The classical interpretation needs so-called hidden variables which determine the state of the cat.

In the quantum mechanics interpretation the state is determined when the interaction (measurement, looking at it) takes place. Before that there are several parallel realities of what could be.

In the case of the cat we cannot actually determine which interpretation is correct. We need a different kind of experiment:

As far as I understood (no physicist here) they use two entangled objects. E.g. two photons which are generated by a special process so that one polarized one way and the other one polarized the other way but we don't know which is which before the measurement.

Then you need to measure both photons separately and determine (statistically) which has which polarization.

These measurements need to be spaced out far enough such that faster than light speed communication would be required for the photon measured first to tell the second photon what was happening.

So far it still doesn't help to exclude the possibility of hidden variables since the measurement setup is static.

Another trick is needed.

They use a random generator to determine at both ends what to measure (which polarization direction) shortly before doing the actual measurement. So, when the photon is generated it doesn't know what would be acutally measured.

When you tabularize the all the possible combinations of how the photons could be measured and calculate the possible probability for the measurement results, there is a difference between what we would expect from classical physics when the status is determined at photon generation and stored inside the photons in a hidden variable and the result we would get from quantum mechanics without a hidden variable but with (faster than light) spooky entanglements.

The experiments show that there is no hidden variable. This is a BIG thing because this means that classical physics are not enough to explain the universe.

QM is like lazy evaluation. The result is not computed until someone actually forces it. This is different from the result being computed but you just don't know what it is.

Another way is to think of entangled states is like two virtual addresses which alias to the same physical address. The addresses could be very far apart (similar to entangled particles being very far apart), and their contents are coupled but "uncomputed", until one of the virtual addresses is accessed, faulting in the physical page. At this point, both addresses instantly point to the same contents.

https://vimeo.com/126833477

The science part here, is that a hypothesis has been proposed about the inner workings, and the "so what?" objection is a valid question. If the "so what?" is nothing, it's just an idea.

The "so what?" part is where the science happens. In trying to promote the idea (hypothesis) as a working idea with evidence (theory) We look for side effects of the idea that are expected to be different than our existing ideas (theories). We expect to see a cat when we open the box, and it's alive or dead. That's not interesting. But let's actually ask a science-like question of the consequences. These aren't related to the quantum mechanics questions really, but it's the sort of "consequences" questions that do get asked.

Maybe we can do the experiment over a longer time, like a week. We put the cat in the box with food and water. In the original experiment the poison is released at a random time during the week. So under our existing understanding (theory) we expect, if we run the experiment a hundred times we would see that random amounts of food and water are left pretty much evenly from 0% to 100% depending on when the cat randomly actually died.

Here is a different idea (hypothesis) about the side effects: If the cat really is alive and dead as hypothesized, it will use only the part of the food the 'alive' part would use, which is reasonably predictable. It changes by percentage over the week from 100% (alive) to 0% (fully dead). This means we should always see the same, or similar amounts of food and water left at the end. This is different than the usual prediction!

Now we have a new idea, a hypothesis that relies on consequences of the first hypothesis. So we run 100 experiments, and unlike our current understanding of cat mortality, we find that in this case the remaining food and water values a grouped up at about 50%! We can even do a control where the cat is observed the entire week, and we find that under these "observed" conditions our expected spread of values does happen. We've proven that our previous idea isn't enough to explain everything, and we have tested a prediction of the new idea, so our previous theory isn't "wrong" but it isn't entirely accurate, and the new hypothesis has gained evidence to become a new more complete theory.

Now in the real world at the cat scale, we definitely would NOT expect this to be the case, and real science is much more complicated. It's easy to make errors in predictions, or have "confounding" factors that throw off our investigations. But the initial investigations into spooky action were every bit as bizarre and unexpected. We're learning how things work at the atomic scale, and it turns out that while they mostly behaved as we though they did, there are some odd edge cases that really turn things on their head. Much like we don't poison cats in our day to day life, just in thought experiments (I hope), we don't interact with spooky action in science a whole lot either. But as we learn more, it informs our understanding, and frequently leads some unexpected concequences.

For an excellent real-world consequence, under the classic model, photosythesis is too efficient. It requires quantum entanglement to explain the efficiency! http://io9.com/new-evidence-that-plants-get-their-energy-usi...