Also, I think I read somewhere that Epicurus also said that dreams are made of atoms, didn't he? (or did I imagine that, or dream it?). So the gods were made of the same stuff as dreams?
We still don't really fully understand dreams, in terms of neurobiology (at least, I don't). Some of us now probably think that dreams, thought, feelings, emotions, etc, are holistic properties of the functioning of the brain, probably connected with processes dealing with the sorting of memories and emotions. There can also be a psychological (eg Jungian) interpretation of dreams etc, much of which seems to me to be quite astute (although Jung was open to everything, including, at times, the supernatural)
I personally tend to believe in the most firmly established principles of modern science (which is not necessarily the same as believing anything that any scientist says is true). So, I don't believe in anything that seems to me to be supernatural (although I do also see the need for numinous art). But we can't expect the ancient Greeks to have known about 'science' as we know it. Or to have known what is 'supernatural' and what is not. They lived in a pre-scientific world. I find it quite remarkable that Epicurus was able to achieve the freedom of thought to figure out that the gods have no interest in human affairs so there is no need to be afraid of them. (He could also have said 'assuming they exist' but he didn't say that, so he apparently assumed they did exist and were composed of 'atoms' - like dreams).
I don't need to think that any of the great thinkers of the past were right about everything, in order to be impressed by the things that they were right about.
Likewise, Epicurus spoke about the 'swerve'. That 'atoms' fall down with a 'swerve', because he recognised the need for indeterminacy. We now know that is not very scientifically accurate. But we can't expect Epicurus to have been able to deduce Heisenberg's uncertainty principle. The mere fact that Epicurus saw the need to introduce indeterminacy into the 'atomic' model seems quite remarkable to me.
If we want anything better, we will have to take the best of what we have learned from the thinkers of the past, whilst recognising their mistakes, and try to figure it out for ourselves.