Does someone know what are the different roles between a Program Committee (PC) member and Reviewer for conferences (e.g., in the field of machine learning)?
The responsibilities of a programme committee member and a mere reviewer could easily be the same, or differ. All depends on the particular conference/workshop setup. E.g., in computer science/AI there the difference is mainly a result of the venue size. Large first-tier conferences, such as e.g., IJCAI have a four layer programme committee structure. Here PC chairs, the top layer, govern the whole process from recruiting the various types of PC members, through conflict-of-interest handling, bidding on submissions, to notification and proceedings composition and publication. For such conferences the PC chairs recruit Senior PC members who are responsible for larger batches of submissions, or small sub-areas and report to PC chairs. For some conferences, the SPCs recruit regular PC members who in turn report to them and are responsible for the reviews. Regular PC members normally do the reviews themselves, but it's relatively common that they sometimes "subcontract" the job to other reviewers they recruit. These lowest level reviewers would do the review and get credit too, but the actual PC member is ultimately responsible for the review and the discussions among the PC members. Upon completing the review and discussion phases, SPCs give final recommendation for acceptance/rejection and possibly write meta-review summarizing discussion to each submission. The PC chairs would distribute the notifications and handle the proceedings and programme schedule composition.
Now, for mid-size conferences, or workshops the senior PC members layer is usually missing and regular PC members report directly to PC chairs and take over the tasks of SPCs for large conferences. The mechanics of sub-reviewing stays the same.
For small-size conferences and workshops, there are virtually no administrative tasks left for the PC members, so they do just the reviewing (possibly recruit sub-reviewers) and the PC chairs finally decide about acceptance and rejection of the individual submissions on the basis of the received recommendations.
What I describe above is regular scheme of things in "applied computer science".
PC = reviewer + administrative work.
They are responsible for: