Editorial: What is Artificial Intelligence?

Abstract

Let me share the first definition I will use in my role as editor-in-chief; not as a hard boundary, but as a guiding definition for a first pass to filter papers that do not fit within the scope of this journal. The definition is short and concise, but will nonetheless likely be met with severe scepticism from readers. Definition 1: Artificial Intelligence is the automation of cognition. Presenting AI from a "cognition" perspective, and recalling that the origin of the word "cognition" is the Latin word cogito, which means "to know," sends a clear message that mere development of software and classic industrial automation are outside the scope of this journal. Digitization of documents and systems is a necessary step for organizations to develop AI capabilities, but such a basic form of digitization is not within the scope of Transactions on Artificial Intelligence. That said, designing the knowledge base or the machine learning algorithms for digitization may be within scope. It is important to note that the above definition did not limit "cognition" to humans. , I see it as potentially limiting when the journal receives a paper on human-AI interaction or the ethics of AI. These papers comprise important subject matter for Transactions on Artificial Intelligence; however, they are not automation of cognition papers. This calls for a second definition to complement that already offered. Definition 2: Artificial Intelligence is social and cognitive phenomena that enable a machine to socially integrate with a society to perform competitive tasks requiring cognitive processes and communicate with other entities in society by exchanging messages with high information content and shorter representations. The above definition does not see AI as a technology or a product alone, but as social and cognitive phenomena. No definition for AI will be error free, sufficiently universal, or concisely unambiguous. Each definition may be more appropriate for particular contexts or even just for a particular timeframe. As discussed above, the purpose here is not to agree on a universal definition, but to have a definition in this context, so we know what we disagree about. The two definitions above do not set hard rules on which papers are within or outside the scope of this journal. They are first layer soft filters that need to be augmented with the experience of the editor-in-chief and associate editors to make professional judgments to shape the scope of IEEE Transactions on Artificial Intelligence. I welcome community responses and opinions via email, and look forward to learning what members of the community see as the pros and cons of these definitions.