Hacker News Reader

Comment by whimsicalism | Hacker News Reader

> And without data, the question is not which solution to the problem is correct, but which one you like best.

I think this is correct to some extent, ultimately unavoidable, and some assumptions can be "inferred" as more reasonable than others, even in the absence of falsification.

By Sabine's stance, the problem of induction [0] hasn't been "solved" because we actually have no reason to assume that past events are at all related to future events. So when we say that induction is possible, we say so because we "like best" that theory of causality, rather than my alternate theory that when you finish reading this sentence, all physics will cease to function and the universe will end.

It didn't happen, but we had no reason to know so ahead of time deductively. However, it is still reasonable (in my view) to believe in induction.

So sure, it's perhaps 'unsolvable', but if a plausible explanation comes around that is consistent with modern physics that seems good to me.

[0]: https://plato.stanford.edu/entries/induction-problem/

View on HN

> By Sabine's stance, the problem of induction [0] hasn't been "solved"

The "problem" of induction is a different kind of problem from the black hole information loss problem.

Induction can't be tested against experimental data. Induction isn't a testable hypothesis; it's a strategy we have no choice but to adopt if we want to plan for the future at all. So there is no "problem" of induction at all: it's just something we're stuck with.

Proposed solutions to the black hole information loss problem can be tested against experimental data; we just don't have the technical capability to acquire such data yet. That doesn't change the fact that until proposed solutions are tested against experimental data, and some proposed solution passes the test, the black hole information loss problem is not solved.

The problem if induction attacks the idea of experiments leading to anything meaningful.

My point is that there are always going to be equally possible alternative theories which make any problem "unsolvable".

I could make a theory that says that gravity works exactly as we think it does, except in about 1000 years will cease to function entirely - and that theory would be equally consistent with observation. I would be equally correct in saying that we do not yet posses the technology (time travel) to falsify this theory.

We have to rely on some sort of proxy for the simplicity or elegance of the theory in order to preclude hypotheses like the above. If we find an elegant solution to the BHIP that uses existing QM + GR (which are empirically verified), then that seems like a pretty good resolution even if it can't be observationally verified directly at a black hole yet.

> We have to rely on some sort of proxy for the simplicity or elegance of the theory in order to preclude hypotheses like the above.

No, we don't. The rule that precludes your hypothesis is much simpler: the laws of physics don't change with time. That doesn't require the laws of physics to be "simple" or "elegant", and doesn't require anyone's subjective judgment about which laws of physics they think are more beautiful.

>The rule that precludes your hypothesis is much simpler: the laws of physics don't change with time

Why should we believe this rule?

You can believe what you like - science is not concerned with persuading every skeptic.

The point isn't skepticism of science, it's the recognition that something akin to Occam's Razor has to be used in theory, as there will always be an infinite number of unfalsifiable theories for any set of observations.

In this context, Sabine is the skeptic of theory writ large.

So many replies taking my argument to be the exact opposite of what it is.

My post above was a specific response to whatshisface's comment, saying that science need not, and will not, be hobbled by inordinate skepicism (which, if persued consistently, collapses into useless solipsism.) In this particular case, science adopts the idea of unchanging laws as a methodological assumption rather than an axiom.

Similarly, science need not, and will not, be hobbled by the "infinite number of unfalsifiable theories" that you raise.

In effect, science tables all these fanciful objections unless and until circumstances and evidence make them relevant. Scence is not metaphysics - it is not attempting to deduce how the world must be; it is trying to find out how it is, and so the sort of arguments that philosophers get entangled in (such as "possible worlds" arguments which ultimately reduce to what a given philosopher can imagine) don't carry much weight. This is not just "something akin" to Occam's razor; it is precisely that, and it is not a vague preference for "beauty" or "elegance".

Elsewhere, Hossenfelder made an insightful comment about falsifiability, along the lines that falsifiability is just table stakes for a hypothesis to make the grade as a theory, and not a guarantor of merit.

> I could make a theory that says that gravity works exactly as we think it does, except in about 1000 years will cease to function entirely.

I suppose you could - but this vacuous armchair theorizing, which, I hope, took you less than a minute to dream up, has no bearing on scientific theories of gravity, which are based on what has been observed. This does not depend on any concept of 'simplicity' or 'elegance', but, at most, on Occam's razor (which is not, as Wikipedia claims, a matter of "the simplest explanation is usually the right one": it is a matter of eliminating premises that are not explaining anything beyond what can already be explained.)

Oh yeah, Occam's razor has nothing to do with any concept of "simplicity". /s

You're literally stumbling into my conclusion and then arguing that it is different from what I was saying. You can even look at my other replies to you referencing Occam's Razor as exactly the sort of thing I'm talking about.

Your sarcasm reflects back on you, as explained by my concurrent edit to my point, which I will repeat here: Occam's razor is not, as Wikipedia claims, a matter of "the simplest explanation is usually the right one": it is a matter of eliminating premises that are not explaining anything beyond what can already be explained. Your armchair "hypothesis" that, in about 1000 years, gravity will cease to function entirely, is not something that needs to be explained.

Rather than delving into whether additional assumptions is related to the complexity of a theory, which I think is a largely pedantic debate, I think we don't disagree as much as you perceive.

> not something that needs to be explained

I agree, and I'm saying that Sabine is discounting the same tool that you used to reject my hypothesis. That is my point. If there is an explanation for BHIP that does not introduce any new assumptions or very small assumptions to existing theory, that lends it credence even in the absence of direct observations of black holes, just as we lack direct observation of gravity 1000 years from now.

I entered this discussion merely to point out what I saw as a problem with an example that you chose to make your point with. If you think that leads to a pedantic debate, then maybe it wasn't a good example to begin with. So long as you continue to insist, as you do in your last sentence above, that this is an appropriate analogy for the point you are trying to make, I will regard it as a weakness in your position.

To put that aside and address your larger point, then if we had just one candidate that provides "an explanation for BHIP that does not introduce any new assumptions or very small assumptions to existing theory", then it would indeed be a plausible candidate explanation. That is not, however, the situation we are faced with: Hossenfelder's argument is that there are many equally plausible (and equally speculative) ways in which the paradox could be resolved (though none that "[do] not introduce any new assumptions or very small assumptions to existing theory"), with no prospect of us getting the data to choose between them (note that Hossenfelder is highly skeptical of choosing between theories on the basis of 'beauty', 'elegance' or other subjective distinctions[1]. As I explained above, Occam's razor is not in this category.)

As far as I know, this argument might fail if some data from a seemingly unrelated field supports just one of the candidate resolutions of the paradox. This might be the point you are trying to make, but if so, your "gravity might fail" analogy isn't helping, as we are not currently faced with multiple, equally-plausible theories of gravity, one of which will be preferred if gravity stops working in 1000 years (speculating that it might do so is neither an observation nor a theory of gravity, it's just speculation.)

[1] https://www.amazon.com/Lost-Math-Beauty-Physics-Astray/dp/04...

The pedantic debate is because you refuse to acknowledge Occam's Razor has anything to do with the simplicity of a theory, not related to the specific example I constructed.

> Hossenfelder's argument is that there are many equally plausible (and equally speculative) ways

This is just not true. Most of the resolutions of the past required new conjectures like EP=EPR, whereas recent solutions rely on newer sort of uses of a path integral, but are still built up from basic QFT+GR.

Claiming agnosticism between those two theories is discarding Occam's Razor entirely, as the latter relies on constructing way less strong of assumptions and provides a believable mechanism for information preservation.

I have explained why Occam's razor is not a simplistic and subjective matter of preferring simple solutions, and your only response to that is to keep saying I'm wrong, without saying why. That is one factor in keeping this discussion going. The other thing was your insistence that your "gravity might fail" scenario is a relevant analogy; at least you did not continue pushing that view in your latest post.

>> Hossenfelder's argument is that there are many equally plausible (and equally speculative) ways

> This is just not true. Most of the resolutions of the past required new conjectures like EP=EPR, whereas recent solutions rely on newer sort of uses of a path integral, but are still built up from basic QFT+GR.

As far as I can tell, what I wrote here seems to reflect what Hossenfelder is saying in the article. Here, for the first time, you seem to be making a reasonable argument that the specific hypotheses are not, in fact, all alike, and that they may differ in a way that at least opens up the possibility of applying Occam's razor correctly - i.e. by rejecting any that make unnecessarily broad or strong assumptions.

This brings us to Hossenfelder's follow-on point: we're going to need more data in order to figure out which, if any, is correct [1]. Without more data, neither Occam's razor nor any other principle (and certainly not subjective simplicity) is sufficient to do this - the most you can do is have a preference, based possibly on which you think is least likely to run into problems with Occam's razor. That's not enough to turn a hypothesis into a theory.

This is the point that you started this thread by objecting to, and I do not think you have yet made a strong enough case to overthrow it, though you are now making better arguments than before.

[1] I'm not sure about Hossenfelder's apparent claim that this data can only come from observing black-hole decay, but that's a separate matter.

Isn’t the information loss problem hypothetical in the first place? I doubt anyone has seriously tried to observe information coming out of a black hole.

I'm not a physicist so if I got anything wrong, please correct me.

The problem is that information conservation is a fundamental part of quantum mechanics. QM is a hugely successful theory which has proven itself to be accurate to absolutely incredible precision.

The black hole information paradox implies that there is no way for information to escape the black hole, breaking conservation of information.

The implications of this is huge: Either information can escape a black hole, which suggests general relativity is wrong. Or information cannot escape a black hole, in which case QM is wrong.

So the key to coming up with a solution to this problem is to explain what happens without breaking physics. One suggestion have been that the information actually never enters the black hole in the first place, it's just spread out over the surface of the event horizon.

> The black hole information paradox implies that there is no way for information to escape the black hole

Not quite. When Hawking radiation is included, it becomes possible in principle for information to escape the hole. That's not the main issue.

The main issue is that any information carried by an object that falls into the hole and hits the singularity inside the black hole gets destroyed. So either that information is lost, which violates QM unitarity, or the information has to get copied into the Hawking radiation that gets emitted, which violates the QM no cloning theorem.

> One suggestion have been that the information actually never enters the black hole in the first place, it's just spread out over the surface of the event horizon.

This doesn't really help, because the information can't just stop at the horizon; if an object falls in, it carries its information with it. So "spread out over the surface of the event horizon" really has to mean the information gets copied there, which, as above, violates the QM no cloning theorem. Some physicists, such as Susskind, claim that this actually isn't an issue because no single observer will ever observe both copies of the information, but this argument strikes me as contrived (and I don't think it's been proven that it must be the case).

Two possible resolutions that seem to me to at least avoid the problems I stated above are:

(1) Quantum gravity effects change the singularity inside the black hole to something else: one commonly suggested possibility is that a "baby universe" is born there, and that baby universe carries the information that fell into the hole. In other words, the information stays inside the hole and doesn't come out in the Hawking radiation, but it doesn't get destroyed either.

(2) Quantum gravity effects prevent a true event horizon and a true black hole from ever forming in the first place. That means there is never any singularity or any place where information ever gets destroyed; quantum gravity effects end up converting all of the objects that fell into the "hole" (which is now not a true black hole, but will still look like one to us on the outside, at least for a very long time) into the Hawking radiation that the hole emits as it evaporates. So no information ever gets destroyed and no information ever gets cloned.

As far as I understand, there are plenty of plausible explanations that are consistent with modern physics. The problem is finding out which one of them (if any) is right, since they are not all mutually compatible.

Sure, but I think there's a difference between "if we conjecture this additional effect that has never been observed but is not inconsistent with observation, then we see this thing" vs. "actually this is sort of resolved by a straightforward application of solutions to models we already had of effects we have already observed"

This is a problem with "philosophical" induction, not logical induction. I don't even know what it means "to believe in [philosophical] induction." Does it mean to believe that future things will behave like things similar to them in the past, except when they don't, in which case we made a mistake in thinking they were similar?

> Does it mean to believe that future things will behave like things similar to them in the past, except when they don't, in which case we made a mistake in thinking they were similar?

It's somewhat more fundamental than that: it's attacking the idea that experience can be used in any way. Not believing in inductive reasoning would mean being constantly surprised at everything you see, as it's theoretically just as likely that an object that you saw 1 second ago will disappear as it is that it will not. You would be surprised every day that the sun has risen again.

I'm confused, isn't the whole schtick around past events being related to future events the basis of causality?

Right, and GP point is that we actually only have data about how past events are related to other past events (and even then, the data is not very good). We have no information at all about how past or current events are related to future events.

What evidence do we have that causality will continue to “exist”?

People who believe that past experience is a good predictor of the future will make decent predictions as a result; people who believe that past experience has nothing to do with the future ("anti-inductivists") will make wrong predictions again and again. The inductivists will therefore outcompete the anti-inductivists.

On the other hand, as a friend of mine pointed out, for the anti-inductivists that manage to exist, although they keep suffering from making wrong predictions, they will not see this as a reason to change their philosophy, so they are stuck in an epistemic trap. Since evidence has no meaning for them, no evidence can change their minds.

> The inductivists will therefore outcompete the anti-inductivists.

You have no deductive or a priori reason to know this, rhetoric about "epistemically traps" aside.

So is this survivorship bias or something else? I don't understand your point. I have a white wall in front of me. If I take a brush and red paint and paint a big red circle on the wall, are you saying that the big red circle "just happens to be there" with no relationship to the brush, red paint and my actions?

You only know that there will be a red spot on the wall based on your past experiences with the physical world. Why should we believe our experiences in the past are at all relevant to the future? We haven’t lived in the future. The laws of gravity could be totally different.

All of our data about gravity is equally consistent with the theory “If I drop an apple it will fall only if it is before December 2020” as it is with our theory that the law of gravity will continue to exist in the future as it does today.

That's true. "Induction works in practice" is merely an empirical fact. But it is a fact. Our bodies embed it already: evolution has worked because the environment has changed slowly enough for advantageous mutations to remain advantageous for long enough to become ubiquitous in the species. But, again, the thesis "empirical evidence doesn't matter" is immune to empirical evidence.

Now, if one were to try to construct a theoretical argument, I suspect it would go as follows: "If we assume the world could change in every aspect from moment to moment (e.g. pieces of matter could teleport anywhere in the universe, physical constants might change by the second, laws of physics might change over time or space, the very concepts of space and time might become invalid, ...), then the space of possible worlds is enormous and incomprehensible. If we don't have any rational way of assigning probabilities to any of the possible worlds, then this method of thinking doesn't have any implications for correct actions, because every action could be the best in some possible world; in that case, as long as you assign nonzero probability to the "laws of physics remain constant" universe most of us believe in, you may as well act as though it's the truth." If someone does claim to have a rational way of assigning probabilities to the possible worlds, then that would have to be addressed on its own terms, and any conclusion may be possible; but most such ways that people have proposed will tend to imply that "simple" possibilities are most likely, and "the laws of physics remain constant" will tend to rank highly among them.

Exactly, and your responses both apply to Sabine's criticism as well.

I like the tenor of your second argument, it reminds me of Pascal's wager. That said, you're not really addressing the problem of induction. We have no reason why you'd assign a higher probability to the "law of physics remain constant" to the "laws of physics are flipped in the next second", outside of gut reasoning.

If we can use gut reasoning about induction, no reason we can when we assume that quarks are a real thing, or think about this solution to the BHIP.

To be clear, a true "anti-inductivist" as I describe it would be completely deranged and unable to function in life. It's not really relevant except to people who start from a normal perspective and are considering the anti-inductivist hypothesis (usually either because they're studying philosophy or because someone else brought philosophy into the discussion, which is what happened here).

For those people, it is important to realize that anti-inductivism can't be defeated on its own terms. By usual standards (for some definition of "usual"), the space of possible observations one could make is enormous, and the fact that they keep very, very consistently being in accord with the laws of physics as we've discovered them—often to many decimal places—is immense evidence in favor of "physics as we've discovered it" as compared to "anything might happen". Any viable competing theory would have to be reasonably simple (i.e. not include "billions of specific items that explain all prior observations" as axioms) and would have to make almost exactly the same predictions.

("Newtonian mechanics" is reasonably simple and makes similar predictions at the level most people can see, but with specialized equipment and experiments it has been disproven. "God did it" sounds simple, but either the word "God" smuggles in the extremely non-simple "God is a being whose psychological makeup led him to make the following billions of specific decisions that explain all prior observations", or it amounts to "God is a being who chooses to run the universe according to something approximating Einsteinian mechanics"; in that case, it is possible that God might decide to intervene and do something physics would say is impossible, but if you want to say that God will do any specific intervention, you would have to make a less-simple theory that explains why God would do that particular thing and not anything else, and the longer time passes without any verified physically-impossible miracles, the more unlikely most such theories get. "Intelligent beings vaguely like us arose, and decided to simulate our universe, following rules that approximate Einsteinian mechanics" is also possible—not much different from the God-based class of theories.)

we don't. It's the central faith of science that causality is a thing.

You can't argue that causality is real because it's always worked; that is begging the question.

If "causality" were to fail, I can't see "evidence" faring any better. What would count as evidence, absent a serviceable notion of causality?

Reply on news.ycombinator.com