Kripke is discussing section 193 of Philosophical Investigations, where Wittgenstein draws a distinction between a machine as an idealized symbol of a rule-following or law-governed mechanism, and a machine as a real object whose behaviours and operations are subject to failure. Wittgenstein here was particularly concerned with trying to shed light on the concept of rule-following, the idea that following a rule means we somehow grasp all the infinite instances of behavior that could be thought to be in accord with the rule. He was saying that the instances are not mysteriously contained 'in' the rule, but rather the idea of a rule is an idealization and to be understood as more like a general ability represented by specific behaviors than an infinite number of possible occurrences.
This could apply to functionalism, one of the traditional tenets of which is that dispositions to behavior are encoded in actual brain states. (It is possible that forms of functionalism can be characterized without reference to brain states, but I do not know of anyone who has actually suggested this). That is to suggest there are actual states in the brain which encode or contain 'in' them an infinite number of possible behavioral outputs. This is the same basic error as thinking of rules as mysteriously 'containing' their instances. Naturally, Wittgenstein does not himself mention functionalism, which originated after his death. Hilary Putnam, more influenced by logical positivism than the later Wittgenstein, was its chief advocate, though he himself later abandoned it.
Kripke's discussion of Wittgenstein on rule-following is notoriously idiosyncratic, and seems overly concerned with reading into W a 'skeptical' method, a reading which has been very controversial. For an alternative, detailed analysis of W on rule following, Baker and Hacker's Language, Rules and Necessity remains the locus classicus.