Placebos are not 'ineffective'. In fact, when three condition trials are run (no treatment vs placebo vs medical treatment), placebo consistently out performs 'no treatment' and of course, not uncommonly, the medical treatment condition as well.
Placebos are not a 'non-specific' treatment. A study on people who take the dopamine-boosting drug L-DOPA for Parkinson's disease but who took a placebo L-DOPA pill, showed almost identical brain changes, as if they'd taken the real thing.
Furthermore, studies done in the 1970s showed that when heroin users inject water (sometimes done deliberately to alleviate cravings when drugs are in short supply), they can experience drug-like euphoria and have been observed to show opiate-like physiological signs such as pupil constriction.