As I understand lasers, you start off with a few photons that are in an identical state, and other photons that are created later tend to have the same quantum numbers due to Einstein-Bose statistics. Isn't each photon that "joins" the group of preexisting ones a clone of the previous ones? Why doesn't this violate the no-cloning theorem?
As Rococo already pointed out, the no-cloning theorem doesn't forbid cloning of all specific states. It just states that you cannot make copies of arbitrary (general) states.
Let me (briefly) reiterate the core of the theorem: To clone a state you need a linear operator C that maps a state $|a\rangle|0\rangle$ to $|a\rangle|a\rangle$. This is not possible for general states: $$C |\lambda a+ \mu b\rangle|0\rangle$$ would have to map to $$|\lambda a+\mu b\rangle|\lambda a+ \mu b\rangle$$ per definition of the operator. But linearity (and homogenity) leads to $$C |\lambda a+\mu b\rangle|0\rangle = \lambda C|a\rangle|0\rangle + \mu C|b\rangle|0\rangle = \lambda|a\rangle|a\rangle + \mu |b\rangle|b\rangle$$ while $$ |\lambda a+ \mu b\rangle|\lambda a+ \mu b\rangle = \lambda^2|a\rangle|a\rangle +\lambda \mu (|a\rangle|b\rangle+|b\rangle|a\rangle)+\mu^2|b\rangle|b\rangle$$
So you see that for e.g. $\lambda=1, \mu=0$ (i.e. a base state) there is no contradiction. But you can't clone a general superposition of your base states.
A short history of the laser:
1917: Einstein describes theory of stimulated emission
Following decades: Lasers (and masers) are thought to be impossible due to quantum theory (Everyone knows Einstein was wrong about quantum theory :)
1951 Charles Townes builds a maser - some people are just awkward (John von Neumann: 'That can't be right', Niels Bohr: 'But that is not possible' - see How the Laser Happened: Adventures of a Scientist)
1982 Nick Herbert invents FLASH - A superluminal communicator based upon a new kind of quantum measurement (see How the hippies saved physics : science, counterculture, and the quantum revival )
Later in 1982: No-cloning theorem is devised, explaining why FLASH won't work.
So how was FLASH supposed to work? Well two entangled photons were produced with the "same" polarization (Actually, I think, they will be opposite polarizations, but that can be dealt with, and I'll ignore it from now on) One of the photons is measured at one place, thus collapsing the wave function. At another place the other photon is fed into a laser, which produces lots of photons with the "same" polarization. Measuring several of these allows you to work out how the wave function has collapsed, and thus what measurement was made at the other location, allowing superluminal information transfer.
The trouble is that, in reality the word "same" has two different meanings. In the entanglement case it means that the result will be the same whatever polarization measurement you do. In the case of the laser it doesn't - as wataya says, it might clone the linear polarization, but not the circular polarization.
I think there is a problem here, but I'm not sure what it is. It may be our ideas of quantum states - saying two states are both represented by |H> doesn't actually tell you how much the same they are. Or it could be our ideas of lasers - that the common idea of one photon stimulation the emission of another is inaccurate, and that the working of lasers has to be thought of as a collective way.
I find the question very interesting. After some thought I think the situation of the laser is actually quite different from the no-clone theorem scenario so now I feel that there is probably no paradox here after all. This is my thoughts:
The laser's working is basically, when there are a huge number of photons already present in a state, an atom will preferably emit into this state due to bosonic enhancements. So in other words this is copying the quantum state into a new photon from a large sample. And this does not feel contradictory to the no-clone theorem. I mean, I can think this way: since there is such a large number of photons already present, one can actually measure a lot of them to get very accurately the quantum state and produce them as abundantly as one wish. But this is precisely what one cannot do in the no-clone-theorem case: since there is only one sample available, one cannot take this measure-and-produce route.
So to sum up, I would say the laser is copying from a large sample which is allowed, and not contradictory to the no-clone theorem.
But then there is of course a natural question: how this large sample of identical photons come into being in the first place. Is it the result from cloning starting from a single photon? I am not familiar enough with the experimental details of lasers. But I think it is not, one cannot deterministically amplify from a source of a single photon using a laser setup, and my reasons are as follows: according to my limited knowledge of lasers, there is no experimental proof for deterministically amplifying a single photon. Furthermore, theoretically, when there is only a single photon present, from an atom's point of view, the bosonic enhancement for emitting into the photon's mode is not large, and the atom will emit with high probability also into other modes, and there is actually no cloning here.