Forgive my ignorance, but isn't that how we carbon date things?
No, not really.
We carbon-date things (and do other kinds of radiometric dating) by measuring how
incredibly large numbers of unstable nuclei act. But we do not carbon-date things by looking at
one particular unstable nucleus.
If you have one atom of carbon-14, you cannot predict when it will decay. You can calculate a probability that it will decay in some specified time period, but for that one nucleus that's the best you can do. It may or may not actually decay in that time period.
However, when you have trillions of carbon-14 atoms, or (as is more usual) trillions of trillions of trillions of carbon-14 atoms, you can predict extremely precisely
how many of those nuclei will decay in any specified time period ... but you cannot predict
which of those nuclei will be the ones that decay. You know that
some of them will decay but not the particular ones that will decay.
Radioisotope dating methods work because any samples we take, even the ones that are incredibly tiny by our everyday standards, have enough atoms in them that the large-number-of-atoms statistics work very precisely.
Or, as it says at
The Law of Radioactive Decay:
"You can't, however, predict the time at which a given atomic nucleus will decay. For example, even if the probability of a decay {of one particular nucleus - JRF} within the next second is 99 %, it is nevertheless possible (but improbable) that the nucleus decays after millions of years."
That site has a nice Java applet that simulates decay.