The problem is that the word "photon" is the most abused concept in historty ;-)). Not only popular-science books but also (otherwise very good) introductory textbooks on quantum mechanics start with a chapter on the "historical development" introducing the subject with a notion of "photon" that's outdated, namely the "old quantum mechanics". Ironically Einstein got the Nobel prize not for his long-lasting achievements in our fundamental understanding of space and time, the theory of relativity (both the general and the special, where as the name says the special theory is a special case of the general in circumstances where gravity is negligible), but for his ideas on photons, mostly the photoelectric effect. That's ironic, because these ideas where already obsolete during his lifetime with the development of modern quantum theory (Heisenberg, Born, Jordan 1925; Schrödinger 1926; Dirac 1925). The correct notion of photons was introduced by Jordan and independently by Dirac when the quantum-field theoretical formulation of many-body quantum theory has been discovered.
The point is that nearly anything, advocated as the proof of the existence of photons is explainable as well in terms of the semiclassical theory, where only the matter particles (mostly electrons if we consider the detection of light with everyday-matter) are described quantum theoretically while the electromagnetic field is treated as classical. From this picture you get precisely the formula describing the energy bilance of the photoelectric effect a la Einstein, ironically showing that the photoelectric effect does not prove the existence of photons at all. On the other hand it shows the necessity for a quantum-mechanical description of matter. In this respect Planck was correct with his scepticism against Einstein's idea of "light particles".
Another statement, often made, is that you can procuce single photons by just diming a laser to very low intensity. The same holds also true for antenna sending out radio waves, as stressed by Vanadium already. At high intensity these devices send out classical electromagnetic waves (light being just the electromagnetic waves with typical frequencies in the range our eyes are sensitive too and which we thus realize as light). If you ask, how to understand such classical waves in a quantum-theoretical way, you come to the idea of coherent states. These are quantum states of the photon field which are a superposition of states containing all numbers of photons. It's hard to explain without the mathematical machinery of quantum field theory (quantum electrodynamics in this case). The point is that, if you have such a state of the electromagnetic field and you ask, "how many photons are there?", you'll get a different answer in any measurement you make on a such prepared state. Quantum theory only tells you the probability distribution to find 0, 1, 2,... photons. What you get is a Poisson distribution,
P(n)=\frac{\lambda^n}{n!} \exp(-\lambda),
with an average photon number
\langle n \rangle=\lambda.
So you can make \lambda as small as you like (even less then 1). If you plot the distribution in such a case, you'll see that the most probable photon number to find is 0, i.e., no photons present!
So the answer is: If you dim down a laser or a radio antenna to very low intensity you don't get a single photon source but a coherent state of very low average photon number, but the photon number is indetermined. You can find with some probability 1 but also with some (lower) probability 2 or more photons.
It's not so easy to really get a single-photon source. Nowadays the quantum opticians employ birefringent crystals. Shooting with a laser on it, under certain circumstances you get out polarization-entangled photon pairs. Besides the fact that you have this entanglement which gives rise to the weirdest quantum phenomena (weird of course only in the sense that we are not used to the quantum behavior in our macroscopic every-day experience) you can also have true one-photon states: By detecting and absorbing one of the entangled photons, you are sure that there must be the other photon, and it's indeed really exactly one photon.
