Can we build a species discovery factory?

Life on Earth is structured into estimated twenty million living species. After 300+ years of study, biological systematics named and classified only about two million (ca. 10%) of them. With such pace, Linnean taxonomy needs thousands of years more to complete this job for all species. This is unacceptable now, when human induced ecosystem depletion calls for an urgent and complete species inventory. One of the bottlenecks which makes species discovery so slow is our dependance on the phenotypic data, the visual species description, even in the genomic era. With help of machine learning, I want to convert tedious morphological species examination into an automatic pipeline to accelerate species discovery at an order of magnitude. Since phenotype is a much more complex system than the nucleic acids of the genetic code, the outcome remains uncertain.