect of shifting velocities is that gases behave differently. At first it looks like changing speeds would change temperature (since we tend to think of the temperature of a gas as how fast the molecules are bouncing around), but actually the kinetic temperature of a gas depends on (you guessed it) the average kinetic energy. So that doesn’t change at all. However, the speed of sound should scale as \propto 1/\sqrt{\mu}: it becomes far higher in the inertia-dampening field, producing helium-voice like effects. Air molecules inside an inertia-decreasing field would tend to leave more quickly than outside air would enter, producing a pressure difference.

Momentum conservation is a headache

Atlas 6Changing the velocity so that energy is conserved unfortunately has a drawback: momentum is not conserved! I throw a heavy object at my inertics machine at velocity v, momentum mv and energy (1/2)mv^2, it reduces is inertia and increases the speed to v/\sqrt{\mu}, keeps the kinetic energy at (1/2)mv^2, and the momentum is now mv/\sqrt{\mu}.

What if we assume the momentum change comes from the field or machine? When I hit the mass M machine with an object it experiences a force enough to change its velocity by w=mv(1-1/\sqrt{\mu})/M. When set to increase inertia it is pushed back a bit, potentially moving up to speed (m/M)v. When set to decrease inertia it is pushed forward, starting to move towards the direction the object impacted from. In fact, it can get arbitrarily large velocities by reducing \mu close to 0.

This sounds odd. Demanding momentum and energy conservation requires mv = mv/\sqrt{\mu} + Mw (giving the above formula) and mv^2 = \mu m(v/\sqrt{\mu})^2 + Mw^2, which insists that w=0. Clearly we cannot have both.

I don’t know about you, but I’d rather keep energy conserved. It is more obvious when you cheat about energy conservation.

Still, as Einstein pointed out using 4-vectors, momentum and energy conservation are deeply entangled – one reason inertics isn’t terribly likely in the real world is that they cannot be separated. We could of course try to conserve 4-momentum ((E/c,\gamma \mu m v_x, \gamma \mu m v_y, \gamma \mu m v_z)), which would look like changing both energy and normal momentum at the same time.

Energy gain/loss to preserve momentum

Buffer stopsWhat about just retaining the normal momentum rather than the kinetic energy? The new velocity would be v/\mu, the new kinetic energy would be K_1=(1/2) \mu m (v/\mu)^2 = (1/2) mv^2 / \mu = K_0/\mu. Just like in the kinetic energy preserving case the object slows down (or speeds up), but more strongly. And there is an energy debt of K_0 (1-1/\mu) that needs to be fixed.

One way of resolving energy conservation is to demand that the change in energy is supplied by the inertia-manipulation device. My ping-pong ball does not change momentum, but requires 0.999 Joule to gain the new kinetic energy. The device has to provide that. When the ball leaves the field there will be a surge of energy the device needs to absorb back. Some nice potential here for things blowing up in dramatic ways, a requirement for any self-respecting space opera.


If I want to accelerate my spaceship in this setting, I would point my momentum vector towards the target, reduce my inertia a lot, and then have to provide a lot of kinetic energy from my inertics devices and power supply (actually, store a lot – the energy is a surplus). At first this sounds like it is just as bad as normal rocketry, but in fact it is awesome: I can convert my electricity directly into velocity without having to lug around a lot of reaction mass! I will even get it back when slowing down, a bit like electric brake regeneration systems.  The rocket equation does not apply beyond getting some initial momentum. In fact, the less velocity I have from the start, the better.

At least in this scheme inertia-reduced reaction mass can be restored to full inertia within the conceptual framework of energy addition/subtraction.

One drawback is that now when I run into interplanetary dust it will drain my batteries as the inertics system needs to give it a lot of kinetic energy (which will then go on harming me!)

Another big problem (pointed out by Erik Max Francis) is that turning energy into kinetic energy gives an energy requirement $latex dK/dt=mva$, which depends on an absolute speed. This requires a privileged reference frame, throwing out relativity theory. Oops (but not unexpected).


Energy addition/depletion makes traditional force-fields somewhat plausible: a projectile hits the field, and we use the inertics to reduce its kinetic energy to something manageable. A rifle bullet has a few thousand Joules of energy, and if you can drain that it will now harmlessly bounce off your normal armour. Presumably shields will be depleted when the ship cannot dissipate or store the incoming kinetic energy fast enough, causing the inertics to overload and then leaving the ship unshielded.


This kind of inertics allows us to accelerate projectiles using the inertics technology, essentially feeding them as much kinetic energy as we want. If you first make your projectile super-heavy, accelerate it strongly, and then normalise the inertia it will now speed away with a huge velocity.


A metal rod entering this kind of field will experience the same type of force as in the kinetic energy respecting model, but here the field generator will also be working on providing energy balance: in a sense it will be acting as a generator/motor. Unfortunately it does not look like it could give a net energy gain by having matter flow through.

Note that this kind of device cannot be simply turned off like the previous one: there has to be an energy accounting as everything returns to \mu=1. The really tricky case is if you are in energy-debt: you have an object of lowered inertia in the field, and cut the power. Now the object needs to get a bunch of kinetic energy from somewhere. Sudden absorption of nearby kinetic energy, freezing stuff nearby? That would break thermodynamics (I could set up a perpetual motion heat engine this way). Leaving the inertia-changed object with the changed inertia? That would mean there could be objects and particles with any effective mass – space might eventually be littered with atoms with altered inertia, becoming part of normal chemistry and physics. No such atoms have ever been found, but maybe that is because alien predecessor civilisations were careful with inertial pollution.

Other approaches

Gravity manipulation

Levitating morris dancersAnother approach is to say that we are manipulating spacetime so that inertial forces are cancelled by a suitable gravity force (or, for purists, that the acceleration due to something gets cancelled by a counter-acceleration due to spacetime curvature that makes the object retain the same relative momentum).

The classic is the “gravitic drive” idea, where the spacecraft generates a gravity field somehow and then free-falls towards the destination. The acceleration can be arbitrarily large but the crew will just experience freefall. Same thing for accelerating projectiles or making force-fields: they just accelerate/decelerate projectiles a lot. Since momentum is conserved there will be recoil.

The force-fields will however be wimpy: essentially it needs to be equivalent to an acceleration bringing the projectile to a stop over a short distance. Given that normal interplanetary velocities are in tens of kilometres per second (escape velocity of Earth, more or less) the gravity field needs to be many, many Gs to work. Consider slowing down a 20 km/s railgun bullet to a stop over a distance of 10 meters: it needs to happen over a millisecond and requires a 20 million m/s^2 deceleration (2.03 megaG).

If we go with energy and momentum conservation we may still need to posit that the inertics/antigravity draws power corresponding to the work it does . Make a wheel turn because of an attracting and repulsing field, and the generator has to pay the work (plus experience a torque). Make a spacecraft go from point A to B, and it needs to pay the potential energy difference, momentum change, and at least temporarily the gain in kinetic energy. And if you demand momentum conservation for a gravitic drive, then you have the drive pulling back with the same “force” as the spacecraft experiences. Note that energy and momentum in general relativity are only locally conserved; at least this kind of drive can handwave some excuse for breaking local momentum conservation by positing that the momentum now resides in an extended gravity field (and maybe gravitational waves).

Unlike the previous kinds of inertics this doesn’t change the properties of matter, so the effects on objects discussed below do not apply.

One problem is edge tidal effects. Somewhere there is going to be a transition zone where there is a field gradient: an object passing through is going to experience some extreme shear forces and likely spaghettify. Conversely, this makes for a nifty weapon ripping apart targets.

One problem with gravity manipulation is that it normally has to occur through gravity, which is both very weak and only has positive charges. Electromagnetic technology works so well because we can play positive and negative charges against each other, getting strong effects without using (very) enormous numbers of electrons. Gravity (and gravitomagnetic effects) normally only occurs due to large mass-energy densities and momenta. So for this to work there better be antigravitons, negative mass, or some other way of making gravity behave differently from vanilla relativity. Inertics can typically handwave something about the Higgs field at least.

Forcefield manipulation

This leaves out the gravity part and just posits that you can place force vectors wherever you want. A bit like Iain M. Banks’ effector beams. No real constraints because it is entirely made-up physics; it is not clear it respects any particular conservation laws.

Other physical effects

Here are some of the nontrivial effects of changing inertia of matter (I will leave out gravity manipulation, which has more obvious effects).

Electromagnetism: beware the blue carrot

It is worth noting that this thought experiment does not affect light and other electromagnetic fields: photons are massless. The overall effect is that they will tend to push around charged objects in the field more or less. A low-inertia electron subjected to a given electric field will accelerate more, a high-inertia electron less. This in turn changes the natural frequencies of many systems: a radio antenna will change tuning depending on the inertia change. A receiver inside the inertics field will experience outside signals as being stronger (if the field decreases inertia) or weaker (if it increases it).

Reducing inertia also increases the Bohr magneton, e\hbar/2 \mu m_e. This means that paramagnetic materials become more strongly affected by magnetic fields, and that ferromagnets are boosted. Conversely, higher inertia reduces magnetic effects.

Changing inertia would likely change atomic spectra (see below) and hence optical properties of many compounds. Many pigments gain their colour from absorption due to conjugated systems (think of carotene or heme) that act as antennas: inertia manipulation will change the absorbed frequencies. Carotene with increased inertia will presumably shift its absorption spectra towards lower frequencies, becoming redder, while lowered inertia causes a green or blue shift. An interesting effect is that the rhodopsin in the eye will also be affected and colour vision will experience the same shift (objects will appear to change colour in regions with a different \mu from the place where the observer is, but not inside their field). Strong enough fields will cause shifts so that absorption and transmission outside the visual range will matter, e.g. infrared or UV becomes visible.

However, the above claim that photons should not be affected by inertia manipulation may not have to hold true. Photons carry momentum, p=\hbar k where k is the wave vector. So we could assume a factor of 1/\sqrt{\mu} or 1/\mu gets in there and the field red/blueshifts photons. This would complicate things a lot, so I will leave analysis to the interested reader. But it would likely make inertics fields visible due to refractive effects.

Chemistry: toxic energy levels, plus a shrink-ray

Projectile warningOne area inertics would mess up is chemistry. Chemistry is basically all about the behaviour of the valence electrons of atoms. Their behaviour depends on their distribution between the atomic orbitals, which in turn depends on the Schrödinger equation for the atomic potential. And this equation has a dependency on the mass of the electron and nucleus.

If we look at hydrogen-like atoms, the main effect is that the energy levels become

E_n = - \mu (M Z^2 e^4/8 \epsilon_0^2 h^2 n^2),

where M=m_e m_p/(m_e+m_p) is the reduced mass. In short, the inertial manipulation field scales the energy levels up and down proportionally. One effect is that it becomes much easier to ionise low-inertia materials, and that materials that are normally held together by ionization bonds (say NaCl salt) may spontaneously decay when in high-inertia fields.

The Bohr radius scales as a_0 \propto 1/\mu: low-inertia atoms become larger. This really messes with materials. Placed in a low-inertia field atoms expand, making objects such as metals inflate. In a high inertia-field, electrons keep closer to the nuclei and objects shrink.

As distances change, the effects of electromagnetic forces also change: internal molecular electric forces, van der Waals forces and things like that change in strength, which will no doubt have effects on biology. Not to mention melting points: reducing the inertia will make many materials melt at far lower temperatures due to larger inter-atomic and inter-molecular distances, increasing it can make room-temperature liquids freeze because they are now more closely packed.

This size change also affects the electron-electron interactions, which among other things shield the nucleus and reduce the effective nuclear charge. The changed energy levels do not strongly affect the structure of the lightest atoms, so they will likely form the same kind of chemical bonds and have the same chemistry. However, heavier atoms such as copper, chromium and palladium already have ordering rules that are slightly off because of the quirks of the energy levels. As the field deviates from 1 we should expect lighter and lighter atoms to get alternative filling patterns and this means they will get different chemistry. Given that copper and chromium are essential for some enzymes, this does not bode well – if copper no longer works in cytochrome oxidase, the respiratory chain will lethally crash.

If we allow permanently inertia-altered particles chemistry can get extremely weird. An inertia-changed electron would orbit in a different way than a normal one, giving the atom it resided in entirely different chemical properties. Each changed electron could have its own individual inertia. Presumably such particles would randomise chemistry where they resided, causing all sorts of odd reactions and compounds not normally seen. The overall effect would likely be pretty toxic, since it would on average tend to catalyze metastable high-energy, low-entropy structures in biochemistry to fall down to lower energy, higher entropy states.

Lowering inertia in many ways looks like heating up things: particles move faster, chemicals diffuse more, and things melt. Given that much of biochemistry is tremendously temperature dependent, this suggests that even slight changes of \mu to 0.99 or 1.01 would be enough to create many of the bad effects of high fever or hypothermia, and a bit more would be directly lethal as proteins denaturate.

Fluids: I need a lie down

Inside a lowered inertia field matter responds more strongly to forces, and this means that fluids flow faster for the same pressure difference. Buoyancy cases stronger convection. For a given velocity, the inertial forces  are reduced compared to the viscosity, lowering the Reynolds number and making flows more laminar. Conversely, enhanced inertia fluids are hard to get to move but at a given speed they will be more turbulent.

This will really mess up the sense of balance and likely blood flow.

Gravity: equivalent exchange

I have ignored the equivalence of inertial and gravitational mass. One way for me to get away with it is to claim that they are still equivalent, since everything occurs within some local region where my inertics field is acting: all objects get their inertial mass multiplied by \mu and this also changes their gravitational mass. The equivalence principle still holds.

What if there is no equivalence principle? I could make 1 kg object and a 1 gram object fall at different accelerations. If I had a massless spring between them it would be extended, and I would gain energy. Beside the work done by gravity to bring down the objects (which I could collect and use to put them back where they started) I would now have extra energy – aha, another perpetual motion machine! So we better stick to the equivalence principle.

Given that boosting inertia makes matter both tend to shrink to denser states and have more gravitational force, an important worldbuilding issue is how far I will let this process go. Using it to help fission or fusion seems fine. Allowing it to squeeze matter into degenerate states or neutronium might be more world-changing. And easy making of black holes is likely incompatible with the survival of civilisation.

[ Still, destroying planets with small black holes is harder than it looks. The traditional “everything gets sucked down into the singularity” scenario is surprisingly slow. If you model it using spherical Bondi accretion you need an Earth-mass black hole to make the sun implode within a year or so, and a 3\cdot 10^{19} kg asteroid mass black hole to implode the Earth. And the extreme luminosity slows things a lot more. A better way may be to use an evaporating black hole to irradiate the solar system instead, or blow up something sending big fragments. ]

Another fun use of inertics is of course to mess up stars directly. This does not work with the energy addition/depletion model, but the velocity change model would allow creating a region of increased inertia where density ramps up: plasma enters the volume and may start descending below the spot. Conversely, reducing inertia may open a channel where it is easier for plasma from the interior to ascend (especially since it would be lighter). Even if one cannot turn this into a black hole or trigger surface fusion, it might enable directed flares as the plasma drags electromagnetic field lines with it.

The probe was invisible on the monitor, but its effects were obvious: titanic volumes of solar plasma were sucked together into a strangely geometric sunspot. Suddenly there was a tiny glint in the middle and a shock-wave: the telemetry screens went blank.

“Seems your doomsday weapon has failed, professor. Mad science clearly has no good concept of proper workmanship.”

“Stay your tongue. This is mad engineering: the energy ran out exactly when I had planned. Just watch.”

Without the probe sucking it together the dense plasma was now wildly expanding. As it expanded it cooled. Beyond a certain point it became too cold to remain plasma: there was a bright flash as the protons and electrons recombined and the vortex became transparent. Suddenly neutral the matter no longer constrained the tortured magnetic field lines and they snapped together at the speed of light. The monitor crashed.

“I really hope there is no civilization in this solar system sensitive to massive electromagnetic pulses” the professor gloated in the dark.


Model Pros Cons
Preserve kinetic energy Nice armour. Fast spacecraft with no energy needs (but weird momentum changes). Interplanetary dust is a problem. Inertics cannons inefficient. Toxic effects on biochemistry.
Preserve momentum Nice classical forcefield. Fast spacecraft with energy demands. Inertics cannons work. Potential for cool explosions due to overloads. Interplanetary dust drains batteries. Extremely weird issues of energy-debts: either breaking thermodynamics or getting altered inertia materials. Toxic effects on biochemistry. Breaks relativity.
Gravity manipulation No toxic chemistry effects. Fast spacecraft with energy demands. Inertics cannons work. Forcefields wimpy. Gravitic drives are iffy due to momentum conservation (and are WMDs). Gravity is more obviously hard to manipulate than inertia. Tidal edge forces.

In both cases where actual inertia is changed inertics fields appear pretty lethal. A brief brush with a weak field will likely just be incapacitating, but prolonged exposure is definitely going to kill. And extreme fields are going to do very nasty stuff to most normal materials – making them expand or contract, melt, change chemical structure and whatnot. Hence spacecraft, cannons and other devices using inertics need to be designed to handle these effects. One might imagine placing the crew compartment in a counter-inertics field keeping \mu=1 while the bulk of the spacecraft is surrounded by other fields. A failure of this counter-inertics field does not just instantly turn the crew into tuna paste, but into blue toxic tuna paste.

Gravity manipulation is cleaner, but this is not necessarily a plus from the cool fiction perspective: sometimes bad side effects are exactly what world-building needs. I love the idea of inertics with potential as an anti-personnel or assassination weapon through its biochemical effects, or “forcefields” being super-dense metal with amplified inertia protecting against high-velocity or beam impact.

The atomic rocket page makes a big deal out of how reactionless propulsion makes space opera destroying weapons of mass destruction (if every tramp freighter can be turned into a relativistic missile, how long is the Imperial Capital going to last?) This is a smaller problem here: being hit by a inertia-reduced freighter hurts less, even when it is very fast (think of being hit by a fast ping-pong ball). Gravity propulsion still enables some nasty relativistic weaponry, and if you spend time adding kinetic energy to your inertia-reduced missile it can become pretty nasty. But even if the reactionless aspect does not trivially produce WMDs inertia manipulation will produce a fair number of other risky possibilities. However, given that even a normal space freighter is a hypervelocity missile, the problem lies more in how to conceptualise a civilisation that regularly handles high-energy objects in the vicinity of centres of civilisation.

Not discussed here are issues of how big the fields can be made. Could we reduce the inertia of an asteroid or planet, sending it careening around? That has some big effects on the setting. Similarly, how small can we make the inertics: do they require a starship to power them, or could we have them in epaulettes? Can they be counteracted by another field?

Inertia-changing devices are really tricky to get to work consistently; most space opera SF using them just conveniently ignores the mess – just like how FTL gives rise to time travel or that talking droids ought to transform the global economy totally.

But it is fun to think through the awkward aspects, since some of them make the world-building more exciting. Plus, I would rather discover them before my players, so I can make official handwaves of why they don’t matter if they are brought up.

How much for that neutron in the window?

Zach Weinersmith asked:

That is a great question. I once came up with the answer “50 tons of neutrons are needed” to a serious problem (you don’t want to know). How cheaply could you get that?

Figuring out roughly how many neutrons there are per kilogram of pure elements is pretty easy. Get their standard atomic weights, A, and subtract the atomic number Z since that is the number of protons: N=A-Z. Now we know how many neutrons there are per atom on average (standard atomic weights include the different isotope weights, weighted by their abundance).

[ Since nucleons (protons and neutrons) are about 1830 times heavier than electrons, we can ignore the electrons for an error on order of 0.05%. There is also a binding energy error, since some of the total atomic mass is because of binding energy between nucleons, which is 0.94% or less. These errors are nothing compared to price uncertainties.]

We know that one nucleon weighs about u=1.660539040\cdot 10^{-27} kg, so the number of nucleons per kilogram is N_{\mathrm{nucl}} \approx 1/(Au) and the number of neutrons per kilo is N_n \approx N_{\mathrm{nucl}}(N/A). This ranges from 7.5\cdot 10^{25} for helium down to 1.2\cdot 10^{24} for Oganesson. Hydrogen just has 4.7\cdot 10^{24} neutrons per kilogram, despite having 5.97\cdot 10^{26} nucleons per kilogram – there isn’t that much deuterium and tritium around to contribute neutrons.

Now, the price of elements is badly defined. I can get a kilogram of coal much cheaper than a kilogram of diamond, and ultra-pure elements are very expensive even if the everyday element is cheap. Plus, prices vary. And it is hard to buy plutonium on the open market. Ignoring all that and taking the numbers from Wikipedia (and ignoring the that some values look odd, and some are for compounds, and that the prices are unadjusted for inflation, and that they are lacking for many elements…) we can actually calculate the number of neutrons per dollar:

Neutrons per dollar if one buys one kilogram of the element.
Neutrons per dollar if one buys one kilogram of the element.

And the winner is… aluminium! You can get 8.8\cdot 10^{24} neutrons per dollar from aluminium.

In second place, nitrogen (7.1\cdot 10^{24}) and in third, hydrogen (6.8\cdot 10^{24})! Hydrogen may be very neutron-poor, but since it is rather cheap and you get lots of nucleons per kilo, this balances the lack.

Given that these prices are dodgy, I would expect an uncertainty on the order of a magnitude (at least). So the true winner, given the cheapest actual source of the element, might be hard to find without excruciating price comparisons. But we can be fairly certain it is going to be something with an atomic number less than 25. Uranium is unlikely to be a cheap neutron source in this sense (and just look at poor plutonium!)

So, given that aluminium is 51.8% neutrons by weight I need 96.5 tons. The current aluminium price is $1,650.00 per ton, so I would have to pay $159,225 for the neutrons in my doomsday weapon – I mean, totally innocuous thought experiment!

The Aestivation hypothesis: popular outline and FAQ

Anders Sandberg & Milan Ćirković

Since putting up a preprint for our paper “That is not dead which can eternal lie: the aestivation hypothesis for resolving Fermi’s paradox” (Journal of the British Interplanetary Society, in press) we have heard some comments and confusion that suggest to us that it would be useful to try to outline and clarify what our idea is, what we think about it, and some of the implications.

Table of contents

The super-short version of the paper

Maybe we are not seeing alien civilizations because they are all rationally “sleeping” in the current early cosmological era, waiting for a remote future when it is more favourable to exploit the resources of the universe. We show that given current observations we can rule out a big chunk of possibilities like this, but not all.

A bit more unpacked explanation

Information processing requires physical resources, not just computers or brains, but energy to run them. There is a thermodynamic cost to perform information processing that is temperature dependent: in principle, running processing become 10 times more efficient if your computer is 10 times colder (measured in Kelvins). Right now the cosmic background radiation makes nearly everything in the universe hotter than 3 Kelvin, but as the universe expands this background temperature will decline exponentially. So if you want to do as much information processing as possible with the energy you have it makes sense to wait. It becomes exponentially better. Eventually the background temperature bottoms out because of horizon radiation in a few trillion years: at this point it no longer makes sense to wait with the computation.

Hence, an advanced civilization may have explored a big chunk of the universe, done what is doable with existing nature, and now mostly have internal “cultural” things to do. These things can be regarded as information processing. If they want to maximize processing they should not do it today but wait until the cold future when they will get tremendously more done (1030 times more!). They should hence aestivate, leaving their domain protected by some automation until they wake up.

If this is correct, there might be old and powerful civilizations around that are hard to observe, not because they are deliberately hiding but because they are inactive for the time being.

However, were this hypothesis true, they would not want to lose their stuff. We should expect to see fewer processes that reduce resources  that could be useful in the far future. In the paper we look at processes that look like they might waste resources: stars converting mass into energy that is lost, stars imploding into black holes, galactic winds losing gas into intergalactic space, galaxy collisions, and galaxy clusters getting separated by the expansion of the universe. Current observations do not seem to indicate anything preventing these processes (and most interventions would be very visible).

Hence, either:

  1. the answer to the Fermi question “where are they?!” is something else (like there being no aliens),
  2. advanced civilizations aestivate but do so with only modest hoards of resources rather than entire superclusters,
  3. they are mostly interested in spreading far and wide since this gives a lot of stuff with a much smaller effort than retaining it.

Necessary assumptions

The aestivation hypothesis makes the following assumptions:

  1. There are civilizations that mature much earlier than humanity. (not too implausible, given that Earth is somewhat late compared to other planets)
  2. These civilizations can expand over sizeable volumes, gaining power over their contents. (we have argued that this is doable)
  3. These civilizations have solved their coordination problems. (otherwise it would be hard to jointly aestivate; assumption likelihood hard to judge)
  4. A civilization can retain control over its volume against other civilizations. (otherwise it would need to actively defend its turf in the present era and cannot aestivate; likelihood hard to judge)
  5. The fraction of mature civilizations that aestivate is non-zero. (if it is rational at least some will try)
  6. Aestivation is largely invisible. (seems likely, since there would be nearly no energy release)

Have you solved the Fermi question?

We are not claiming we now know the answer to the Fermi question. Rather, we have a way of ruling out some possibilities and a few new possibilities worth looking for (like galaxies with inhibited heavy star formation).

Do you really believe in it?

I (Anders) personally think the likeliest reason we are not seeing aliens is not that they are aestivating, but just that they do not exist or are very far away.

We have an upcoming paper giving some reasons for this belief. The short of it is that we are very uncertain about the probability of life and intelligence given the current state of scientific knowledge. They could be exceedingly low, and this means we have to assign a fairly high credence to the empty universe hypothesis. If that hypothesis is not true, then aestivation is a pretty plausible answer in my personal opinion.

Why write about a hypothesis you do not think is the most likely one? Because we need to cover as much of possibility space as possible, and the aestivation hypothesis is neatly suggested by considerations of the thermodynamics of computation and physical eschatology. We have been looking at other unlikely Fermi hypotheses like the berserker hypothesis to see if we can give good constraints on them (in that case, our existence plus some ecological instability problems make berzerkers unlikely).

What is the point?

Understanding the potential and limits of intelligence in the universe tells us things about our own chances and potential future.

At the very least, this paper shows what a future advanced human-derived civilization may try to achieve, and some of the ultimate limits on far-future information processing. It gives some new numbers to feed into Nick Bostrom’s astronomical waste argument for working very hard on reducing existential risk in the present: the potential future is huge.

In regards to alien civilizations, the paper maps a part of possibility space, showing what is required for this Fermi paradox explanation to work as an explanation. It helps cut down on the possibilities a fair bit.

What about the Great Filter?

We know there has to be at least one the unlikely step between non-living matter and easily observable technological civilizations (“the Great Filter”), otherwise the sky would be full of them. If it is an early filter (life or intelligence is rare) we may be fairly alone but our future is open; were the filter a later step, we should expect to be doomed.

The aestivation hypothesis doesn’t tell us much about the filter. It allows explaining away the quiet sky as evidence for absence of aliens, so without knowing if it is true or not we do not learn anything from the silence. The lack of megascale engineering is evidence against certain kinds of alien goals and activities, but rather weak evidence.

Meaning of life

Depending on what you are trying to achieve, different long-term strategies make sense. This is another way SETI may tell us something interesting about the Big Questions by showing what advanced species are doing (or not):

If the ultimate value you aim for is local such as having as many happy minds as possible, then you want to spread very far and wide, even though the galaxy clusters you have settled will eventually drift apart and be forever separated. The total value doesn’t depend on all those happy minds talking to each other. Here the total amount of value is presumably proportional to the amount of stuff you have gathered times how long it can produce valuable thoughts. Aestivation makes sense, and you want to spread far and wide before doing it.

If the ultimate value you aim for is nonlocal, such as having your civilization produce the deepest possible philosophy, then all parts need to stay in touch with each other. This means that expanding outside a gravitationally bound supercluster is pointless: your expansion will halt at this point. We can be fairly certain there are no advanced civilizations trying to scrape together larger superclusters since it would be very visible.

If the ultimate value you aim for is finite, then at some point you may be done: you have made the perfect artwork or played all the possible chess games. Such a civilization only needs resources enough to achieve the goal, and then presumably will shut down. If the goal is small it might do this without aestivating, while if it is large it may aestivate with a finite hoard.

If the ultimate goal is modest, like enjoying your planetary utopia, then you will not affect the large-scale universe (although launching intergalactic colonization may still be good for security, leading to a nonlocal instrumental goal). Modest civilizations do not affect the overall fate of the universe.

Can we test it?

Yes! The obvious way is to carefully look for odd processes keeping the universe from losing potentially useful raw materials. The suggestions in the paper give some ideas, but there are doubtless other things to look for.

Also, aestivators would protect themselves from late-evolving species that could steal their stuff. If we were to start building self-replicating von Neumann probes in the future, if there are aestivations around they better stop us. This hypothesis test may of course be rather dangerous…

Isn’t there more to life than information processing?

Information is “a difference that makes a difference”: information processing is just going from one distinguishable state to another in a meaningful way. This covers not just computing with numbers and text, but having one brain state follow another, doing economic transactions, and creating art. Falling in love means that a mind goes from one state to another in a very complex way. Maybe the important subjective aspect is something very different from states of brain, but unless you think that it is possible to fall in love without having the brain change state there will be an information processing element to it. And that information processing is bound by the laws of thermodynamics.

Some theories of value place importance on how or that something is done rather than the consequences or intentions (which can be viewed as information states): maybe a perfect Zen action holds value on its own. If the start and end state are the same, then an infinite amount of such actions can be done and an equal amount of value achieved – yet there is no way of telling if they have ever happened, since there will not be a memory of them occurring.

In short, information processing is something we instrumentally need for the mental or practical activities that truly matter.


Like hibernate, but through summer (latin aestus=heat, aestivate=spend the summer). Hibernate (latin hibernus=wintry) is more common, but since this is about avoiding heat we choose the slightly rarer term.

Can’t you put your computer in a fridge?

Yes, it is possible to cool below 3 K. But you need to do work to achieve it, spending precious energy on the cooling. If you want your computing done *now* and do not care about the total amount of computing, this is fine. But if you want as much computing as possible, then fridges are going to waste some of your energy.

There are some cool (sorry) possibilities by using very large black holes as heat sinks, since their temperature would be lower than the background radiation. But this will only last for a few hundred billion years, then the background will be cooler.

Does computation costs have to be temperature dependent?

The short answer is no, but we do not think this matters for our conclusion.

The irreducible energy cost of computation is due to the Landauer limit (this limit or principle has also been ascribed to Brillouin, Shannon, von Neumann and many others): to erase one bit of information you need to pay an energy cost equal to kT\ln(2) or more. Otherwise you could cheat the second law of thermodynamics.

However, logically reversible computation can work without paying this by never erasing information. The problem is of course that eventually memory runs out, but Bennett showed that one can then “un-compute” the computation by running it backwards, removing the garbage. The problem is that reversible computation needs to run very close to the average energy of the system (taking a long time) and that error correction is irreversible and temperature dependent. Same thing is true for quantum computation.

If one has a pool of negentropy, that is, something ordered that can be randomized, then one can “pay” for bit erasure using this pool until it runs out. This is potentially temperature independent! One can imagine having access to a huge memory full of zero bits. By swapping your garbage bit for a zero, you can potentially run computations without paying an energy cost (if the swapping is free): it has essentially zero temperature.

If there are natural negentropy pools aestivation is pointless: advanced civilizations would be dumping their entropy there in the present. But as far as we know, there are no such pools. We can make them by ordering matter or energy, but that has a work cost that depends on temperature (or using yet another pool of negentropy).

Space-time as a resource?

Maybe the flatness of space-time is the ultimate negentropy pool, and by wrinkling it up we can get rid of entropy: this is in a sense how the universe has become so complex thanks to matter lumping together. The total entropy due to black holes dwarfs the entropy of normal matter by several orders of magnitude.

Were space-time lumpiness a useful resource we should expect advanced civilizations to dump matter into black holes on a vast scale; this does not seem to be going on.

Lovecraft, wasn’t he, you know… a bit racist?

Yup. Very racist. And fearful of essentially everything in the modern world: globalisation, large societies, changing traditions, technology, and how insights from science make humans look like a small part of the universe rather than the centre of creation. Part of what make his horror stories interesting is that they are horror stories about modernity and the modern world-view. From a modernist perspective these things are not evil in themselves.

His vision of a vast universe inhabited by incomprehensible alien entities far outside the range of current humanity does fit in with Dysonian SETI and transhumanism: we should not assume we are at the pinnacle of power and understanding, we can look for signs that there are far more advanced civilizations out there (and if there is, we better figure out how to relate to this fact), and we can aspire to become something like them – which of course would have horrified Lovecraft to no end. Poor man.

Håkan’s surface

Here is a minimal surface based on the Weierstrass-Enneper representation f(z)=1, g(z)=\tanh^2(z). Written explicitly as a function from the complex number z to 3-space it is \Re([-\tanh(z)(\mathrm{sech}^2(z)-4)/6,i(6z+\tanh(z)(\mathrm{sech}^2(z)-4))/6,z-\tanh(z)]).

Håkan’s surface, a minimal surface with Weierstrass-Enneper representation f=1,g=tanh(z)^2.

It is based on my old tanh surface, but has a wilder style. It gets helped by the fact that my triangulation in the picture is pretty jagged. On one hand it has two flat ends, but also a infinite number of catenoid openings (only two shown here).

I call it Håkan’s surface, since I came up with it on my dear husband’s birthday. Happy birthday, Håkan!

Why fears of supersizing are misplaced

I am a co-author of the paper “On the Impossibility of Supersized Machines” (together with Ben Garfinkel, Miles Brundage, Daniel Filan, Carrick Flynn, Jelena Luketina, Michael Page, Andrew Snyder-Beattie, and Max Tegmark):

In recent years, a number of prominent computer scientists, along with academics in fields such as philosophy and physics, have lent credence to the notion that machines may one day become as large as humans. Many have further argued that machines could even come to exceed human size by a significant margin. However, there are at least seven distinct arguments that preclude this outcome. We show that it is not only implausible that machines will ever exceed human size, but in fact impossible.

In the spirit of using multiple arguments to bound a risk (so that the failure of single arguments do not decrease the power of the joint argument strongly) we show that there are philosophical reasons (the meaninglessness of “human-level largeness”, the universality of human largeness, the hard problem of largeness), psychological reasons (acting as an error theory based on motivated cognition), conceptual reasons (humans plus machines will be larger) and scientific/mathematical reasons (irreducible complexity, the quantum-Gödel issue) to not believe the possibility of machines larger than humans.

While it is cool to do exploratory engineering to demonstrate what can in principle be built, it is also very reassuring to show there are boundaries of what is possible. That allows us to focus on the (large) space within.


Catastrophizing for not-so-fun and non-profit

T-valuesOren Cass has an article in Foreign Affairs about the problem of climate catastrophizing. It is basically how it becomes driven by motivated reasoning but also drives motivated reasoning in a vicious circle. Regardless of whether he himself has motivated reasoning too, I think the text is relevant beyond the climate domain.

Some of FHI research and reports are mentioned in passing. Their role is mainly in showing that there could be very bright futures or other existential risks, which undercuts the climate catastrophists that he is really criticising:

Several factors may help to explain why catastrophists sometimes view extreme climate change as more likely than other worst cases. Catastrophists confuse expected and extreme forecasts and thus view climate catastrophe as something we know will happen. But while the expected scenarios of manageable climate change derive from an accumulation of scientific evidence, the extreme ones do not. Catastrophists likewise interpret the present-day effects of climate change as the onset of their worst fears, but those effects are no more proof of existential catastrophes to come than is the 2015 Ebola epidemic a sign of a future civilization-destroying pandemic, or Siri of a coming Singularity

I think this is an important point for the existential risk community to be aware of. We are mostly interested in existential risks and global catastrophes that look possible but could be impossible (or avoided), rather than trying to predict risks that are going to happen. We deal in extreme cases that are intrinsically uncertain, and leave the more certain things to others (unless maybe they happen to be very under-researched). Siri gives us some singularity-evidence, but we think it is weak evidence, not proof (a hypothetical AI catastrophist would instead say “so, it begins”).

Confirmation bias is easy to fall for. If you are looking for signs of your favourite disaster emerging you will see them, and presumably loudly point at them in order to forestall the disaster. That suggests extra value in checking what might not be xrisks and shouldn’t be emphasised too much.

Catastrophizing is not very effective

The nuclear disarmament movement also used a lot of catastrophizing, with plenty of archetypal cartoons showing Earth blowing up as a result of nuclear war or commonly claiming it would end humanity. The fact that the likely outcome merely would be mega- or gigadeath and untold suffering was apparently not regarded as rhetorically punchy enough. Ironically, Threads, The Day After or the Charlottesville scenario in Effects of Nuclear War may have been far more effective in driving home the horror and undesirability of nuclear war better, largely by giving a smaller-scale more relateable scenarios. Scope insensitivity, psychic numbing, compassion fade and related effects make catastrophizing a weak, perhaps even counterproductive, tool.

Defending bad ideas

Another take-home message: when arguing for the importance of xrisk we should make sure we do not end up in the stupid loop he describes. If something is the most important thing ever, we better argue for it well and backed up with as much evidence and reason as can possibly be mustered. Turning it all into a game of overcoming cognitive bias through marketing or attributing psychological explanations to opposing views is risky.

The catastrophizing problem for very important risks is related to Janet Radcliffe-Richards’ analysis of what is wrong with political correctness (in an extended sense). A community argues for some high-minded ideal X using some arguments or facts Y. Someone points out a problem with Y. The rational response would be to drop Y and replace it with better arguments or facts Z (or, if it is really bad, drop X). The typical human response is to (implicitly or explicitly) assume that since Y is used to argue for X, then criticising Y is intended to reduce support for X. Since X is good (or at least of central tribal importance) the critic must be evil or at least a tribal enemy – get him! This way bad arguments or unlikely scenarios get embedded in a discourse.

Standard groupthink where people with doubts figure out that they better keep their heads down if they want to remain in the group strengthens the effect, and makes criticism even less common (and hence more salient and out-groupish when it happens).

Reasons to be cheerful?

An interesting detail about the opening: the GCR/Xrisk community seems to be way more optimistic than the climate community as described. I mentioned Warren Ellis little novel Normal earlier on this blog, which is about a mental asylum for futurists affected by looking into the abyss. I suspect he was maybe modelling them on the moody climate people but adding an overlay of other futurist ideas/tropes for the story.

Assuming climate people really are that moody.

An elliptic remark

I recently returned to toying around with circle and sphere inversion fractals, that is, fractal sets that are invariant under inversion in a given set of circles or spheres.

That got me thinking: can you invert points in other things than circles? Of course you can! José L. Ramírez has written a nice overview of inversion in ellipses. Basically a point P is projected to another point P' so that ||P-O||\cdot ||P'-O||=||Q-O||^2 where O is the centre of the ellipse and Q is the point where the ray between O, P', P intersects the ellipse.

In Cartesian coordinates, for an ellipse centered on the origin and with semimajor and minor axes a,b, the inverse point of P=(u,v) is P'=(x,y) where x=\frac{a^2b^2u}{b^2u^2+a^2v^2} and y=\frac{a^2b^2v}{b^2u^2+a^2v^2}. Basically this is a squashed version of the circle formula.

Many of the properties remain the same. Lines passing through the centre of the ellipse are unchanged. Other lines get mapped to ellipses; if they intersect the inversion ellipse the new ellipse also intersect it at those points. Hence tangent lines are mapped to tangent ellipses. Ellipses with parallel axes and equal eccentricities map onto other ellipses (or lines if they intersect the centre of the inversion ellipse). Other conics get turned into cubics; for example a hyperbola gets mapped to a lemniscate. (See also this paper for more examples).

Now, from a fractal standpoint this means that if you have a set of ellipses that are tangent you should expect a fractal passing through their points of tangency. Basically all of the standard circle inversion fractals hence have elliptic counterparts. Here is the result for a ring of 4 or 6 mutually tangent ellipses:

Invariant set fractal (blue) for inversion in the red ellipses. Generated using an IFS algorithm.

Invariant set fractal (blue) for inversion in the red ellipses. Generated using an IFS algorithm.
Invariant set fractal (blue) for inversion in the red ellipses. Generated using an IFS algorithm.

These pictures were generated by taking points in the plane and inverting them with randomly selected ellipses; as the process continues they get attracted to the invariant set (this is basically a standard iterated function system). It also has the known problem of finding the points at the tangencies, since the iteration has to loop consistently between inverting in the two ellipses to get there, but it is likely that a third will be selected at some point.

One approach is to deliberately recurse downward to find the points using a depth first search. We can take look at where each ellipse is mapped by each of the inversions, and since the fractal is inside each of the mapped ellipses we can then continue mapping the chain of mapped ellipses, getting nice bounds on where it is going (as long as everything is shrinking: this is guaranteed as long as it is mappings from the outside to the inside of the generating ellipses, but if they were to overlap things can explode). Doing this for just one step reveals one reason for the quirky shapes above: some of the ellipses get mapped into crescents or pears, adding a lot of bends:

Mappings of the ellipses by their inversions: each of the four main ellipses map the other three to their interior but distort the shape of two of them.
Mappings of the ellipses by their inversions: each of the four main ellipses map the other three to their interior but distort the shape of two of them.

Now, continuing this process makes a nested structure where the invariant set is hidden inside all the other mapped ellipses.

Nested mappings of the ellipses in the chain, bounding the invariant set. Colors are mixtures of the colors of the generating ellipses, with an increase in saturation.
Nested mappings of the ellipses in the chain, bounding the invariant set. Colors are mixtures of the colors of the generating ellipses, with an increase in saturation.

It is still hard to reach the tangent points, but at least now they are easier to detect. They are also numerically tough: most points on the ellipse circumferences are mapped away from them towards the interior of the generating ellipse. Still, if we view the mapped ellipses as uncertainties and shade them in we can get a very pleasing map of the invariant set:

Invariant set of chain of four outer ellipses and a circle tangent to them on the inside.
Invariant set of chain of four outer ellipses and a circle tangent to them on the inside.

Here are a few other nice fractals based on these ideas:

Using a mix of circles and ellipses produces a nice mix of the regularity of the circle-based Apollonian gaskets and the swooshy, Hénon fractal shape the ellipses induce.

Appendix: Matlab code


% center=[-1 -1 2 1; -1 1 1 2; 1 -1 1 2; 1 1 2 1];
% center(:,3:4)=center(:,3:4)*(2/3);
%center=[-1 -1 2 1; -1 1 1 2; 1 -1 1 2; 1 1 2 1; 3 1 1 2; 3 -1 2 1];
% center=[-1 -1 2 1; -1 1 1 2; 1 -1 1 2; 1 1 2 1];
% center(:,3:4)=center(:,3:4)*(2/3);
% center=[center; 0 0 .51 .51];
% egg
% center=[0 0 0.6666 1; 2 2 2 2; -2 2 2 2; -2 -2 2 2; 2 -2 2 2];
% double
%center=[-r 0 r r; r 0 r r; 2 2 2 2; -2 2 2 2; -2 -2 2 2; 2 -2 2 2];
% Double egg
center=[0.3 0 0.3 0.845; -0.3 0 0.3 0.845; 2 2 2 2; -2 2 2 2; -2 -2 2 2; 2 -2 2 2];
M=size(center,1); % number of ellipses
N=100; % points on fill curves
hold on
alpha 0.2
for i=1:M
    for j=1:M
        if (i~=j)
            recurseDown(X,[i j],10,center)


function recurseDown(X,ellword,maxlevel,center)
i=ellword(end); % invert in latest ellipse
% Perform inversion
if (norm(max(X)-min(X))<0.005) return; end
coco=mean([1 1 1; 1 1 1; co(ellword,:)]);
%    plot(X(:,1),X(:,2),'Color',coco)
if (length(ellword)<maxlevel)
    for j=1:size(center,1)
        if (j~=i)
            recurseDown(X,[ellword j],maxlevel,center)

The frightening infinite spaces: apeirophobia

Bobby Azarian writes in The Atlantic about Apeirophobia: The Fear of Eternity. This is the existential vertigo experienced by some when considering everlasting life (typically in a religious context), or just the infinite. Pascal’s Pensées famously touches on the same feeling: “The eternal silence of these infinite spaces frightens me.” For some this is upsetting enough that it actually count as a specific phobia, although in most cases it seems to be more a general unease.

Fearing immortality

Circle of life

I found the concept relevant since yesterday I had a conversation with a philosopher arguing against life extension. Many of her arguments were familiar: they come up again and again if you express a positive view of longevity. It is interesting to notice that many other views do not elicit the same critical response. Suggest a future in space and some think it might be wasteful or impossible, but rarely with the same tenaciousness as life extension. As soon as one rational argument is disproven another one takes its place.

In the past I have usually attributed this to ego defence and maybe terror management. We learn about our mortality when we are young and have to come up with a way of handling it: ignoring it, denying it by assuming eternal hereafters, that we can live on through works or children, various philosophical solutions, concepts of the appropriate shape of our lives, etc. When life extension comes up, this terror management or self image is threatened and people try to defend it – their emotional equilibrium is risked by challenges to the coping strategy (and yes, this is also true for transhumanists who resolve mortality by hoping for radical life extension: there is a lot of motivated thinking going on in defending the imminent breakthroughs against death, too). While “longevity is disturbing to me” is not a good argument it is the motivator for finding arguments that can work in the social context. This is also why no amount of knocking down these arguments actually leads anywhere: the source is a coping strategy, not a rationally consistent position.

However, the apeirophobia essay suggests a different reason some people argue against life extension. They are actually unsettled by indefinite or infinite lives. I do not think everybody who argues has apeirophobia, it is probably a minority fear (and might even be a different take on the fear of death). But it is a somewhat more respectable origin than ego defence.

When I encounter arguments for the greatness of finite and perhaps short spans of life, I often rhetorically ask – especially if the interlocutor is from a religious worldview – if they think people will die in Heaven. It is basically Sappho’s argument (“to die is an evil; for the gods have thus decided. For otherwise they would be dying.”) Of course, this rarely succeeds in convincing anybody but it tends to throw a spanner in the works. However, the apeirophobia essay actually shows that some religious people may have a somewhat consistent fear that eternal life in Heaven isn’t a good thing. I respect that. Of course, I might still ask why God in their worldview insists on being eternal, but even I can see a few easy ways out of that issue (e.g. it is a non-human being not affected by eternity in the same way).


I found infinity on the stairsAs I often have to point out, I do not believe immortality is a thing. We are finite beings in a random universe, and sooner or later our luck runs out. What to aim for is indefinitely long lives, lives that go on (with high probability) until we no longer find them meaningful. But even this tends to trigger apeirophobia. Maybe one reason is the indeterminacy: there is nothing pre-set at all.

Pascal’s worry seem to be not just the infinity of the spaces but also their arbitrariness and how insignificant we are relative to them. The full section of the Pensées:

205: When I consider the short duration of my life, swallowed up in the eternity before and after, the little space which I fill, and even can see, engulfed in the infinite immensity of spaces of which I am ignorant, and which know me not, I am frightened, and am astonished at being here rather than there; for there is no reason why here rather than there, why now rather than then. Who has put me here? By whose order and direction have this place and time been allotted to me? Memoria hospitis unius diei prætereuntis.

206: The eternal silence of these infinite spaces frightens me.

207: How many kingdoms know us not!

208: Why is my knowledge limited? Why my stature? Why my life to one hundred years rather than to a thousand? What reason has nature had for giving me such, and for choosing this number rather than another in the infinity of those from which there is no more reason to choose one than another, trying nothing else?

Pascal is clearly unsettled by infinity and eternity, but in the Pensées he tries to resolve this psychologically: since he trusts God, then eternity must be a good thing even if it is hard to bear. This is a very different position from my interlocutor yesterday, who insisted that it was the warm finitude of a human life that gave life meaning (a view somewhat echoed in Mark O’Connell’s To Be a Machine). To Pascal apeirophobia was just another challenge to become a good Christian, to the mortalist it is actually a correct, value-tracking intuition.

Apeirophobia as a moral intuition

Infinite ShardI have always been sceptical of psychologizing why people hold views. It is sometimes useful for emphatizing with them and for recognising the futility of knocking down arguments that are actually secondary to a core worldview (which it may or may not be appropriate to challenge). But it is easy to make mistaken guesses. Plus, one often ends up in the “sociological fallacy”: thinking that since one can see non-rational reasons people hold a belief then that belief is unjustified or even untrue. As Yudkowsky pointed out, forecasting empirical facts by psychoanalyzing people never works. I also think this applies to values, insofar they are not only about internal mental states: that people with certain characteristics are more likely to think something has a certain value than people without the characteristic only gives us information about the value if that characteristic somehow correlates with being right about that kind of values.

Feeling apeirophobia does not tell us that infinity is bad, just as feeling xenophobia does not tell us that foreigners are bad. Feeling suffering on the other hand does give us direct knowledge that it is intrinsically aversive (it takes a lot of philosophical footwork to construct an argument that suffering is actually OK). Moral or emotional intuitions certainly can motivate us to investigate a topic with better intellectual tools than the vague unease, conservatism or blind hope that started the process. The validity of the results should not depend on the trigger since there is no necessary relation between the feeling and the ethical state of the thing triggering it: much of the debate about “the wisdom of repugnance” is clarifying when we should expect the intuition to overwhelm the actual thinking and when they are actually reliable. I always get very sceptical when somebody claims their intuition comes from a innate sense of what the good is – at least when it differs from mine.

Would people with apeirophobia have a better understanding of the value of infinity than somebody else? I suspect apeirophobes are on average smarter and/or have a higher need for cognition, but this does not imply that they get things right, just that they think more and more deeply about concepts many people are happy to gloss over. There are many smart nonapeirophobes too.

A strong reason to be sceptical of apeirophobic intuitions is that intuitions tend to work well when we have plenty of experience to build them from, either evolutionarily or individually. Human practical physics intuitions are great for everyday objects and speeds, and progressively worsens as we reach relativistic or quantum scales. We do not encounter eternal life at all, and hence we should be very suspicious about the validity of aperirophobia as a truth-tracking innate signal. Rather, it is triggered when we become overwhelmed by the lack of references to infinity in our lived experience or we discover the arbitrarily extreme nature of “infinite issues” (anybody who has not experienced vertigo when they understood uncountable sets?) It is a correct signal that our minds are swimming above an abyss we do not know but it does not tell us what is in this abyss. Maybe it is nice down there? Given our human tendency to look more strongly for downsides and losses than positives we will tend to respond to this uncertainty by imagining diffuse worst case scenario monsters anyway.

Bad eternities

I do not think I have apeirophobia, but I can still see how chilling belief in eternal lives can be. Unsong’s disutility-maximizing Hell is very nasty, but I do not think it exists. I am not worried about Eternal Returns: if you chronologically live forever but actually just experience a finite length loop of experiences again and again then it makes sense to say that your life just that long.

My real worry is quantum immortality: from a subjective point of view one should expect to survive whatever happen in a multiverse situation, since one cannot be aware in those branches where one died. The problem is that the set of nice states to be in is far smaller than the set of possible states, so over time we should expect to end up horribly crippled and damaged yet unable to die. But here the main problem is the suffering and reduction of circumstances, not the endlessness.

There is a problem with endlessness here though: since random events play a decisive role in our experienced life paths it seems that we have little control over where we end up and that whatever we experience in the long run is going to be wholly determined by chance (after all, beyond 10100 years or more we will all have to be a succession of Boltzmann brains). But the problem seems to be more  the pointlessness that emerges from this chance than that it goes on forever: a finite randomised life seems to hold little value, and as Tolstoy put it, maybe we need infinite subjective lives where past acts matter to actually have meaning. I wonder what apeirophobes make of Tolstoy?

Embracing the abyss

XXI: Azathoth PleromaMy recommendation to apeirophobes is not to take Azarian’s advice and put eternity out of mind, but instead to embrace it in a controllable way. Learn set theory and the paradoxes of infinity. And then look at the time interval t=[0, \infty) and realise it can be mapped into the interval [0,1) (e.g. by f(t)=1/(t+1)). From the infinite perspective any finite length of life is equal. But infinite spans can be manipulated too: in a sense they are also all the same. The infinities hide within what we normally think of as finite.

I suspect Pascal would have been delighted if he knew this math. However, to him the essential part was how we turn intellectual meditation into emotional or existential equilibrium:

Let us therefore not look for certainty and stability. Our reason is always deceived by fickle shadows; nothing can fix the finite between the two Infinites, which both enclose and fly from it.

If this be well understood, I think that we shall remain at rest, each in the state wherein nature has placed him. As this sphere which has fallen to us as our lot is always distant from either extreme, what matters it that man should have a little more knowledge of the universe? If he has it, he but gets a little higher. Is he not always infinitely removed from the end, and is not the duration of our life equally removed from eternity, even if it lasts ten years longer?

In comparison with these Infinites all finites are equal, and I see no reason for fixing our imagination on one more than on another. The only
comparison which we make of ourselves to the finite is painful to us.

In the end it is we who make the infinite frightening or the finite painful. We can train ourselves to stop it. We may need very long lives in order to grow to do it well, though.

Calabi-Yau and Hanson’s surfaces

I have a glass cube on my office windowsill containing a slice of a Calabi-Yau manifold, one of Bathsheba Grossman’s wonderful creations. It is an intricate, self-intersecting surface with lots of unexpected symmetries. A visiting friend got me into trying to make my own version of the surface.

First, what is the equation for it? Grossman-Hanson’s explanation is somewhat involved, but basically what we are seeing is a 2D slice through a 6-dimensional manifold in a projective space expressed as the 4D manifold z_1^5+z_2^5=1, where the variables are complex. Hanson shows that this is a kind of complex superquadric in this paper. This leads to the formulae:

z_1(\theta,\xi,k_1)=e^{2\pi i k_1 / n}\cosh(\theta+\xi i)^{2/n}

z_2(\theta,\xi,k_2)=e^{2 \pi i k_2 / n}\sinh(\theta+\xi i)^{2/n}/i

where the k’s run through 0 \leq k \leq (n-1). Each pair k_1,k_2 corresponds to one patch of what is essentially a complex catenoid. This is still a 4D object. To plot it, we plot the points


where \alpha is some suitable angle to tilt the projection into 3-space. Hanson’s explanation is very clear; I originally reverse-engineered the same formula from the code at Ziyi Zhang’s site.

The Hanson n=4 Calabi-Yau manifold projected into 3-space.
The Hanson n=4 Calabi-Yau manifold projected into 3-space.

The result is pretty nifty. It is tricky to see how it hangs together in 2D; rotating it in 3D helps a bit. It is composed of 16 identical patches:

The patches making up the Hanson Calabi-Yau surface.
The patches making up the Hanson Calabi-Yau surface.

The boundary of the patches meet other patches except along two open borders (corresponding to large or small values of \theta): these form the edges of the manifold and strictly speaking I ought to have rendered them to infinity. That would have made it unbounded and somewhat boring to look at: four disks meeting at an angle, with the interesting part hidden inside. By marking the edges we can see that the boundary are four linked wobbly circles:

Boundary of the piece of the Hanson Calabi-Yau manifold displayed.
Boundary of the piece of the Hanson Calabi-Yau manifold displayed.

A surface bounded by a knot or a link is called a Seifert surface. While these surfaces look a lot like minimal surfaces they are not exactly minimal when I estimate the mean curvature (it should be exactly zero); while this could be because of lack of numerical precision I think it is real: while minimal surfaces are Ricci-flat, the converse is not necessarily true.

Changing N produces other surfaces. N=2 is basically a catenoid (tilted and self-intersecting). As N increases it becomes more like a barrel or a pufferfish, with one direction dominated by circular saddle regions, one showing a meshwork of spaces reminiscent of spacefilling minimal surfaces, and one a lot of overlapping “barbs”.

Hanson's Calabi-Yau surface for N=2, N=3, N=5 and N=8.
Hanson’s Calabi-Yau surface for N=2, N=3, N=5 and N=8.

Note that just like for minimal surfaces one can multiply z_1, z_2 by e^{i\omega} to get another surface in an associate family. In this case it circulates the patches along their circles without changing the surface much.

Hanson also notes that by changing the formula to z_1^{n_1}+z_2^{n_2}=1 we can get boundaries that are torus-knot-like. This leads to the formulae:

z_1(\theta,\xi,k_1)=e^{2\pi i k_1 / n_1}\cosh(\theta+\xi i)^{2/n_1}

z_2(\theta,\xi,k_2)=e^{2 \pi i k_2 / n_2}\sinh(\theta+\xi i)^{2/n_2}/i

Knotted surface for n1=4, n2=3.
Knotted surface for n1=4, n2=3.

Appendix: Matlab code

%% Initialization
edge=0; % Mark edge?
coloring=1; % Patch coloring type
s=0.1; % Gridsize
alp=1; ca=cos(alp); sa=sin(alp); % Projection
% Color scheme
tt=2*pi*(1:200)'/200; co=.5+.5*[cos(tt) cos(tt+1) cos(tt+2)];
%% Plot
hold on
for k1=0:(n-1)
for k2=0:(n-1)
if (coloring==0)
switch (coloring)
case 1
C=z1*0+(k1+k2*n); % Color by patch
case 2
case 3
case 4
case 5
case 6
if (edge>0)
view([2 3 1])
set(h,'Color',[1 1 1]*.5)
axis equal
axis vis3d
axis off

The capability caution principle and the principle of maximal awkwardness

ShadowsThe Future of Life Institute discusses the

Capability Caution Principle: There being no consensus, we should avoid strong assumptions regarding upper limits on future AI capabilities.

It is an important meta-principle in careful design to avoid assuming the most reassuring possibility and instead design based on the most awkward possibility.

When inventing a cryptosystem, do not assume that the adversary is stupid and has limited resources: try to make something that can withstand a computationally and intellectually superior adversary. When testing a new explosive, do not assume it will be weak – stand as far away as possible. When trying to improve AI safety, do not assume AI will be stupid or weak, or that whoever implements it will be sane.

Often we think that the conservative choice is the pessimistic choice where nothing works. This is because “not working” is usually the most awkward possibility when building something. If I plan a project I should ensure that I can handle unforeseen delays and that my original plans and pathways have to be scrapped and replaced with something else. But from a safety or social impact perspective the most awkward situation is if something succeeds radically, in the near future, and we have to deal with the consequences.

Assuming the principle of maximal awkwardness is a form of steelmanning and the least convenient possible world.

This is an approach based on potential loss rather than probability. Most AI history tells us that wild dreams rarely, if ever, come true. But were we to get very powerful AI tools tomorrow it is not too hard to foresee a lot of damage and disruption. Even if you do not think the risk is existential you can probably imagine that autonomous hedge funds smarter than human traders, automated engineering in the hands of anybody and scalable automated identity theft could mess up the world system rather strongly. The fact that it might be unlikely is not as important as that the damage would be unacceptable. It is often easy to think that in uncertain cases the burden of proof is on the other party, rather than on the side where a mistaken belief would be dangerous.

As FLI stated it the principle goes both ways: do not assume the limits are super-high either. Maybe there is a complexity scaling making problem-solving systems unable to handle more than 7 things in “working memory” at the same time, limiting how deep their insights could be. Maybe social manipulation is not a tractable task. But this mainly means we should not count on the super-smart AI as a solution to problems (e.g. using one smart system to monitor another smart system). It is not an argument to be complacent.

People often misunderstand uncertainty:

  • Some think that uncertainty implies that non-action is reasonable, or at least action should wait till we know more. This is actually where the precautionary principle is sane: if there is a risk of something bad happening but you are not certain it will happen, you should still try to prevent it from happening or at least monitor what is going on.
  • Obviously some uncertain risks are unlikely enough that they can be ignored by rational people, but you need to have good reasons to think that the risk is actually that unlikely – uncertainty alone does not help.
  • Gaining more information sometimes reduces uncertainty in valuable ways, but the price of information can sometimes be too high, especially when there are intrinsically unknowable factors and noise clouding the situation.
  • Looking at the mean or expected case can be a mistake if there is a long tail of relatively unlikely but terrible possibilities: on the average day your house does not have a fire, but having insurance, a fire alarm and a fire extinguisher is a rational response.
  • Combinations of uncertain factors do not become less uncertain as they are combined (even if you describe them carefully and with scenarios): typically you get broader and heavier-tailed distributions, and should act on the tail risk.

FLI asks the intriguing question of how smart AI can get. I really want to know that too. But it is relatively unimportant for designing AI safety unless the ceiling is shockingly low; it is safer to assume it can be as smart as it wants to. Some AI safety schemes involve smart systems monitoring each other or performing very complex counterfactuals: these do hinge on an assumption of high intelligence (or whatever it takes to accurately model counterfactual worlds). But then the design criteria should be to assume that these things are hard to do well.

Under high uncertainty, assume Murphy’s law holds.

(But remember that good engineering and reasoning can bind Murphy – it is just that you cannot assume somebody else will do it for you.)