Forbidden knowledge in machine learning -- Reflections on the limits of research and publication

TL;DR

This paper discusses the concept of 'forbidden knowledge' in machine learning, emphasizing the need for ethical norms and review parameters to prevent misuse of sensitive research findings.

Contribution

It introduces the idea of applying forbidden knowledge discourse to machine learning and proposes review parameters for ethical decision-making in research dissemination.

Findings

Highlights potential misuse of generative models and other ML applications.

Proposes a framework for ethical review of sensitive ML research.

Calls for community standards to prevent harmful dissemination.

Abstract

Certain research strands can yield "forbidden knowledge". This term refers to knowledge that is considered too sensitive, dangerous or taboo to be produced or shared. Discourses about such publication restrictions are already entrenched in scientific fields like IT security, synthetic biology or nuclear physics research. This paper makes the case for transferring this discourse to machine learning research. Some machine learning applications can very easily be misused and unfold harmful consequences, for instance with regard to generative video or text synthesis, personality analysis, behavior manipulation, software vulnerability detection and the like. Up to now, the machine learning research community embraces the idea of open access. However, this is opposed to precautionary efforts to prevent the malicious use of machine learning applications. Information about or from such applications may, if improperly disclosed, cause harm to people, organizations or whole societies. Hence, the goal of this work is to outline norms that can help to decide whether and when the dissemination of such information should be prevented. It proposes review parameters for the machine learning community to establish an ethical framework on how to deal with forbidden knowledge and dual-use applications.