According to Wikipedia it's because there are never enough repressor molecules to go around, so the molecules are always being produced. When lactose is present, however, it will neutralize more repressor molecules, speeding up the process.
An astute observer might wonder how allolactose could bind LacI if the genes necessary for the conversion of lactose to allolactose are under the control of the lac promoter. It turns out that the number of repressor molecules in a bacterium is low enough that at any given time, some percentage of the cells will not have enough to inhibit transcription. This is an example of biological noise. Given time, more cells in a culture will transiently have no LacI inhibition and will express the lac operon, temporarily conferring the ability to take up lactose and convert it into allolactose. This allolactose binds LacI, increasing the probability of more transcripts being made. This positive feedback loop allows for a small signal (cytoplasmic allolactose concentration) to be amplified and induce a significant change in the cell's gene expression profile. This induced state is epigenetic and somewhat heritable: in cell division, each daughter cell will likely have enough inducer to bind and deactivate LacI.