Permalink https://philpapers.org/rec/TURTMA-5

epistemic status: preliminary write-up on an interesting topic. AI was used only for grammar and style editing.

TL;DR: The laws of physics seem to prevent us from accidentally erasing our own history through time travel. However, this same protection might mean that any civilization advanced enough to build time machines gets destroyed before they can do so – nature's way of maintaining consistency.

Short summary

Does the potential creation of a time machine present an existential risk to our current timeline? Time travel is theoretically possible under general relativity, and there is steady progress (similar to Moore's Law) in developing ideas about how to create time machines with decreasing effort. While time travel may seem like a remote possibility due to its dependence on space travel to black holes, there is a concept of a quantum time machine (suggested by Deutsch in 1991 and further developed by Lloyd in 2011) that would require much less energy to send a few bits back in time and could still cause paradoxes. An artificial superintelligence (ASI) might soon create such a device, especially in scenarios where multiple ASIs compete for control with each other.

However, the catastrophic implications of time machines are largely unexplored in scientific research. Two common, yet contradictory, perspectives exist: either time machines are impossible, rendering risk moot, or Novikov’s self-consistency principle eliminates any chance of timeline alteration. I argue that Novikov’s principle itself can precipitate various catastrophes, for instance, by annihilating civilizations capable of building time machines. This paper classifies the catastrophic risks associated with potential future time machines. The aim is not to declare time machines inherently risky, but to emphasize the necessity of risk assessment given ongoing theoretical advancements in their construction.

Longer Summary

The possible global catastrophic risks from a possible time machine (TM) appear underestimated, largely because of the common belief in its physical impossibility. However, anthropic considerations suggest a greater likelihood of our existence in a world where time travel induced time loops are feasible, as such worlds would possess a higher measure. Consequently, TM is more probable in our reality. Furthermore, evolution might have leveraged time loops for future anticipation, thereby astronomically boosting fitness. This implies our much higher probability of existing in such a world.

Progress in theoretical time machine construction is steady, with increasingly simpler proposals demanding fewer “impossibilities.” Advanced AGI could accelerate this development.

Six primary global catastrophic risks of TM have been identified:

1. Erasing of a timeline. The main risk of the time machine, erasing the current timeline due to a change in the past, seems unlikely because of Novikov’s self-consistency principle (NP). 

• NP links the local non-possibility of a change in the past with the global property that the universe is free of paradoxes. The principle asserts that if an event would cause a paradox or any "change" to the past whatsoever, then the probability of that event is zero.
  • We also suggest the non-cancel principle: an event which has happened, can’t be erased or turned into “unhappened.”
    • In the quantum multiverse, a change in the past will just create a new branch without violating both principles, but consistent time loops will appear in this case too, and thus Novikov’s principle will still work.
  • Dangerous idea: Novikov’s principle works only statistically, as an endpoint, after several unstable loops have happened.
    • That means some timelines can be completely wiped out until a stable circle is found.
      • Or they get low measure, which is also bad.
      • There is a way to erase the current timeline without causing a grandfather paradox, so Novikov’s principle doesn’t provide full protection.

2. Self-fulfilling catastrophic prophecies. Novikov’s principle increases the chances of bad, strange prophecies as loops.

• Fictional example: The Oedipus prophecy that he would kill his father, and all attempts to prevent this prophecy, only ended in its implementation.
  • Only data which cannot change the future can be sent to the past. This creates a stable loop but also ensures the strangeness and incompleteness of the data.

3. Novikov's catastrophe: The simplest way Novikov's Principle can implement itself in preventing time travel and paradoxes is a future catastrophe which stops technological development and prevents the creation of time machines.

• Future catastrophes are more likely. Novikov's principle increases the chances of a data erasing event, that is, of a global catastrophe. Because a catastrophe cleans all causal power from the changes in the past, time travel before a catastrophe doesn't create paradoxes.
• Technological development will never reach the point where time machines are possible.
• No actual time travel is required for such a catastrophe.
• The difference between this and the previous type of catastrophe is that in the previous model, time travel can happen, but only before the catastrophe, and here time travel is prevented by catastrophe completely.

4. Novikov's Bomb: This refers to a time machine designed to be consistent only if a global catastrophe occurs. Such a device could be wielded as a powerful tool for global blackmail.

5. Causal Catastrophe of Spacetime: If Novikov's principle were violated – for instance, by operating two time machines that produce mutually contradictory outputs – it could trigger a spacetime catastrophe, akin to a Big Rip or false vacuum decay.

Minor risks also include:

• Superintelligence powered by the Novikov principle.
• The Butterfly Effect: minor alterations in the past leading to significant current personal value destruction.
• Acquiring hazardous information from the past.
• Invasion or 'SETI-attack' originating from the future.
• The deployment of destructive devices to the past.
• The potential inter-timeline dissemination of diseases, resulting in pandemics in unprepared eras.
• Funny if the first thing an ASI did was rush off to the nearest rotating black hole so it could arrive back on earth before it left (James Miller comment).

Safeguarding measures:

• Regulation and risk analysis must precede the creation of the first time machine.
• A precommitment to abstain from time machine development is crucial to avert a Novikov catastrophe.
• The development of time-locks or barriers is essential to prevent unauthorized temporal displacement.
• Establishment of timekeepers or regulatory agencies is necessary to oversee and manage time-travel activities.
• Time machines themselves could be employed to thwart perilous time travel endeavors.

Combating dangerous AI by altering the past (as depicted in the Terminator movie):

• Forecasting the future to avert most existential risks.
  • Utilizing time travel to escape the heat death of the universe and other undesirable outcomes.
  • Inter-timeline cooperation to accelerate progress.
  • Employing a non-AI-based universal problem solver based on the Novikov self-consistency principle.
  • Avoiding costly errors.

Several potential advantages of the time machine include:

• Resurrection of the dead by reading data from the past and reintroducing extinct species.
• Curing past sufferings by bringing people to the future and replacing them in the past with automata.
• Creating interesting new timelines.
• Explanation of unidentified aerial phenomena (UAP):
  • As strange loops, but not as tourists from the future.
  • Check old cases.
• Curing aging:
  • If the time arrow can be locally reversed.
  • Getting personal stem cells from the past.
• Validation of the truths about the past (e.g., was Christ real?)
• Crime prevention.
• Exploring cosmic events, like the Big Bang or supernovae, up close, offering unparalleled insights into the cosmos.

Introduction

The prevention of existential risks begins with their identification. This often involves examining risks with initially low probability estimates to provide them with proper evaluation. Science fiction can be a valuable tool for this identification process. I've previously explored several underreported x-risks, including SETI-attack, wireheading, chemical contamination, simulation termination, UFOs, and planet explosions. Now, it's time to consider the time machine.

The existential risk posed by a time machine is, at its core, simple: someone creates a time machine, travels to the past, alters it, and our entire timeline vanishes. Many science fiction works treat this scenario as a given; for instance, "Terminator" is not primarily a film about AI risk, but about the erasure of timelines to combat either AI or humanity. One might assume that Novikov's principle (which posits no time travel paradoxes) would prevent such timeline erasure. However, as I will demonstrate, the situation is far more intricate and ominous: Novikov's principle itself can enable catastrophes to halt the creation of time machines or erase timelines in ways that do not trigger paradoxes. Sabina Hossenfelder offers a concise introduction to time machines at this video.

1. Nature of time and time machines

Only “eternalism” allows the existence of a time machine

There are two main theories of time, eternalism and presentism. In the presentism theory, only the now moment of time actually exists, and in the externalism there is a timeless universe, in which past, present and future coexist simultaneously.

The validity of eternalism is a necessary condition (but it is not enough) for time travel, as the past still exists in it.

Paradoxically time travel requires that time doesn’t exist: past and future should exist similarly to the objects in space dimension in the same modal status as now.

Types of time machines

In fiction, time machines can move to the past and to the future, and can send both data and material objects; however, an actual time machine may turn out to be able to do only some of these things.

In table 1, the directions of time travel and the types of objects that could travel are listed for different time machines. Some fictional examples are listed to illustrate the idea.

Table 1. Types of time machines and fictional examples of them

Receiver time machine

As Sabina Hossenfelder noted, real time travel or data transfer between different times may be possible only when we build a signal (or probe) receiver machine, the same way there is no radio transmission without a receiver. This machine will start getting signals from the future, amplify them, and present them to the current observer. The lack of such a receiver can explain why we do not observe time travel. Natural receivers are possible, with black holes or human brains being two candidates.

The most important thing from section 3 is the Non-cancel principle, which states that if some timeline has existed, it can not be canceled by any changes in the past: it can not instantly turn into non-existence. It would require double propagating of causality through time. In MWI the non-cancel principle is certainly true. It is not exactly equivalent to the Novikov principle that time paradoxes are impossible, as we can imagine non-paradoxical cancelation.

{Section 3 “Physics of time machines” is moved at the end for better readability}

2. Principles of time travel which prevent paradoxes

Paradoxes of time travel

Grandfather paradox

It is a self-denying time travel similar to the liar paradox.

Its main property is that it causes complete destruction of all future timeline by a change of the past, so causality is moving twice through time: first, normally, and then, as a canceling wave, which turns all future timeline into non-existence as if it never happened. More on this later.

This is not the end, as the situation now enters into auto-oscillation, because after canceling the timeline the time machine will never appear and no change into the past will ever happen.

This may mean that subjectively inhabitants of such a loop may experience it infinitely many times – even if they were not sent back in time in the time machine and only a few bits of information were sent.

This has enormous anthropic and utilitarian consequences. Anthropic consequences are connected with an overwhelming number of my copies in loops, and utilitarian consequences – with that most moral patients will be in such loops.

Bootstrapping paradox

In the bootstrapping paradox, a piece of information circles infinitely inside a time loop, and it is not clear where its origin is.

For example, a person sends a poem from a book back in time, and it is published in the past. But where did this poem originally come from?

This paradox can be answered by the Novikov self-consistency principle which acts as a super intelligent mind emulator (more on this later): after many unstable loops, one stable loop is selected, and this selection process is a powerful optimization that is similar to intelligence.

A quantum wave (which explores all possible trajectories) effect may help to find stable solutions and thus explain the paradox https://iopscience.iop.org/article/10.1088/1361-6382/ac8198/meta.

Novikov self-consistency principle

The Novikov self-consistency principle (NP) posits that any event leading to a time travel paradox has a zero probability of occurring. This principle links the local impossibility of altering the past with the global requirement that the universe remains free of paradoxes. Essentially, if an event would create a paradox or any "change" to the past, its probability is zero.

As described by Novikov et al., this viewpoint is embodied in "a principle of self-consistency, which states that the only solutions to the laws of physics that can occur locally in the real Universe are those which are globally self-consistent." (Phys. Rev. D 42, 1915).

To uphold this principle, the universe must, in some way, "know" when a particular local arrangement of warp drives and black holes constitutes a time machine transmitting data to the past. The most obvious mechanism for the universe to possess this knowledge is through a quantum multiverse (or at least a wave function), where all potential forward paths are explored, and only those that avoid paradoxes persist.

While the principle aims to avert paradoxes in general relativity based time machines, Everettian quantum mechanics easily prevents such paradoxes through branching.

Note that the Novikov self-consistency principle (NP) not only prevents the grandfather paradox, but also validates the bootstrapping paradox. In some sense, NP performs a search through all possible time loops and selects only stable loops.

But here arises the problem of interpretation of quantum mechanics: how exactly does the Novikov’s principle test "wrong" trajectories?

1. Pilot wave. Does some kind of pilot wave pass through all of them?
2. Many-Worlds Interpretation (MWI). All trajectories are implemented, but only those which have a loop get a probability boost?
3. Objective collapse. Or are unstable timelines actually destroyed after it is found that they cause inconsistency? This variant suggests a "dangerous" possibility: Novikov's principle works only statistically, after several unstable loops have passed and been canceled and we can be in such a timeline.

The danger here means that some timelines can be completely wiped out until a stable circle is found. This requires, however, that the non-cancel principle is false.

Even if Novikov's principle works only for the majority of cases, our timeline seems to be fine based on the non-cancel principle, which doesn't have exceptions.

However, here arises the problem of timelines with very low measure (share of total probability mass). At first glance, a timeline's measure doesn't matter if I am already inside it. But low measure timelines start to have similar measure as Boltzmann brains or some other low probability things, so strange futures are much more likely in them. See also R. Hanson's article about dissolving worlds with low measure. https://philpapers.org/rec/HANDIM

Low-measure timelines also could be more easily simulated by hostile intelligence and minds from them will be "stolen" via indexical uncertainty (more in my article "Curing past sufferings").

So, it is not good to have very low measure, even if it is not equal to complete wiping out.

The good news is that I am unlikely to find myself in a timeline with low measure. But in some situations, the measure of my timeline could become very thin, e.g., after a low probability survival in a quantum immortality event.

Non-cancel principle 

I suggest a similar “non-cancel principle”:

A change of an event in the past will not cancel all future events that were causally caused by that event and that had already happened. 

It means that if something has happened, it can’t be turned into a never-happening state. Double causality waves are impossible. The Novikov principle is mostly an argument against paradoxical time-travel, but the non-cancel principle is an argument against disappearing timelines. Under the non-cancel principle, time travel is possible, but it will create a new timeline after intervention in the past. (See below about how Novikov principle allows changes in timeline if they will not cause paradox.)

If this principle is true, then traveling to the past or sending data to the past is safe.

Even if someone is able to change the past, this will not “cancel” the timeline following this past event, maybe because a new timeline will appear starting from that past change, but the previous one will still be valid.

Non-cancel principle also prevents time travel paradoxes, but it does it in a different way than Novikov’s principle. 

Novikov’s principle prevents time travel events if they cause paradoxes. Non-cancel principle prevents “secondary propagation” of causality from the past event to the future, which will turn an existing thing into the state of never existing things.

Non-cancel principle allows the creation of a new timeline, or a parallel world; Novikov’s principle prevents time travel. This is because Novikov’s principle was designed for general relativity which typically does not assume new multiverse branches.

Non-cancel principle implies that there is some difference between events that happened in the past and just possible events. Is it similar to the eternalism? The past must have some ‘afterglow’ of existence. In common perception of time, it is true. A crime that has happened in the past is different from just a possible crime. 

Non-cancel principle can also be called one-time causality principle. If A has caused B once, no changes in A will affect B after that. 

Non-cancel principle’s applications

If non-cancel principle is false, the universe is finite

If a timeline can disappear as if it has never existed, it means that there are no copies of that timeline anywhere and thus the universe is finite. 

In that case, the Everettian multiverse and inflationary infinite universe are both false.

Theory of time and Youngness paradox

The non-cancel principle also has anthropic consequences. 

If the presentism is true, the SSA (self-sampling assumption in anthropics, that I am randomly selected from all real observers of my reference class) can be only applied to the currently existing observers. 

This causes the Youngness paradox by Gutt. In the infinite inflational universe most observers will be in the youngest possible universe. But we need to define the now moment as cross-section through the whole universe which is difficult because of relativity. 

If the eternalism is true, observers are selected from the whole history of the universe. 

However, if the eternalism is true, but we all assume Everettian branching, there are difficulties:

• If the measure-weight of each branch is declining, then the Youngness paradox appears again.

• Or the total measure-weight of the whole multiverse is increasing and, in that case, the anti-youngness paradox is valid:

• I am located just before the end of the world, in the last second before Big Rip.
  • But it could be compensated by quantum immortality.

If non-cancel principle is false, a sort of perpendicular time is true 

• If the principle is false, it also means that there is a second dimension in time, which is perpendicular to normal time.
  • In the first moment in this dimension an event is happening. In the next step, the whole timeline disappears.

• It also means that the eternalism is valid for normal time dimension, but the perpendicular time dimension is presentist's time, so anything in the past in the perpendicular time disappears as if it never existed.

If non-cancel principle is false, continuous causality is true

• Continuous causality means that a stream of causality is constantly flowing from the past event to the future events, and if there is a change in the past, it will affect the whole future.

  • This, for example, is how gravity works in housing: if we remove a stone from the basement, the whole building will fall.

  • However, causality works only locally in the ball games: only the event when one ball is hitting another is the real moment of causality, and it can’t be canceled by future events.

• Obviously, this reminds us of the presentism and eternalism.
  • The difference is that in eternalism or block-time theory, all blocks are actually existing.
  • In eternalist of causality, the causality continues to support all future.
    • We could imagine that blocks continue to exist even after the causality stream stops, so the theories are not exactly the same.
• Note that epistemic status of these principles are just our ideas about how things should work. 

However, if the principle is not valid, it means that theoretically we can cure past sufferings. 

We could do this by going to the past before the moment of suffering and cancel them. This seems counterintuitive. 

This also means that past crimes could be made non-existing.

Note that if the non-cancel principle is not false, there is still a more speculative way to cure past sufferings via creating multiple copies of the person in pain and using then indexical uncertainty. 

If non-cancel principle is false, timeline can still first continue to eternity

One contradiction in the science fiction about timeline cancellation is that the cancelation is happening exactly in the subjective moment of time when the time machine was sent to the past. This raises the question: with what speed the cancellation propagates through the timeline?

If it goes with the speed of normal causality, it will never reach the now moment, and the timeline will continue as if it was never canceled which makes cancellation meaningless. 

If it is instant, the timeline will never appear. 

Time-loops, measure and anthropics

Time loops in Everettian world’s time travel

Imagine that we send some data back in the Everettian multiverse. Any data which was sent to the past will fork the timeline and create additional Everettian branches. Maybe there will be a very large number of them as there could be a stochastic element in the data incorporation into the timeline. 

There will be a small share of the branches which will be almost the same as the initial branch, and thus in these branches the same data will be sent back again. This subjectively looks like a time loop, as the same chain of events will repeat indefinitely. It is like a fictional situation when you go to the past and become your own grandfather. 

However, it is not a real loop (but a strange attractor), as it is a new Everettian branch every time, and small changes could happen and even accumulate, until the escape happens. 

Example: Let's consider a hypothetical scenario where we have developed a technology in 2023 to send data back in time. We decided to send a simple message: "Invest in XYZ Company in 2020". 

This message is sent back to our past selves in 2019.

In the Everettian multiverse, this action would create a new branch of reality where the 2019 version of ourselves receives this message. In this new branch, we decided to follow the advice and invest in XYZ Company in 2020. In that case, there is no time loop.

However, there is a chance that we didn’t believe that message in 2019, and didn't invest. In that case, we will send the message again in 2023 maybe with small variations hoping to persuade the past self in 2019. After a few cycles, a perfect persuasion message will appear and the cycle forms (but the observer is passing the cycle only once subjectively). 

In the infinite universe there is no difference between past events and possible events

Everything possible will exist actually in an infinite (and chaotic) universe. Thus, any past event is also possible and exists somewhere.

Therefore, in the infinite universe there is no problem with the existence of the “afterglow” which highlights past events that actually happened compared with past variants which did not happen. All events have some existence, but different measure. 

The non-cancel principle also works in the infinite universe: as everything exists, nothing can be canceled. 

Anthropics in time loops

The important property of the infinite time-loop is that it consists of different but very similar cycles. Thus, there could be an astronomically large number of convolutions. Linear time passes just once for an observer, but if there is an infinite time-loop, its measure of existence (roughly equal to the number of copies of an observer) will be very high.

From this follows that the biggest part of the possible observers in the universe are finding themselves in such time loops. 

Thus, we also more likely to be:

• In the world in which time loops (and time travel) are possible,
  • In a such a time loop, and
    • Even if we are in simulation, it could be also in a time loop

We are more likely to live in a world where a time machine is possible, because in this world there will be more observers or at least existing observers will have higher measure. The reason for this is that if there is a time loop, an observers’ timeline will spend an astronomically large number of circles in it, before leaving it because of some stochastic perturbation.    

Such a loop is like a strange attractor for observers’ timeline. Spending a very large number of circles in the loop is equal to having many “copies” or having higher measure. The universes where time-loops are possible, will have an astronomically larger number of observers’ copies. So based on self-sampling-assumption in multiverse [link], we are likely to be in this case. 

Note that only the message is in the cycle, but the observer passes it once. However, the measure of the cycle is so high, that at the beginning of the cycle the observer is much more likely to be sucked into the cycle.

Novikov’s principle doesn’t mean a perfect loop

While Novikov’s principle states that paradoxical timelines with time travel have zero probability, and only consistent closed timelines can exist, such a non-paradoxical timeline may be not circular.

E.g., in a classical thought experiment, a ball can pass through wormholes and hit itself in the past on the right side in the way that the next time it will hit itself on the left side. Therefore, right and left balls strikes will follow each other. 

[spoiler] It was illustrated in Dark TV series, in which three timelines supported each other. [spoiler] But the non-cancel principle is violated in Dark, as eventually the whole trefoil is deleted as if it never happened – or at least stopped.

As there are infinitely many possible non-paradoxical trajectories for a self-hitting ball, such trajectories can follow one another in different combinations. 

This means that my unique timeline can be effectively erased without causing a paradox which will violate Novikov’s principle. So Novikov’s principle does not preserve timelines, it only cares that the timeline will not cause a paradox; non-cancel principle instead preserves the timeline. 

Where are the time travelers?

Hawking famously invited future time travelers to his house – and they never appeared when they were expected. Where are they? It is a time-travel analogue of the Fermi paradox. 

One possible explanation is that time travel is a self-limiting technology which can be implemented only a few times, as loops will prevent too many interventions in them. Also, this doesn’t mean that the total number of time-travelers will be small: but their arrivals to the past will be diluted by branching, as each time-travel creates a new fork.

The too obvious appearance of time travelers would have too strong effects on the future of civilization and will have too strong grandfather-paradox effects. 

Time travel may be possible not for material large objects, but only for some probabilistic data or small nanobots. 

Humanity may go extinct before time travel is invented for other reasons, like wars or AI. 

So, it can be formulated as trilemma:

• Either time travel is not possible,

  • Or its use is very limited because of paradoxes,

    • Or humanity will go extinct before the invention of time travel. 

Note that some think that UFOs are time travelers. 

Chronology Protection Conjecture

Stephen Hawking later proposed the Chronology Protection Conjecture, suggesting that quantum effects would inevitably destroy any time machine at the moment it could first create a paradox (essentially, nature conspires to prevent time travel).  “Do our laws of physics mean all time machines are destined to self-destruct?”

One of the ways to destroy time machines is to prevent civilizations from reaching such a level when they create them. This can be an explanation of the Fermi paradox of time travel – and a big risk for us, as it limits the Kardashev level which we can reach. 

4. Global catastrophic risks from time machines

In some sense, the Terminator movie is not about AI risk, but it is about a time machine’s risk, which can completely erase a future timeline, according to the plot.

Table 2. Catastrophic risks from time machine

1. Erasing timeline

At first glance, a time machine is a perfect example of x-risks: if we travel to the past, we will change it, and this will prevent our existence. Stop building time machines! 

However, this assumes that the past continues to affect me now, so some changes in the past will again propagate via causal lines and affect my existence. This is a much stronger claim than just a time machine. 

This also violates the non-cancel principle discussed in section 2. 

This also implies the grandfather paradox, as if the future disappears, the probe and the fact of time travel will also disappear. That violates Novikov’s principle also, because if a time line can be erased, it would appear to have not happened at all. 

But we can’t be sure about our protection from the changes in the past. So, sending data (or a probe) to the past seems to be a very dangerous idea.

The difference between sending data to the past and traveling to the past is the location of the observer. Sending objects to the past without an observer is equivalent to sending data. And sending data means influencing the casual chain of events, so if nobody observes the data (or its physical consequences), it is not time travel.

In most cases erasing a time line doesn’t mean that humanity goes extinct, as there will be another timeline with humans. However, if the change is in the remote past, before the appearance of contemporary humans, this could prevent the existence of humanity as a whole. Maybe other intelligent beings will evolve, but they will likely have less of what we value. 

Changing the past could be used to prevent some other x-risks, like unaligned AI if it appears only in some timelines. AI is more likely to have access to a time machine, than humans – but in Terminator, both humans and AI got TMs. For AI, the change of timeline is not a big problem, if an AI with the same set of preferences exists in another timeline: the number of “paperclips” will be the same. For humans it is catastrophic, as all currently living people will be un-existed, which is more or less equal to killing.

The timeline disappearance looks differently for different observers. From the grandfather’s friend's point of view this is just a replacement of one possible future to another. From the time traveler’s point of view, it means that his world will disappear, but some other will appear. 

But the worst situation is for a friend of the time traveler who remains in now moments and does not participate in the time travel: his timeline abruptly ends at the moment when the time-traveler starts his journey into the past; it isn’t replaced with some other timeline, but it is erased by propagating wave of canceling from the past (if non-cancel principle is false). 

1a. Replacing timeline or Butterfly effect

The Novikov principle allows multiple futures, if all of them create time machines and send similar data points in the past. This creates a risk that our timeline is not “erased” but merely replaced with a very similar one. The difference between “erased” and “replaced” is more verbal. Even in the case of the “erased timeline”, the planet remains and some process on it continues, so some other timeline continues and life on Earth remains. To erase human civilization, one needs to go a million years in the past and prevent the appearance of hominids. But even in that case some other species of primates may achieve civilization.   

The idea of the Butterfly effect underscores that even small pieces of information sent to the past will cause significant changes in the future. No need to kill Hitler, just one small bit of data will change the whole future. 

However, the whole world will be different – no same people will be born, no known for us art created etc. Surely, it is catastrophic from a personal point of view. Nothing that has a personal value for us will remain. 

Time-line erasure which does not contradict Novikov’s principle

This can be illustrated by the following thought experiment:

1. Bob sends a bomb to the past which kills his grandfather in childhood. 

2. His timeline instantly disappears. 

3. His grandmother marries another man, and they eventually have granddaughter Alice who also sends a bomb to the past at the same moment of space time.

4. The bomb kills Bob’s grandfather again, and a self-confirming loop forms and there is no paradox.  

Here Alice’s timeline forms a loop. The question arises: does this timeline accumulate measure and subjective chances to be in it? If yes, then being in such a loop is almost inevitable. If no, then being Bob and having timeline destruction has equal chances as being Alice. Anthropic probabilities in time-loops seem to be a difficult question. 

2. Time-travel-invariants as bad prophecies

It is possible to imagine “time-travel invariants”: a piece of data which, if it is sent to the past, creates a time loop. From inside such a loop will look like a situation when whatever you do, you can’t change the future. 

The idea is not new. Greek tragedies like Oedipus are built around unescapable prophecies. Oedipus was to kill his father and have sex with his mother; to prevent this outcome the baby was abandoned in childhood by its parents. Eventually, Oedipus came back and did what he was prophesied to do. An important property of time travel invariants is their incompleteness: only part of the data about the future could be sent into the past. https://es.wikipedia.org/wiki/Edipo 

Time-travel invariants tend to be bad and catastrophic prophecies. One of the prevalent scenarios of bad prophecy is that the attempts to escape some catastrophe only increase its chances. 

Time travel invariants are similar to Death in Damascus logical paradox: Death tells a person: “Tomorrow I will wait for you in Damascus, but if you decide to escape to Aleppo, I will wait for you in Aleppo”. So, Death is an absolute predictor of the person’s behavior and any escape attempts are predicted. 

However, a time travel invariant does not prevent the creation of the time machine in the future, so the message can be sent to the past. Thus, the catastrophe can’t be an extinction event, if the TM is artificial.

Another variant of the bad prophecy is the Cassandra case, where the prediction is valid and correct, but nobody is hearing it or acting based on it to prevent the catastrophe.

A prophecy about future catastrophe becomes self-fulfilling if any attempt to escape it actually increases its chances. The risk is that the attempt to escape a time loop only increases its chances (as in the case of Oedipus). 

Maybe a precommitment to ignore any prophecies will protect us from bad prophecies, the same way as precommitment to ignore blackmail lowers chances of being blackmailed.

3. Probability boost for future remote catastrophes by any time travel

The main risk from bad time loops is that catastrophic events get a probability boost. That is because a catastrophe destroys a lot of data, so a bad prophecy can get enough distortions that the senders will eventually send the same piece of data. 

For example, they send a plate in the past with inscription “Don’t create time machines”, but “don’t” part is eventually broken, so the future gets ‘create time machine’ message, but wants to change that and sends back again "don' t create time machine”. Such time loop is more probable in the worlds where plate-breaking events are more likely: wars, earthquakes. Therefore, the time loop is selecting more catastrophic world independently of the type of the message in the loop. 

But why should they send anything at all knowing about potential risks and not expecting any benefits, as benefits will be in another timeline? Loop selection process – testing all possible ways until loop was found – ensures that the senders will have the “urge” to send the message despite knowing all this.

A large catastrophe will destroy all causal connections between now and the future and thus prevent the grandfather paradox: sending data into the past will have no consequence in the future. Catastrophes are the simplest way to prevent time paradoxes, but also some senders have to survive. 

For example, they could be a million years from now and coming from the next and non-human civilization on Earth; their existence requires humanity’s demise but all details of our lives don’t matter, and could be changed without causing a paradox. 

So a very special type of catastrophe is needed: it erases all data, but does not prevent organic life on Earth. This doesn’t mean that all causal connection is cut: the signal needs to be mudded enough so it can’t be decrypted and causes paradoxes. 

Time travel without paradoxes can happen in the time period before such a catastrophe. This could be seen as an argument that we are more likely to be in times before such an erasing catastrophe – if some time travel at all is possible. 

This sounds similar to what was discussed in section (2): self-fulfilling prophecies about catastrophes. The difference is that in the first case, the looped data is a prediction about a catastrophe which will happen in the near term.

In the second case, any time loop is more likely to work if between the moment when the data is sent in the future and the moment in the past when it is received, is present a catastrophic event which erases causal connections between past and future. This prevents paradoxes and fulfills Novikov’s principle. 

Another risk of a bad prophecy is that it is absurd (so its absurdity may be used to explain some absurd things as UAP). The absurdity results from the fact that time loops evolve not for having any meaning in our world, but only for their ability to self-replicate. It is absurd that Oedipus should have sex with his mother no matter what. Absurd events destroy meaning in our world and are thus catastrophic. 

4. Novikov's catastrophe which destroys time machines

Novikov’s principle states according to wiki: “The principle asserts that if an event exists that would cause a paradox or any "change" to the past whatsoever, then the probability of that event is zero. It would thus be impossible to create time paradoxes.” 

The most effective way to prevent the past changes is preventing the creation of time machines, and the most simple, reliable and universal way to do it is a global catastrophe which stops technological development.

In other words, we get a form of Doomsday argument: 

A civilization will never reach a level when it can easily create time machines as they are forbidden. 

Because if a time machine is created it will cause paradoxes in the past, which is prohibited by Novikov’s self-consistency principle. So, it is not the time travel, but the rule against time travel, that causes the catastrophe. This is a meta-level catastrophe. See similar ideas in “Definitely maybe” by Strugatskys’ and explanation of large hadron collider LHC failures via anthropics.

Obviously, there could be other ways to fulfil Novikov’s principle: physical impossibility of time travel, pseudo-random accidents preventing time-machine creation or time-travel-invariants (AKA self-fulfilling prophecies).

Novikov’s catastrophe is not the destruction of the casual structure of the universe discussed below but an attempt of the universe to prevent this type of catastrophe. It is more like an unexpected asteroid strike in a scientific center than sudden dimming of all the stars.

See also https://en.wikipedia.org/wiki/Chronology_protection_conjecture

5. Creating artificial paradoxes via abusing Novikov principle

Novikov bomb: destroying the world by attempt to create time machine

We could imagine a Novikov’s bomb: an idle time machine which would obey Novikov’s principle if started but in which our timeline will be self-contradictory and will have zero probability. 

If the bomb “explodes”, our timeline will be erased. But if you already exist, you will likely fail to create a time machine violating Novikov’s principle for you, as it violates Novikov’s principle. 

Anyway, such a hypothetical doomsday weapon can be used for global blackmail, as all other doomsday weapons. 

Novikov’s doomsday and time loops

It seems that Novikov’s doomsday theory excludes the theory that we already live in a world prone to time loops. And given that time loops are getting a probability boost by anthropics, this makes a time loop world more likely than the world where Novikov’s doomsday is valid. 

However, natural time loops could be “naturally mild” like they would present themselves as a small probability shift for more evolutionary fitness of some beings. They also can have some plausible deniability and “look” (for Novikov principle) like just luck.

Artificial time travel paradoxes could lack this probabilistic arbitrariness and therefore trigger Novikov’s principle to prevent them.  The difference I am speaking of here is a difference between small, e.g. 0.1%, update of reproductive success, which is not causing any paradoxes, as it is based on a positive feedback loop – and the grandfather paradox which is based on negative feedback and is causing oscillations.

Novikov time machine (Outcome pump)

Let’s first introduce the Novikov time machine. 

This time machine sends in the past only one bit of information, which is called “consistency bit”. This bit is 1, if some initial condition is met. The bit is 0 if the condition is not met, and in that case the receiver, located in the past, tries different initial conditions to satisfy this measure. 

For example, a Novikov machine can be used to guess a cipher. 

• If 0 is returned from the future, a random number generator produces a new number and after that it is tested.
• If it is not the correct cipher, 0 returns to the past, and everything starts from the beginning.
• But if the cipher is correct, then 1 is sent to the past, and the loop stops.

If we get 1, we know that we got the correct cypher even before it was tested: we can generate it now, but the test in the future is in times when the time machine will be read, so it is equal to sending the whole cypher to the past.

In other words, sending just one bit to the past allows us to send large messages to the past using Novikov’s principle!

It also allows us to solve complex problems. Even the main principle of the time machine can be found this way. We just need to build a receiver, but not the sender.

Yudkowsky: “This humble-seeming tool can be exploited to achieve nigh-omnipotence; one need merely program the time machine to put the universe into an inconsistent state—send back a “0” if a “1” is recorded as having been received, or vice versa—unless some goal state is achieved. For example, to factor a large composite number, you could generate random numbers using thermal noise or quantum events, and test those random numbers to see if they represent the prime factors of a large composite number; and if not, send back an inconsistent temporal message (Aaronson and Watrous 2009). Let us term such a system an Outcome Pump – it drains probability from some possible futures and pours it into others.” https://intelligence.org/files/ComplexValues.pdf 

He cites “Closed Timelike Curves Make Quantum and Classical Computing Equivalent” by Scott Aaronson.

Two Novikov time machines with contradicting outputs

Two Novikov time machines could be created in a way that the output of one of them will contradict the output of the other. For example, one of them sends a consistency bit in the past only if the other fails to send such a bit.

Obviously, the most likely outcome is failure of one of the machines, so consistency remains. A global catastrophe, like an asteroid impact may be one of the solutions.

Logical catastrophe of space-time

But another outcome of conflicting TMs is a logical catastrophe, something like black hole or self-propelling wave of inconsistencies which crashes the universe. 

If the world disappears immediately, other branches will survive in the multiverse, and something akin to quantum immortality will ensure survival.

But breaking the logic of reality may not kill everyone immediately, but put all minds in an infinite maze of time loops.

Benford et al. wrote about such paradoxes in general, offering a scenario in which two parties are able to send a message two hours into the past:

“The paradoxes of backward-in-time communication are well known. Suppose A and B enter into the following agreement: A will send a message at three o'clock if and only if he does not receive one at one o'clock. B sends a message to reach A at one o'clock immediately on receiving one from A at three o'clock. Then the exchange of messages will take place if and only if it does not take place. This is a genuine paradox, a causal contradiction.” https://en.wikipedia.org/wiki/Tachyonic_antitelephone

See also “Time storms” by Randel. In this book are analyzed reported cases of observations of animalities in the flow of time in normal life, like spontaneous observations of the past. 

0-bit Novikov time machine

While the Novikov time machine described above is assumed to be able to send one bit of information, no actual transfer of information is happening: we can observe it only sending 1, as getting 0 causes a paradox. 

A real 0-bit time machine is the one where the lack of information from the future can be interpreted as some prediction about the future. For example, nobody came to Hawking, when he invited time travelers. He correctly interpreted this as a sign that physical time travel is impossible or at least very difficult. 

The lack of observable time travel gives us an estimation of difficulty and rareness of time travel: an upper bound. This upper bound is depend on a few assumptions: 

• Claimed UFOs are not time travellers, 

  • Reported prophecies and predictions are not real and are not data transfer from the future, 

    • no receiver time machine is needed to get a signal. This is the strongest argument here.

Traveling to the future is equivalent to superintelligence

We can solve P=NP (actually, PSPACE complexity class, see Aaronson) problems if we can get information from the future. We can test as many possible solutions as we want, until we find the working one.

There was an idea to use time travel for computations. Even if it can’t produce AGI, it will increase efficiency of any computations and experimentation. For example, we don’t need to test a new rocket, but instead we will learn the result of the test from the future.

This creates all the risks related to superintelligence, including quick development of new weapons, perfect knowledge about the enemy, and misalignment or wrong interpretation of human commands, if the criterion on which the future tested is wrongly understood. Like in Yudkowsky’s example of rescuing grandmother trapped in the house on fire via Novikov time machine which ends with her flying in the air https://intelligence.org/files/ComplexValues.pdf 

Assume you had a time machine which could send one bit backward in time, in a universe obeying the Novikov self-consistency principle. Suppose your grandmother was trapped in a burning house. How would you use the Outcome Pump to get her out? The obvious approach would be to have a button on the time machine which sends back a consistent bit (if the button is not pressed, the time machine sends back an inconsistent bit), and you only press this button if your grandmother ends up being rescued. This initially seems like the obvious general form for converting the time machine to a genie: (1) you only press the button if you get what you want, and (2) the Novikov self-consistency principle guarantees a timeline in which the button ends up being pressed, so the chain of rules (1) and (2) seems like it should ensure that you always get what you want. In which case (we might imagine) you trip over your own feet and land on the button, pressing it. The Outcome Pump does not really ensure that you always get what you want. It ensures that the button always ends up being pressed. But the user might overlook that minor subtlety after a while, if, the first few times they pressed the button, they got what they wanted.

Outcome pump as a catastrophe generator and preventor

The outcome pump also can be used as a weapon, if it is conditioned to work only in the event of some global catastrophe, like “Enemy country is destroyed”. But it can’t be complete global destruction, as the sender part of the machine has to survive until a catastrophe has happened. This could be overcome by a situation when the catastrophe becomes inevitable, but doesn’t happen yet. For example, a large inbound asteroid is observed.

Alternatively, the Novikov Machine can be conditioned on the survival of a catastrophe. In that case it will be a universal x-risks prevention method. For example, we could select a world where no nuclear war has happened until some moment. 

Time machine as a doomsday machine for blackmail

Attempts to build a time machine can be viewed with suspicion by an opponent, as he may suspect that you are going to the past to change it in the ways that will help you get an advantage from the opponent.

The opponent may think that some small changes in the past will not affect you and your core values and identity, but will damage one’s of the opponent. E.g., humanity will survive, but Skynet creation will be postponed.

Thus, building the time machine will be a very provocative move, even if it is not actually possible, and may cause a nuclear strike on the development site.

Alternatively, even non-existing time machines may be used to blackmail the opponent: “We will erase you from existence!”

Delivery of an ordinary weapon via time machine

A material time machine is also a transportation machine – for example if we move to the past, Earth will change its location and we will be in another place in space. This means that travelling even 1 second in the past will cover 300 km in space (given the Solar system’s speed of rotation around the Galactic center.)

The idea was also in the movie Tenet, when small bombs were placed in a place after a combat, but they move to the past and explode in the past. This obviously completely ignores the self-consistency principle.

Sending dangerous information via time travel: viruses, prophecies, AIs

Prophecies could be sent even without a time machine: false but self-fulfilling prophecies.

If sending data through time is possible, we can get something like SETI-attack from the future: a malicious AI sends its source code through time to us and replicates in this way. Any medium where information can pass, can be exploited by viral behavior, if no protection exists.

Getting alien AI on Earth is dangerous because it could use all our material and resources for self-replication and therefore destroy life on Earth. For example, tile the whole planet with time machines.

Getting dangerous information via time travel

The same way, some past information may be dangerous for us, like DNA of a long-extinct virus. This already happens when we find time-capsules.

Paradox which will be destructive to the whole Universe

Time travel could have unintended consequences to the whole causal structure of the universe. If we are in a simulation, it will halt. The whole universe could terminate (like a false vacuum decay event), or linearity of time could collapse, causing widespread “Time storms”.

In addition, the use of a time machine could lead to a runaway effect, where the time machine is used to repeatedly alter history, leading to further and further changes that could eventually destabilize the fabric of reality.

Time machines replication and shock-wave of time machines in spacetime

Knowledge of how to create TM can propagate to the past, and thus TM becomes a self-replicating unit similar to memes. 

TM may also create copies of objects, if they are sent in the past short time before now. This somehow may result in physical replication of time machines, propagating in space, time and even between different branches of the multiverse. Similar to the idea of intelligence shock wave after superintelligence creation.

This raises questions similar to the Fermi paradox. 

Observers stuck in the loop: no future

Additionally, the use of a time machine could create a "time loop," allowing someone to repeat an event an infinite number of times, thus creating a never-ending cycle. This however requires sending the whole civilization of a group of observers back in time.

Being in a loop prevents someone from having a future. If a whole civilization – or a large group of observers are stuck in the loop, it will be not only personal but also existential risk.

Invasion from the future and dilution of observations of TM

If time travel is possible, bad actors can come from the future. But where are they?

One idea is that each forking of timelines dilutes the future – so any given timeline gets only a few visitors (details of this are not clear – there are as many futures as now moments, so futures remain and thus visitors will not be diluted). But forking can happen only at the moment when future visitors arrive. In that case, new arrival creates a new fork and this explains dilution of observation of arrivals of future travelers.

Viruses from the past and future

Potential spread of diseases across different timelines, leading to pandemics in eras unprepared for them. 

5. Preventing x-risks of time machine

Protection measures against dangerous TM

Here is the list of some ideas:

• Regulation and risk analysis before the first TM is made.
• A precommitment not to create TM in order to prevent Novikov’s catastrophe.
• Development of time-locks or barriers to prevent unauthorized time travel.
• Establishment of time guardians or agencies to monitor and regulate time-travel activities.
• Use TM to prevent dangerous time travel attempts.
• International treaty

Possible benefits of time machines

Several possible benefits of time machine:

• Fighting dangerous AI by changes in the past
  • Predicting future can help escape moist x-risks
  • Time travel can be used to escape heat death of the universe and other bad outcomes
  • Cooperating between timelines will accelerate progress
  • Non-AI-based universal problem solver using Novikov principle
    • Escaping costly mistakes
    • Escaping risks of agential AI
  • Resurrection of the dead by reading data from the past and reintroduce extinct species
  • Curing past sufferings by bringing people to the future and replacing them in the past with automata
  • Creating interesting new time-lines
  • Explanation of UAP
    • As strange loops, but not as tourists from the future
    • Check old cases

Stopping dangerous AI using TM

In the Terminator movie time machine is used to send a person to the past with the goal to stop rogue AI Skynet. This is obviously not very satisfying, as if it succeeds, the whole timeline with Skynet will be canceled, including all people who will live there, and a completely new timeline with new people will appear. If the timeline remains, then Skynet will remain in it too.

Only a perfect utilitarian may agree to replace one timeline with another, as he will not exist in the new timeline. Utilitarians would measure timelines by the number of people who live in them and their happiness. 

Doing extreme good via time travel – sending nanobot to the past

Abram Demsky asked what is the best thing one can do if one has a one-time time machine. My answer:

I will wait until nanotech and advanced AI appear. After that I will send one nanobot to the beginning of the universe. It will secretly replicate and cover all visible universes with its invisible copies – so no paradoxes. It will go inside every person’s brain and upload this person at the moment of death. It will also turn off pain during intense suffering. Only after the time-machine its existence will be revealed. Thus, this will solve the problem of curing past sufferings and achieve resurrection of the dead.

Resurrection of the dead via time-travel

In some sense, time-travel is the most direct way to immortality: go to the past and save the people! In that case, humanity from the future can send a machine which secretly takes the brain of a dead person just seconds before the death and replaces it with a dummy, without leaving any visible traces. This brain then is moved to the future and resurrected.

Or at least scan the brain states and transfer or record in an un-destructive way the information from the brain.

Counterfactual history with Time safety movement

In some counterfactual universe, preventing creation of a time machine could be as popular as AI safety in our universe. They are actually close to each other: AI will find all the ways to create TM and TM may be used as a universal problem solver thus playing the role of AI or helping in its creation. 

3. Possible physics behind time machine

{This section is moved here from above}

Here we will give a very quick overview of the ideas how a TM can be built without going deep and making specific claims about the possibility. This is not a complete list of ideas, it is only for illustrative purposes, and is not intended to prove that TM is possible. The goal of the article is not to prove TM possibility or design a TM, but explore the risks of time travel, which may warrant preemptive ban. 

The same way as with AI, a good idea how to build a time machine could be an infohazard.

General relativity and time travel

Superluminal travel=time machine

It was shown that if faster than light travel is possible, then in another timeframe this will be travel to the past. Allen Everett (not Hugh) showed that using warp drives, it is possible to perform two trips which also include a change in reference frame and return to the moment before the beginning of the trip. See details about this here.

However, recent research suggested a simpler configuration in which a warp drive moves in rotating trajectory:

We can speculate on what a machine able to create the metric of Eq.5 might look like. Suppose some future technology is able to create a device that via some means can produce the bubble curvature, momentarily, and on command. A large number of such devices could be arranged around the ring and pre-synchronized to fire in sequence in such a way as to create a bubble travelling at varying velocities (including superluminal) around the ring. Imagining this all constructed on a cylindrical platform, then spinning the entire platform at high speeds would result in the desired metric. Obviously, this is a gedanken experiment, however it provides a clear physical picture for attaching to the metric and hence a solid basis upon which to discuss the apparent paradoxes that arise from considering time travel in general relativity. https://arxiv.org/pdf/2007.14627.pdf 

General relativity allows the existence of wormholes and time machines based on them

If the exit of a worm hole is moving with near light speed, it will be slower in time relative to the hole’s entrance. 

Moreover, wormholes can connect any two points of space-time. However, traversing them requires infinitely strong materials, so likely only information can be sent. 

There is a theory that, on the quantum Planck level, many small worm holes appear and disappear constantly, similar to virtual particles, and one could be blown into macroscale existence somehow.

An object passing through such a wormhole will eventually appear in the past.

Complex objects like warp drive as time machines

Some solutions to general relativity allow configurations where time flows backwards, described in the Sabina Hossenfelder video. Initially these solutions required objects which are infinitely long, have infinite energy and have infinite density. But the progress has continued to simpler solutions.

1. Infinitely long rotating cylinder, 1936 https://en.wikipedia.org/wiki/Tipler_cylinder 

2. Two infinite and infinitely dense cosmic strings moving against each over, 1991 https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.66.1126 

3. Short strings or rings; but infinite density is still needed, 2017 https://arxiv.org/abs/1708.09505 

4. Rotating warp drive discussed above is even simpler, 2020 https://arxiv.org/pdf/2007.14627.pdf 

We can see that the size, number of impossibilities and energy requirements for time machine proposes are steadily declining.

Rotating black holes and closed time-like curves

See a detailed explanation in Sabina Hossenfelder’s video. A rotating of black hole has two horizons and closed loops are possible between horizons. It is theoretically possible to enter this space and exit it. 

However, where one will appear after such an exit, even if the observer performed time travel, is not clear. It could be the same time, or past, or another universe according to Sabine Hossenfelder.

Very small subatomic black holes can be instruments for data transfer to the past. Note that it appears that the black holes could be used to perform very quick and efficient computations, so future AI may prefer to create small black holes and even completely disappear from our world (John Smart suggested this). Time travel also increases the efficiency of computations.

Alternatively, very large black holes may be more stable and allow passage of large spaceships. 

Quantum theory and time travel

Quantum time machine

I asked o3 to write about current state of research on quantum time machine and it is unlikely that I can write better. See resulting file here. 

It says: "Quantum time machines explore how quantum mechanics might permit time travel-like effects without requiring black holes or exotic matter. Key proposals include David Deutsch’s model, which resolves paradoxes by allowing mixed quantum states to interact with their past selves, and Seth Lloyd’s model, which uses quantum teleportation combined with post-selection to enforce consistency by filtering out paradoxical outcomes. Both models bypass classical causality while preserving logical consistency, albeit at the cost of introducing nonlinearity or other exotic features into quantum theory. While real closed timelike curves remain unobservable, laboratory simulations using entangled photons and post-selection have successfully demonstrated the internal logic of these models, deepening our understanding of quantum causality and time."

Quantum time machine will be likely a small unsurprising installation on someone experimental table with a few lasers and mirrors capable to send a few bits of information in the past for a few milliseconds back with some probability. Maybe not very useful except high frequency trading, but still able to cause time travel paradoxes. 

Relativistic time machine in the Everettian multiverse

We can update the idea of a relativistic (or other) time machine if we combine it with constantly splitting Everettian universe theory. 

In that case, travel to the past will just create one more timeline, which will not affect my current timeline, so there is no existential risk – at the first glance – and no grandfather paradox. (Traveling to the past and killing own grandfather in childhood will prevent a person's existence and thus he will not perform time travel to the past – more below.)

This is a view of D. Deutsch (see article ”Quantum mechanics near closed timelike lines”).

Everettian quantum mechanics allows existence of multiple pasts

If the multiverse is constantly branching, there will be many branches appearing in the past and they will present different variants of the past. 

Interaction between branches is not completely prohibited in some situations like quantum bomb testers. (A quantum experiment in which a bomb is tested dude-or-not without exploding it via quantum interference https://en.wikipedia.org/wiki/Elitzur%E2%80%93Vaidman_bomb_tester)

If we create a TM which can travel or exchange data between branches, some form of time travel is also possible, as it is possible to imagine a variant of Earth, similar to the current planet, but delayed in its existence for some period of time. 

Travel to such delayed branches will look like time travel but there will be no causal effects on us from changes in this “past”.

Quantum entanglement in time is also possible. 

Basic physical laws are time-symmetric and illusion of arrow of time

Most of the equations describing elementary particle behavior are time symmetrical. This may suggest that data can be sent back in time. 

But it all depends on the nature of the arrow of time. 

Retrocausality theory

One of the interpretations of quantum mechanics uses the idea of retrocausality. It is studied by Hew Price, former director of CSER. 

In some sense, Bohm’s pilot wave ‘explores’ possible ways where a particle can send data back in time to the particle telling it where to go.

Other ideas

Tachyons

There is an idea that there are natural particles called tachyons, which are moving back in time.

https://en.wikipedia.org/wiki/Tachyonic_antitelephone

Acausal trade as analogue of time travel

In this case, prediction and trade play the role of a time machine, no new physical effects are needed. Example: Roko Basilisk. But the dangers are obvious: acausal blackmail is not nice. 

We live in a simulation

The simulation’s “script” may define its future. 

The simulation’s past is recorded somewhere on a host computer and could be accessed via some hack (similar to the fictional Akashi Chronicles).

Note that vanilla simulation excludes time travel: it just simulates world in the current state – that is, 3D-structure of the world in the current moment of time. Thus, eternalism is not applicable to simulation: future is not computed in it yet (except the cases where simulation is going according the script). This means that observed time travel is an argument against simulation. Obviously, time travel can be simulated inside simulation to fool inhabitants. 

Perpendicular time

Maybe there is a second arrow of time (similar to revisions of a novel relative to linear time in a plot; therefore, perpendicular time can exist in simulations).

An observer may have complex movements on the time plane (t1, t2).

Reversal of the arrow of time

One idea to reverse the arrow of time is reconstruction of the past state based on the current state using some superintelligence. But maybe it can be done physically.

Time arrow reversal in quantum computers

https://bigthink.com/hard-science/particle-time-travel/

‘The second law of thermodynamics states that order always moves to disorder, which we experience as an arrow of time. Scientists used a quantum computer to show that time travel is theoretically possible by reverting a simulated particle from an entropic to a more orderly state. While Einstein's general theory of relativity permits time travel, the means to achieve it remain improbable in nature.’

Everything is now

An idea is that the real nature of the universe is “holographic”. Time is an illusion. The past coexists with now. Therefore, there is no violation of any physical laws in time travel.

Dust theory

This theory is that observations just follow one from another according to some predicting principle, like Kolmogorov complexity. Thus, there could be correlations between observer-moments which are located far apart in time or space. See https://arxiv.org/pdf/1712.01826.pdf 

6. Natural time machines

Anthropics, natural selection and precognition

Let’s consider the following line of reasoning:

• If precognition (sending data to the past) is physically possible, it will be exploited by biological evolution, as it gives enormous fitness boost.
  • Even if it gives a small probability of survival increase, 1 per cent increase of survival will accumulate.
• As we showed above, sending data back in time creates loops, and such loops have higher “measure” or higher chances for an observer to be in such a loop, which likely shifts the mechanism of probability shift in the desired direction.
• A planet in which life has the ability to create time loops will have higher measure compared with planets without such ability.
• Thus, we are more likely to be on a planet in which life is exploiting time loops to create probability shifts
  • Note that such probability shifts will be subjective for any observer, so external observers will not be able to see them.
  • See Anthropic Trilemma as an example of subjective probability shift.
• Moreover, the ability to exploit such “precognition” may also evolve in time, and humans may be not only most intelligent beings but also most “magical”.
• Thus, a dangerous time machine may be “natural”.
  • No machine is needed.
  • Bad prophecies and other things related to time travel paradoxes are already happening.
  • Even a natural time machine can be enhanced and exploited. 

Natural vs. technological time travel

Natural time travel assumes that living beings had evolved a very useful capability to get some data from the future in the form of premonitions about dangers. A lot of so-called “psy” could be explained by getting data from the future. Natural time travel will create the needed loops which will increase measure. Absurd UFOs could be the only stable loops for the future time travelers in our time. 

Most premonitions are either probabilistic, or paradoxically uncertain if put in words (“и примешь ты смерть от коня своего”). Paradoxical uncertainty makes them unescapable. 

UFOs as time travelers

It was often suggested that UFOs are coming from the future, like tourists. But why are they not afraid of creating the Grandfather paradox? Maybe their absurd behavior can be explained by time loops: absurd things are products of time loops.

Also, they may come from time after such a large catastrophe happened that no information about us survived. For example, if the Sun goes supernova – but in that case they will come not from our future, but will be aliens, and calling this time travel will be meaningless.

Causal loops as strange attractors

Below is a quote from my “UAP and global risks”:

“If we continue to explore this idea, we could suggest that the causal loops of time travel will be not perfect, but similar to attractors: in the attractors, there is some noise and derivations from the main path, but the “plot” eventually returns to the attractor. Very improbable events could happen if the plot of the causal curve starts to significantly deviate from the attractor, and these events will return the story back to the plot. This could be used to explain the slow disappearance of UAP evidence and MIB phenomena (if it is real). UAP can’t leave undeniable traces, or they will cause a grandfather paradox or break the loop.

Also, the causal loops put the future advanced civilization in a strange causal dependency from us: they must send travelers back in time. In the multiverse, there are infinitely many futures, and the needed strange civilizations will always be there.

Small risks here are that we “jump out” of the attractor and actually destroy our future, but it is very improbable as attractors are by definition very stable. 

There is not much empirical evidence which supports the hypothesis. Strange things about time were discussed in the book “Time Storms” (Randles, 2001) and have some similarities to the UAP observations, that is, “missing time” during observations and desynchronization of watches. Missing time events, if real, suggest that the UFO phenomenon is associated with the capability to manipulate time, and thus time travel is plausible. 

However, time storms themselves (as hypothetical natural events) could be an explanation of UAP if UAP are just random observations of some other past or future world during such storms.

Recent book and discussion about the possible explanation of UAP as time travelers (Masters, 2019) and here. 

Butterfly effect variations in time loops

Butterfly effect is amplification with time of small changes in the past. Time loops suppress it, but small variation is possible. Thus, the loops are actually a strange attractor and small oscillations are happening. 

Prevention plans

The obvious preventive measure would be a universal ban on creating time machines until we have a better understanding of their risks. However, the chances of such a ban succeeding are slim, as even efforts to Pause AI development have failed. Additionally, raising awareness about this issue may have opportunity costs, as there is only a finite amount of attention that can be devoted to global problems. Therefore, I suggest that anyone attempting to build even the smallest time machine should carefully consider these risks.

Acknowledgments: I would like to thank James Miller for his useful comments.