Pushing the Boundaries
Cosmology (the study of the origin and evolution of the universe) is an area of science where things are considerably less settled. This is largely due to the inherent difficulty in collecting data from events that happened 13.7 billion years ago. It's pretty amazing that we have any data at all, but we do (at least from about 400,000 years after the start). Then there's the fact that cosmology is trying to answer that little question, "How did it all begin?"
A basic summary of what we know:
Data
|
Conclusion(s)
|
The light coming from nearly every galaxy
we look at (other than the ones very nearby) is doppler shifted to a lower
frequency (like the sound of an ambulance moving away from you).
|
Everything in the universe is moving
away from us, so the universe must be expanding. If the universe will be
larger in the future, it was smaller in the past.
|
There is a small amount of microwave radiation
coming from every direction in the sky.
|
A very small universe in the past
(consistent with the conclusion above) with a high energy density could have
grown to the large universe we see today, with the energy spread out so much
that it is a weak microwave signal.
|
The microwave radiation (known as the
cosmic microwave background or CMB) is incredibly uniform, but not completely
(it is uniform to one part in 100,000).
|
In order to have the non-uniform
universe we see today (galaxies in some places, voids in others) the early
universe would need to have some non-uniformity. The irregularities measured
in the CMB are of just the right size to match the irregularities in the
universe today.
|
The CMB along with supernovae data
reveal that the universe is expanding at a greater rate today than it was in
the past (i.e. the expansion is accelerating).
|
General relativity predicts that
negative pressures can lead to repulsive gravity, so the force that is
pushing outward on the universe is just gravity. This negative pressure could
be associated with a form of vacuum energy (energy contained in the fields in
otherwise empty space) that has come to be known as dark energy.
|
This stuff is all pretty well established, but when you get into the details of Big Bang cosmology there are some significant problems that are best remedied with an addendum known as Inflation.
Inflation posits that the very early universe was suffused with something very similar to the dark energy that we see today, but much more extreme. This energy lead the universe to double in size every ~10-38 seconds. Here's where things start to go a little too far afield for many scientists (physicists included). The mathematics of Inflation lead directly to a prediction of many universes (possibly an infinite number).
This brings me to the main question of this post: How much
can we trust our equations? For physics, this is tantamount to asking how much
can we trust our theories (and I mean theory in the scientific, not colloquial
sense).
While history is replete with scientific theories that have
made predictions that were later confirmed, we have yet to find (invent?) a
theory that doesn't have limitations. Theories can only be pushed so far until
they breakdown. Regarding Inflationary Cosmology and the Multiverse, it's
possible that there will never be a way to directly confirm the existence of
other universes. If a theory agrees with everything we do see, should we trust
it to tell us about things we don't see? (this book argues that we should not)
When this question is considered with respect to quantum
mechanics (QM), many feel that it isn't worth asking: the purview of a
scientific theory is to systematically predict/confirm data, not to attach
meaning/interpretations to those predictions/confirmations. While I disagree
with that reasoning (science should tell us about the nature of physical
reality, or at least our experience with it), I can see how one could take that
stance when it comes to QM. But with Cosmology, the "shut-up and calculate"
approach seems harder to defend. How can one attempt to explain the states of
the universe that lead to the particularities measured in the CMB without
explaining how those states came to be? I guess one could say, "Well, the
data tell us that universe looked like this at this time, and then like this at
another time," and leave it at that. But that certainly isn't very
satisfying, and it's not likely to be as fruitful as developing an explanatory
model.
So explanations are good. We like theories that explain how
systems (even universes) came to be and how they evolve from one state to
another. If a theory then goes on to explain things beyond what can be
observed...well at that point I think one is justified in choosing to
either believe or disbelieve.
Comments
I think extrapolating from data into a law is always difficult, but well worth the effort. That's really the only time we make progress in science, right? --when we can say with certainty X causes Y?
I take your point with cosmology to be that just because a particular theory fits the data doesn’t mean that the theory is correct. I think the issue with cosmology is primarily too little data. I appreciate that scientists have to work within the limitations of what they can observe, and I personally find it fun to speculate about things that are beyond observation. But many theories can be consistent with the data. Just because a theory is consistent with the evidence, doesn’t mean it’s correct. It's a question of probabilities. Some things can be reliably inferred from the evidence, and sometimes the inference is a stretch.
I feel like I see the same error in pop-evolutionary biology. Someone comes up with a theory that a human behavior or emotion (e.g., love or friendship) is an adaptation to survive in a state of nature. The theory explains the "data" --that is, the observed behavior-- and it posits a cause-and-effect relationship, but it doesn't really offer any evidence that what the theory posits actually happened as a historical matter. The only way to know what happened is to get some really good evidence of what occurred historically. Even historical evidence is usually subject to competing interpretations.
This was kind of a ramble, but hopefully, I made a point in there somewhere.
1) What if two vastly different facial recognition algorithms both work equally well? By looking at the workings of a given algorithm, a computer scientist might feel that they have a sense for what makes a face, a face. Yet the programs would be telling different stories (like newtonian gravity versus einsteinian gravity). This illustrates a danger of imbuing physical laws with philosophical interpretations (though doing otherwise is ultimately impossible).
2) What if all facial recognition algorithms share many similarities and work in more or less the same way? We will definitely be more tempted to make a formal definition of what constitutes a face.
Say our face-ology developed to a point where we can speak to the conditions necessary to bring about faces and the face-ologists are making pronouncements on the likelihood of faces existing on planets in distant galaxies. Maybe the conditions necessary to explain the existence of faces also strongly suggest the existence of other universes, universes of which we can never confirm the existence.
Does the success of our laws of faceness persuade us to accept the existence of the multiverse as scientific fact?
But I actually do like the multiverse.
Statistics would say no. The most predictive model (the theory that agrees best with what we see) is almost never the true model. Additionally, the true model is almost never the most predictive model.
Perhaps most interestingly, these claims can be formally proven.
The use of the term "true model" is an interesting choice. It assumes that a model can have perfect predictive power rather than simply being a useful approximation. I'm not convinced that such models exist, but granting that they do can you give me an example of a true model that is not the most predictive? I would think that quantum mechanics is the most true model of the physical world as well as the most predictive.
This is something that comes up in statistics a lot. People often assume that a predictive model is the true model (or the model that is closest to the truth out of those under consideration), and this leads to all sorts of bad data analysis. Unfortunately I don't have a good reference for you on this at the moment, but I have a half written blog post that I'll email to you once it's written.