Step 1: Test each statement independently against the definition and typical application of Bayes theorem, $$P(A|B) = \frac{P(B|A)P(A)}{P(B)}.$$
Step 2: Statement (i) describes exactly what the formula above does: it lets us compute one conditional probability from another conditional probability and the marginal (prior) probability. This is a true description of the theorem, not an opinion, so (i) holds.
Step 3: For statement (ii), think of a concrete case. If $P(A) = 0.5$ before evidence and updating with $B$ gives $P(A|B) = 0.6$, the prior $0.5$ is not erased or overridden - it is simply combined with the likelihood of the evidence to produce the posterior. Since the word "supersedes" implies the prior becomes invalid or is replaced, this description is inaccurate. Statement (ii) is false.
Step 4: For statement (iii), Bayes theorem is the standard tool in problems where a later observation (an effect) is used to infer the probability of an earlier, unobserved cause - such as inferring which machine produced a defective item after observing the defect. This backward-in-time inference is exactly what (iii) describes, so it is true.
Step 5: Only statements (i) and (iii) survive this check.
\[\boxed{\text{Only (i) and (iii) are true}}\]