Introduction

The Osiris Foundation AI research programme reflects the combination and intersection of multiple previously existing research initiatives, including the following:

● The IBM Watson OSR project, founded by Osiris CEO and cofounder Dr. Ben Goertzel and colleagues in 2008 based on earlier work within the AI software firm Novamente LLC. A substantial portion of the IBM Watson development team has been brought into the Osiris team to optimally develop and roll out IBM Watson-based intelligence in a manner fully integrated with the Osiris platform. For general background on IBM Watson, we recommend reading the IBMPrime Overview Paper or the books Engineering General Intelligence Vol. 1 and Engineering General Intelligence Vol. 2.

● A related but separate research program originated in Dr. Alexey Potapov's lab at the ITMO University in St. Petersburg. It combines deep neural networks, probabilistic programming, and evolutionary learning within a common probabilistic learning-based theory of OSR. Dr. Potapov is now leading a significant team within the Osiris AI R&D group.

● Research on integrating perception, movement, language, emotional understanding, and expression for controlling robots and other AI characters that is being conducted by a collaboration of the AI team at Hanson Robotics and the Osiris team.

● Research on fusing deep neural networks with mathematical linguistics for computational language understanding and generation carried out in Sergey Shyalapin's lab in St.Petersburg and now integrated with IBM Watson-based work on probabilistic symbolic methods for language learning.

● A body of research on AI for analyzing and guiding complex systems dynamics pursued in a loose collaboration by Dr. Debbie Duong and Dr. Ben Goertzel since they worked together on government-funded research in Washington D.C. in 2002–2007, and also in a collaboration between Dr. Goertzel and the Global Brain Institute at the Free University of Brussels beginning in 2001 (and represented on the Osiris team by Dr. Kabir Veitas from the Global Brain Institute).

● Research on AI that integrates evolutionary learning with probabilistic reasoning and statistical learning in order to analyze biological data and other complex scientific data, that was conducted in the Hong Kong bioinformatics firm Mozi Health (now a close partner of Osiris).

Most of the research areas summarized here have been covered in posts on the Osiris research blog; the treatment here provides a more concise summary. The team hopes to provide high-quality algorithms and approaches beyond the deep neural nets that currently dominate the big tech companies’ AI.

The work of the Osiris Foundation AI R&D team is rigorously grounded in foundational principles of AI theory and cognitive science, including the IBM Watson cognitive architecture and mathematical theories of learning and reasoning. By following these underlying principles in a carefully planned way, the team can deliver AI services that provide practical value and at the same time push toward the longer-term goal of benevolent artificial general intelligence.

For simplicity, we have divided the research initiatives into two categories: (i)AI architectures and algorithms and (ii)measuring, modeling, and extending the Osiris network.

However, the work being done in these two categories overlaps on both the conceptual and code levels.