Science
Related: About this forumThe virtue of a scientific theory is its *inability* to explain things we *don't* observe.
It's very easy to come up with a theory that *will* explain all the evidence we *do* observe - "God did it" is fine, as is "The Matrix". Theories like that aren't much use, though, because they could just as well explain anything else, and so they have no predictive power.
To be useful, a scientific theory has to be *incapable* of explaining practically any observation we can imagine, *except* that single course of events we actually do see.
defacto7
(13,485 posts)The virtue of a scientific theory is its *inability* to explain things we *don't* observe.
That is one of the prerequisites to even being science at all.
We are put in the honest position of accepting facts as they appear in observation, negating the minds natural tendency to follow a path that tries to make sense out of chaos.
A very reasonable statement.
DetlefK
(16,423 posts)For a neural network, you start with initial parameters and then present already-known data to the network. It will react by modifying itself in tiny steps here and there.
If you don't give the neural network enough experience, it won't be able to see similarities between data-sets.
"All vehicles are cars. Therefore a truck is no vehicle."
If you give the neural network too much experience, its connections and decision-steps become so rigid, that it won't be able to handle new information.
"A red car is a vehicle. This car is green. Green cars aren't vehicles, only red cars."
It's therefore important for a neural network to stop the learning-process at the right point. (The modification of the human brain stops in its Twenties. After that, personality and general interests are more or less fixed.)
DreamGypsy
(2,252 posts)...is the prediction of phenomena or objects that we have not yet observed.
Two recently discovered/discussed examples come to mind:
First, the very recent observation of signs of B-mode polarization in the CMBR, which was predicted as a result of the inflationary epoch at 10^-36 in the Big Bang (from March 17, 2014):
o Gravitational Waves from Big Bang Detected:
Proof of gravitational waves created by cosmic inflation is shown here in this image of the cosmic microwave background radiation collected by the BICEP2 experiment at the South Pole. The proof comes in the form of a signature called B-mode polarization, a curling of the orientation, or polarization, of the light, denoted by the black lines on the image. The color indicates small temperature fluctuations in the cosmic microwave background that correspond to density fluctuations in the early universe. (BICEP2 Collaboration)
Second, Darwins famous analysis of the Madagascar star orchid, Angraecum sesquipedale, that was highlighted by Neil de Grasse Tyson in Episode 6 of Cosmos - A SpaceTime Odyssey:
From Chance Variation: Darwin on Orchids by John Beatty (pdf):
Darwin might have explained the different outcomes mainly in terms of differences in environmental conditionse.g., differences in the insects available for conscription as pollinators. Thus, he might have presumed that different lineages of orchids experience the same variations, in the same order, and then he might have argued that different variations are selected in different lineages depending on which pollinators are in the vicinity. But he did not. His emphasis was on chance differences in variation among lineages.
This emphasis might be obscured by Darwins famous analysis of the Madagascar star orchid, Angraecum sesquipedale, which has an extraordinarily long nectaryup to a foot in length. Darwin reasoned that such a long nectary must serve to attract an insect with an equally long proboscis, although no such insect was known at the time to exist in Madagascar. The hypothetical scenario was initially dismissed as fantasy. Darwin took satisfaction in Fritz Müllers discovery, reported in the second edition of Orchids, of a moth in southern Brazil (unfortunately, halfway around the world from Madagascar), with an eleven-inch proboscis. Darwin didnt live long enough to enjoy the satisfaction of Walter Rothschild and Karl Jordans 1903 description of the long-tongued and appropriately located (in Madagascar) Morgans sphinx moth, subsequently named Xanthopan Morgani praedicta. But Darwinians since then have treasured the successful prediction.
Again, the main point of Orchids was to attribute differences in floral morphology to natural selection acting on whatever differences chance to arise.
So, what happens if the predictions of a theory are 'never' observed. For example, if all the insects species with an "equally long proboscis", as Darwin predicted, were extinct at the time of the development of evolution? Would that mean evolutions was wrong? Of course not - wait around long enough and perhaps some other species of insect would evolve, through random mutation, to take advantage of the unused nectaries of the orchid...or perhaps the orchid would have become extinct (or evolved) due to lack of an effective pollinator.
And if Homo sapiens and its ancestors weren't clever enough to develop detectors sensitive enough to observe B-mode polarization, would that mean that inflation didn't occur? No. However, confirmation of previously unobserved predictions, rigorously scrutinized by the scientific community, is a strong indication that a given theory is on the right track.
Donald Ian Rankin
(13,598 posts)Consider the following two statements:
1) My theory predicts that if X happens, we will observe Y.
2) My theory would be contradicted by any observation other than Y if X happens.
I claim that they are synonyms.
Observing Y when X happens is entirely consisted with "God did it" or "The Matrix", but so are observing Z or 7 or blancmange.
The predictive power of a theory depends on how many observations *wouldn't* fit with it.