Software Diagnostics Institute

If wrong labels are ML bugs, then wrong answers are ML crashes.

We propose to define Diagnostics of AI as a combination of ideas and approaches from different disciplines such as software engineering, computer science, mathematics, physics, chemistry, biology, literary theory and criticism, narratology, semiotics, humanities, sociology, medicine, forensic science, psychology, psychiatry, and psychopathology with the following main directions of our research activities:

• Pattern-Oriented Software Diagnostics and Data Analysis
• Systemic Software Diagnostics
• Software Narratology
• Software Pathology

It is also important to differentiate diagnostics of artificial intelligence from diagnostics of artificial intelligence systems which is just software and hardware diagnostics, where diagnostics of AI may be just a part. Diagnostics of AI systems is metaphorically like medicine, whereas diagnostics of AI is like psychiatry and psychopathology.

We also organized a few groups for information sharing, such as relevant books and articles:

LinkedIn Diagnostics of Artificial Intelligence group
Facebook Diagnostics of Artificial Intelligence group