Editorial: What Makes JAGI Special

Abstract

This Inaugural Editorial attempts to provide some additional explanation (beside what is already in the documents at the JAGI website) on why this journal is started on this topic, at this moment, and in this way. The research of Artificial Intelligence (AI) began with the aim of using electronic computers to produce "intelligence" comparable with that of the human mind. After a short optimistic period, people realized that the task is much harder than was initially estimated, and consequently the mainstream AI research turned to narrowly defined tasks that can be handled by domain-dependent and problem-specific solutions. Influenced by this atmosphere, the existing AI journals tend to encourage research on isolated topics, rather than research aimed at intelligence as a whole. They often reward incremental improvements on existing theories and applications, while ignore issues in the basic assumptions of the work, which exhibit it to be extended and scaled up to problems outside the initial consideration, even slightly. In recent years, more and more AI researchers began to feel the necessity and possibility of returning to the original goal of the field, that is, to treat "intelligence" as a whole, and as a domain-independent and general-purpose mechanism. Of course, here "general-purpose" does not mean to be able to solve all problems perfectly, but to be able to work in various situations, including novel and unexpected ones. General-purpose systems and special-purpose systems have different design considerations and evaluation criteria. The history of AI research hints that the successes in the latter do not naturally extend to the former, nor can they be simply pooled together to achieve the former, without the guidance of a clear overall design. Therefore, different review criteria and procedure are needed for publications of research on general-purpose AI systems. Under this consideration, in the First Conference on Artificial General Intelligence (http://agi-08.org), those in attendance passed a motion calling for the establishment of an academic journal focusing on "Artificial General Intelligence (AGI)". By starting JAGI, we, the editors, are not claiming that there is already a widely accepted theory and technique to reach the goal of general intelligence, or even that there is a proof on the achievability of such a goal. Instead, we haven't been convinced that such a goal is impossible. Therefore, we believe the research in this field should be encouraged, especially given its great potential, which is hard to overestimate. We fully understand the challenge this field is facing. Even though this kind of research has been carried out for more than half a century, the field now labeled "AGI" is still in its infancy. Almost all of the previous projects failed to deliver their planned functionalities, and it is very likely that most of the current projects will end up the same. However, the difficulty of the problem does not imply its impossibility. On the contrary, if nobody dares to try, it will really become impossible. Even though it is rational for the outside people to keep their skepticism about this research, the dedicated researchers should be allowed to continue their exploration. This journal is primarily established for those researchers, here they can exchange ideas, and therefore gradually put this research on a solid ground.