It's difficult to turn biological knowledge into helpful treatment actions for patients. Because the brain is such a complicated organ with so many non-linear feedback pathways, neurology and psychiatry have one of the lowest success rates for clinical development of any indication. Furthermore, the brain is not as accessible as other organs, therefore biomarkers are limited. Many experimental medications "treat" preclinical transgenic animal models of "CNS diseases" satisfactorily, only to fail in clinical trials later. Incomplete pathology, variable PK and metabolism of experimental medicines, distinct pharmacology on human vs rat targets, absence of critical shared genetic polymorphisms, and significant differences in neural circuits are all possible reasons for this translational mismatch. Psychiatry and neurology longitudinal and cross-sectional clinical data with deep phenotyping are being collected and made publicly available to the scholarly community at the same time. These can be utilized in machine learning methodologies to train artificial intelligence networks. However, the number of conceivable combinations of "confounding" factors, such as age, underlying comorbidities, genetics, Comedications, and disease stage, significantly outnumbers the number of accessible patients, raising questions about the generalizability of AI or ML predictions.