The NEKO Project is an open source effort to build a model of equivalent scale and capability as that reported in DeepMind’s 2022 Paper, A Generalist Agent. This constitutes the first major step in a longer goal of building multimodal, multiobjective models that work well across a variety of domains.
Ultimately, collective high performance across many objectives in varied modalities constitutes a SoTA direction. We hope that in building such models and making them open source, we can move humanity’s collective understanding and capability in building complex AI systems forward.
Practically, building these systems provides us deeper understanding into the utility and mechanics of modern deep learning approaches in building better representational capability and scaling to new problem domains. Put another way, this research direction serves as an excellent foundation to test other, new kinds of general AI approaches to build next generation learning systems.