The usual practice in traditional artificial intelligence (AI) is to train learning algorithms that behave and respond as close as possible to humans, hence machines are usually viewed as “students” and human domain experts as “oracle”, “teacher”, etc. In this project, we argue that with the fast proliferation of digital data in scientific metadata, AI predictors of scientific discoveries can do much more than just imitating their human counterparts. By not mimicking, but rather avoiding human inferences we can design “alien” AIs that radically augment rather than replace human cognitive capacity. Identifying the bias of collective human discovery, we will engineer alien algorithms that broaden the scope of discoveries by proposing promising hypotheses unlikely for scientists and inventors to imagine without machine intervention. Generating such unfamiliar sets of hypotheses will be the starting point to exploratory studies that are key for punctuating disruptive discoveries through “out-of-the-box” ideas.