by Brian Tomasik
First written: 2006; last edit: 10 Dec. 2015

Summary

There's a small but non-negligible probability that humans or their descendants will create infinitely many new universes in a laboratory. This would cause infinitely many additional instances of the Holocaust, infinitely many acts of torture, and worse. Creating lab universes would be very bad according to several ethical views.

Background on lab universes

Some physical theories predict that it may be possible to create new, "baby" universes out of a small amount of matter. Technical reviews of the topic can be found in Stefano Ansoldi and Eduardo I. Guendelman, "Child Universes in the Laboratory," and Gordon McCabe, "How to Create a Universe." Popular-level introductions include the following:A Swarm of Ancient Stars - GPN-2000-000930

McCabe explained the concept clearly (p. 6):

Now, one of the most intriguing possibilities opened up by inflation, is the possible creation of a universe 'in a laboratory'. Creation in a laboratory is taken to mean the creation of a physical universe, by design, using the 'artificial' means available to an intelligent species. It is the ability of inflation to maintain a constant energy density, in combination with a period of exponential expansion, which is the key to these laboratory creation scenarios. The idea is to use a small amount of matter in the laboratory, and induce it to undergo inflation until its volume is comparable to that of our own observable universe. The energy density of the inflating region remains constant, and because it becomes the energy density of a huge region, the inflating region acquires a huge total (non-gravitational) energy.

Andrei Linde, one of the founders of inflationary cosmology, put it this way (p. 8):

Indeed, one may need to have only a milligram of matter in a vacuum-like exponentially expanding state, and then the process of self-reproduction will create from this matter not one universe but infinitely many!

Another pioneer of inflation is Alan Guth, the subject of a 1987 New York Times article:

PHYSICISTS often probe the workings of nature on a cosmic scale, but Prof. Alan H. Guth and his colleagues at the Massachusetts Institute of Technology may have set themselves the ultimate research goal. They are seeking a mechanism by which humans might create a new universe from scratch.

Outrageous though such a notion may be, Dr. Guth and his collaborators are perfectly serious about their investigation. "Ten years ago, we couldn't even have posed the question of whether a man-made universe would be possible," he said. "But physics has progressed a long way since then, and today we can ask this and related questions in the real hope of finding scientifically testable answers. We are working in a new and exciting environment."

In his 1997 book, The Inflationary Universe (pp. 268-69), Guth wrote:

To put the story in perspective, one should remember that the process of eternal inflation [postulated by the theory of the self-reproducing inflationary universe ...] leads to an exponential increase in the number of pocket universes on time scales as short as 10-37 seconds. Since the time needed for the development of a super-advanced civilization is measured in billions of years or more, there appears to be no chance that laboratory production of universes could compete with the "natural" process of eternal inflation.

On the other hand, a child universe created in a laboratory by a super-advanced civilization would set into motion its own progression of eternal inflation. Could the super-advanced civilization find a way to enhance its efficiency? We may have to wait a few billion years to find out.

Infinite suffering

Starting a chain of eternal inflation in the laboratory would produce infinitely many new universes. But what types of universes would emerge? Suppose we assume -- as do Jaume Garriga and Alex Vilenkin in their 2001 article "Many worlds in one" -- that there are only finitely many possible universe histories of a particular duration (say, 13.7 billion years, the age of our universe); call these "histories" for short. The existence of infinitely many universes needn't, in general, imply the existence of all possible histories. As Alex Vilenkin notes in his 2006 book Many Worlds in One, the sequence 1, 3, 5, 7, ... contains infinitely many integers but doesn't contain all possible integers, and one might imagine an analogous situation for universe histories (p. 114). However, because "the initial conditions at the big bang are set by random quantum processes during inflation" (p. 114), the theory of inflation does imply that lab universes would instantiate all possible histories infinitely many times (with probability one -- see the second Borel-Cantelli lemma). This would, of course, include infinitely many replications of the Holocaust, infinitely many acts of torture, and so on. Indeed, there would be infinitely many universes in which Hitler won World War II, as well as infinitely many universes that would be as close as physically possible to "hell on earth" (or on any other planet). The assumption of finitely many possible histories is not really important. As long as we assume that the probability is greater than zero that suffering will emerge in a random universe, creating infinitely many universes would create infinite amounts of suffering.

Cosmic evil

There are many moral principles suggesting that creating lab universes would be wrong:

  • "Never again": Lab universes would, among other things, contain infinitely many repetitions of the Holocaust. Would your conscience be okay with carrying out the Holocaust infinitely many times?
  • Problem of evil: What kind of a good god would create a world like ours with so much suffering? And yet, if we created lab universes, we would be doing just that -- as well as creating worlds much worse than our own. Think about a person in one of the universes that future humans might create whose skin is being burned off as part of her torture prior to death. In between screams, she asks: "Why God? Why?" What would be our answer to her?
  • Ends don't justify means: Even if future humans want to create lab universes because of the happiness and beauty they would contain, this doesn't justify the necessary co-creation of infinitely many people being tortured.
  • Asymmetric population ethics: It's more wrong to create a life that suffers than fail to create one that's happy.
  • Other: Prioritarianism, non-positive-focused utilitarianism, wrongness of "playing god."

Nonetheless, I am afraid that potential creators of lab universes would fail to heed these concerns. They might view their project as "cool" or "groundbreaking" without thinking hard about the consequences that playing around with physics would have on real organisms. (In a similar way, few people reflect upon the massive amounts of expected suffering in the universe when they learn about cosmology.) I fear that, because potential universe creators would have lived generally happy lives -- never having been brutally tortured, eaten alive, or slaughtered while conscious -- they would be less sensitive to how bad pain can really be. In general, the lives of humans are far better than the lives of almost all other animals, so even if the would-be universe creators deferred the decision as to whether to create lab universes to the volition of humanity as a whole, that the decision might be biased against giving weight to suffering.

How likely is this scenario?

Is universe creation physically possible?

As McCabe notes in his review (p. 6), Edward Farhi and Alan Guth initially proposed universe creation using false-vacuum "bubbles." However, as they noted in their 1987 "An obstacle to creating a universe in the laboratory," creating a child universe would, under plausible assumptions, require an initial singularity. Still, they suggested, this obstacle might be overcome by quantum tunneling into an inflationary state.

As this article explains:

Recasting the problem to include quantum effects makes the [initial] singularity unnecessary, unfortunately, it also causes the bubble to vanish before it can expand.

What must be realized is that including quantum mechanics in such circumstances is never straightforward. It turns out that if a different version of inflation is used, the instability vanishes. The theory of inflation used by the theoreticians is one that is based around monopoles, which are theoretical magnets with a north or a south pole but never both. Monopoles were thought to exist very early in the universe and are used to explain why our universe is not finely tuned. They are extremely heavy particles, which would, with a small extra kick, contain sufficient energy to create a vacuum bubble that is stable and large enough to experience inflation. The new universe will disconnect from our own and continue on its merry way. From our point of view, the child universe will look like a microscopic black hole that emits a bit of Hawking radiation and then vanishes.

The magnetic-monopole approach was suggested by Nobuyuki Sakai et al. in their 2006 paper, "Is it possible to create a universe out of a monopole in the laboratory?" McCabe notes (p. 12):

Magnetic monopoles are predicted to exist by certain unified field theories, and whilst a magnetic monopole has yet to be discovered, a collision between an electron and a positron could, in principle, create a monopole-anti-monopole pair. Monopoles have masses much greater than those of electrons and positrons, however, and the kinetic energies required to create them by such a collision are beyond the capabilities of contemporary particle accelerators. Universe creation in a laboratory therefore remains beyond current technology, but theoretically possible.

According to New Scientist:

Ironically, one of inflation theory's greatest successes was to explain why we have had such difficulty finding these elusive monopoles, despite theoretical predictions that they should exist all around us. Inflation argues that our visible universe grew from a quantum fluctuation so small it contained only one monopole. That particle is out there somewhere, but the odds are against our bumping into it.

So if we aren't likely to run into a natural monopole any time soon, just how will we get our hands on one? Maybe we could make one in the lab, [Willy] Fischler concedes. Colliding an electron with a positron in a particle accelerator could, in principle, create a monopole-antimonopole pair. And, according to Sakai, we could then trigger inflation by crashing other particles onto our new monopole, adding more and more mass to it. [...]

"I think our model is one of the most realistic for creating a universe in the lab because it uses materials that may well already be out there," Sakai says. [Eduardo] Guendelman agrees. "Ours was just a theoretical idea, but they get a similar effect using something that is predicted to exist by well-known theories," he says.

This article includes some current opinions on the overall likelihood of the lab-universe scenario:

Linde calls that idea [of creating lab universes] "extremely speculative" but has been known to ask, not entirely in jest, how we can be sure our universe isn't the tinkering of a physicist from some other universe. [...]

A paradoxical feature of inflationary theory is that a mere hundred-thousandth of a gram of matter would suffice to create an eternal, self-reproducing universe. [...]

We have a lot of matter around. Does that mean we can create a new universe in a lab? Physicists remain divided on such bold speculations. "It is possible in principle," says Vilenkin at Tufts. "It's all up in the air," says Guth at MIT. "It seems a little less likely now than it did 10 years ago. Even if it is possible, though, it's still far from practical." [...]

Just because it might be theoretically possible doesn't make it easy. Researchers would be unable to verify their experiment's success because of the rapid separation of the new universe from our own universe.

Still, creating a universe is not purely an academic question. The closer physicists get to figuring out how to make a universe, the more they understand our own. "It's natural to be passive when dealing with cosmological problems," says Linde. "The universe is a big place, existing on an utterly inhuman scale. But nevertheless we want to be active, not mere observers. If we can create a new universe, however improbable this might seem, then we should know more about it and think about its possible implications."

The technology to set up this ultimate experiment is still generations away. [... However,]

The day may be approaching when physicists start debating not how our universe was created, but how to create a new one to test their predictions.

Would our descendants have reason to create lab universes?

Our descendants would probably not find lab universes directly useful. As Linde notes (p. 23):

Leaving aside the possibility to use the universe as a universal trash compactor, we were hardly able to find any good reason to spend our time and energy for its creation. Indeed, one cannot "pump" energy from the new universe to ours, since this would contradict the energy conservation law. One cannot jump into the new universe, since at the moment of its creation it is microscopically small and extremely dense, and later it decouples from our universe. One even cannot send any information about himself to those people who will live in the new universe. If one tries, so to say, to write down something "on the surface of the universe", then, for the billions of billions years to come, the inhabitants of the new universe will live in a corner of one letter. This is a consequence of a general rule: All local properties of the universe after inflation do not depend on initial conditions at the moment of its formation. Very soon it becomes absolutely flat, homogeneous and isotropic, and any original message "imprinted" on the universe becomes unreadable.

The only way to communicate with the inhabitants of a new universe would be to encode a message into its physical constants (p. 24).

Despite this, I imagine that many people would have a certain fascination with the idea of creating universes. Linde remarked somewhat sarcastically, "So, what's to stop us from creating a universe in a lab? We would be like gods!" Brian Greene said in his interview with NPR, "I might have a little trouble resisting this possibility [of creating a universe]. Just because it's so curious, this idea that because of your volitional act, you are creating a universe that could give rise, perhaps, to things we see around us." This author said, "I am suitably impressed that the Japanese may be willing to try this and eagerly await the results," and one of the (tongue-in-cheek) comments on that page reads, "Once they've got the Device working, I'll start up my company that sells universes to people. All the people willing to pay to have a star named after them should be willing to pay at least three times as much for full ownership of a universe!" And from the comment on the New Scientist piece: "I wanna be a god though!" See also the range of responses to the question, "If you had the ability to create a new universe and observe its development would you exercise it?" Of course, for lab universes, observing their future development wouldn't be an option.

On Mechanical Turk, I asked ~100 people if they'd favor creating lab universes. 37 said "yes", and 62 said "no". We should do further surveys to make these results more robust.

In the above paragraphs I've been focusing on how current human psychology might regard lab universes, but it's likely that advanced Earth-originating intelligence won't have current human psychology. How would other types of goal systems view lab universes? Many agents that are smart enough to avoid wireheading will work toward goals based on models of how reality is changed by their actions. Since lab universes would be part of such a reality if lab universes are possible, it seems as though many superintelligent agents would at least consider how their goals would be advanced or inhibited within newly created universes. This is true for most of the simplest utility functions one can imagine, including the notional paperclip maximizer. That said, it's also plausible that some utility functions would focus on their own universe. For instance, if we were to codify the current values of the United States military into an artificial intelligence (AI), those values would plausibly be indifferent to other universes and would aim only to maintain US hegemony throughout the galaxy. Or maybe the AI would realize that the US also exists in some of the universes that it could create, which might either inspire it to create lab universes or might inspire it to refrain from creating lab universes because in some of those alternate universes, the US loses its hegemony. Anyway, this is all speculation for now and deserves further exploration.

Manipulating constants?

According to Linde (p. 9):

if chaotic inflation starts at a sufficiently large energy density, then it goes forever, creating new and new inflationary domains. These domains contain matter in all possible "phase states" (or vacuum states), corresponding to all possible minima of the effective potential and all possible types of laws of physics compatible with inflation. However, if inflation starts at a sufficiently low energy density, as is often the case with the universes produced in a laboratory, then no such diversification occurs; inflation at a relatively small energy density does not change the symmetry breaking pattern of the theory and the way of compactification of space-time. [...] Hopefully, one may [set the laws of physics in the new universes] by choosing a proper combination of temperature, pressure and external fields, which would lead to creation of the universe in a desirable phase state.

How feasible is Linde's scenario for manipulating the physical constants of the new universes? Would this be possible under other scenarios of lab-universe creation, e.g., that proposed by Sakai et al.? How much control would the creators have over the specific constants that developed? What is a plausible probability distribution for the types of universes that would emerge conditional on the creators setting constants to particular values?

Does the raw number of universes matter?

Creating lab universes means adding infinitely many new "artificial" universes to an already infinite sea of "natural" universes. How can we quantify the extent of the effect? One approach could be to evaluate the measure of artificial universes within the total. It's not clear that this could be done uniquely, and maybe the question is not even meaningful for some reason.

One can question whether creating lab universes matters at all. Carl Shulman argues:

Baby universes (a pretty remote possibility) require talking about measure in an infinite multiverse, which is a bit tricky for this context, but basically given the existence of infinite baby universes there are infinite instances of suffering and happiness no matter what, all one might do is affect measure. And while there is no satisfactory 'preferred' measure for physicists and cosmologists, on various accounts creating baby universes would leave the measure unchanged, or could be more than offset by other actions affecting measure.

While I have a strong intuition that creating more universes in an already infinite sea of universes does matter, I could imagine my intuitions on this subject changing if it became clear that, say, the idea of counting individual universes was somehow confused and that only the measure was ontologically real.

For example, imagine that the multiverse is an infinite checkerboard, with just black and white squares in all directions. In this case, it seems as though "adding squares" to the checkerboard (whatever that means) wouldn't matter, because the resulting checkerboard would be exactly, mathematically identical to the original. However, if a better analogy is "adding bubbles in an already infinite sea of bubbles", the result after adding bubbles is not necessarily identical, and maybe it's plausible to argue that there are more total bubbles afterward. For instance, if we imagine a bubble over every natural number and then add a second bubble above multiples of 25, it does seem intuitively that we've increased the total number of bubbles (even though, e.g., Cantor's cardinal numbers would consider the cardinality of bubbles the same as before).

Given that there's some risk that humans will be tempted to create lab universes to increase the numerosity of things they value (including copies of themselves), I might prefer for people who aren't focused on reducing suffering to take Shulman's stance, so that they'll have less motivation to work toward lab universes. Of course, as noted above, it may also be that artificial lab-universe creation is inefficient compared with other goal-directed actions that are mirrored across both natural and artificial universes by our near copies.

If the multiverse is finite, then it's more clear that adding new universes does increase total suffering.

Cosmological natural selection

Lee Smolin has proposed a speculative possibility: if black holes create new universes with similar constants as the mother universe, this could yield a kind of cosmological natural selection. The approach has been adapted by Louis Crane and James N. Gardner into hypotheses in which intelligent agents play a central role in conducing to the creation of baby universes.

Smolin's theory looks physically unrealistic for several reasons, including that information apparently can't be transferred to baby universes from a black hole. Hence, these ideas seem very likely to be wrong. But given model uncertainty, we should maintain some probability (say, 0.1%) that they actually are onto something. The ethical implications would be similar as for the lab universes discussed above: Contributing to new universes created by black holes would create possibly infinitely more suffering.

Is it possible to reduce universe production?

Since, as Guth says, new universes are bubbling into existence all the time, is it possible that an advanced civilization could reduce the rate of universe formation rather than increase it via lab universes? I've never heard this proposed, so I'm doubtful that it would be possible even if lab universes are possible, but I don't understand the relevant physics enough to have a strong opinion.

Even if it is possible to reduce universe production, I would guess that most intelligent civilizations would rather increase than decrease universe production, since most people probably think it's good to have more universes, and many non-human AI agents might also want to increase things of value by creating more universes. (One exception would be an AI whose goal is to minimize some quantity.)

Also, unless cosmological natural selection is true, the universe isn't optimized for universe production, so it's probably easier to increase universe production significantly than to decrease it significantly.a This suggests that the expected harm of advanced civilizations increasing universe production may outweigh the expected benefit of those civilizations decreasing universe production, even ignoring my previous points. On the other hand, if cosmological natural selection is true, then universe production will be increased by someone somewhere even if our descendants don't do it, and our efforts to stall that outcome won't have much impact in the long run.

Other universes in a lab

Proposals have been made for creating miniature "universes" using metamaterials. The metamaterials can create conditions equivalent to a 2+1 Minkowski spacetime. Regardless of whether we define these as new universes, I assume they don't have the same ethical import as the lab universes discussed previously because the spacetimes are so small.

Pages That Link Here

"Infinite Suffering" at McCabism.

Acknowledgments

A discussion with Magnus Vinding inspired a few of the paragraphs of this piece.

Footnotes

  1. This point relies on the premise that when a quantity hasn't been optimized, it's generally very far from the optimum on the dimension of interest. For instance, suppose your goal was to maximize paperclips. Earth is very far from the optimum of paperclip production right now. For many criteria you can think of, Earth is far from the optimum along those criteria.  (back)