WBAI had the following objectives at its inauguration in August 2015.
Through discussions, it has been made clear that WBAI should encourage researches by the WBA approach rather than doing research itself. It is to avoid rivaling with other research entities and to be consistent with our policy for open technology.
The following are three major trends with regard to policy making after our inauguration.
In the WBA approach, we believe co-productive development of WBA by a community will be an important option for hastening the reach to AGI (Fig. 2). With , development in a community becomes possible by decomposing the entire R&D into the R&D of brain modules and that of brain-like cognitive architecture. To support this development, micro and macro level neuroscientific knowledge is required. Knowledge in AI and cognitive sciences is also required for the construction for cognitive architecture.
With regard to the acceleration of technology, we set out for forming an open AI R&D community as large as having the population of 10K people to promote rather short-term R&D. The aim here is to expand faster open AI R&D by combining engineers and experts across the boundary of institutions. We have already formed a SIG where hackathons with activities around Deep Predictive Network and LIS (Life in Silico) the learning environment simulator for AI. By this means, AGI development from the WBA approach will be accelerated and the shared knowledge could be passed on to the organizations affiliated to the participants for the source of competitiveness.
WBAI carries out various R&D. We continue developing , the Brain-inspired Computing Architecture, as a technology for massive distributed processing in brain-like architecture. It also develops , the tool for monitoring cognitive architecture’s activity while mapping it onto connectome, and (Life in Silico: framework that makes intelligent agents live and learn in a virtual environment). The results have been used in the open development community as well as public and private research projects. We expect that these tools will be useful in areas where brain-like AI are applied.
Though various machine learning techniques are getting open, it requires knowledge and skills to integrate them into WBA. As this task is generally not so easy for engineers interested in AI, it would be valuable to have support from researchers by, for example, breaking down the construction of the entire architecture into development themes.
Instead of employing PIs (principal investigators), we are going to financially support researchers who recently started to work on the WBA approach.
In FY 2016, we shall be mainly engaged in the support of specific researches with PIs, entry level researchers and students (Fig. 3 γ), an open AI community activity centered around engineers (Fig. 3 α), and facilitative R&D lead by WBAI (Fig. 3 β), to accumulate R&D result in an open repository, so that we could pass it generally on to the society as well as to our stakeholders.