Why is `predict_proba` not `predict_proba`?

fpredictor.predict_proba(test_data)
Out[4]: 
         <=50K      >50K
0     0.931291  0.740701
1     0.999250 -0.239742
2     0.043029  0.716479
3     0.998207  0.673785
4     0.988926 -0.229418
        ...       ...
9764  0.977868  0.694124
9765  0.941992 -0.182484
9766  0.866390 -0.106882
9767  0.999667 -0.240159
9768  0.742606  0.016902

Negative predict_proba isn't a thing. How are users supposed to use this output if they care about the prediction probabilities?

autogluon / autogluon-fair

Why is `predict_proba` not `predict_proba`? #19