I can't be 100% sure based on listening alone, but...
IMHO it's "baby" with a really-effective pop filter (or deliberate singing technique) and an amplitude ramp of the waveform.
Had a quick look around and every lyric site has it as "baby"... not that it means anything really, but given the stupendous amount of times it's been listened to and talked about over the years you'd think if it were otherwise there'd be some ongoing debates out there. Can't find any.
Also, it's been remixed a kazillion times so you'd think that all that exposure to the individual tracks would've revealed the truth if it were otherwise.
Remember, we tend to hear things as we imagine them. Very-powerful phenomenon and difficult to "unhear" / unlearn.