Frequently Asked Questions
What do you mean by neuro-symbolic AGI?
We believe that agent systems are the future of general artificial intelligence and robotics. Our goal is to build an agent system that solves real-world problems by using an intermediary language interpretable by both humans and machines. If we want to keep humans in the loop in the coming years, we need to design AI systems for that purpose. Coming from a robotic background, we understand how the debate in the ML community between multi-agent systems and neuro-symbolic ones will unfold.
How is HybridAGI different from a toolbox?
Our aim is to develop an entire ecosystem around our technology. Like an autonomous car is composed of several sub-systems, robotic softwares are composed by different and complementary sub-systems coordinating themselves. With neuro-symbolic systems, you need to train your neural networks with the whole architecture to increase efficiency by taking into account the contraints and new knowledge created by this neuro-symbolic architecture.
Why HybridAGI is superior to React-based architectures (and by extension to LangChain/LangGraph or Llama-index)?
By controlling the behavior of the system in a end-to-end fashion, without giving the choice of the tool to use at each step, we can actually build explainable agent sytems that behave as expected. Moreover because we don't let the agent choose the tool to use, we can use an infinite number of tools as the prompt to infer the parameters of the tool is only provided when needed. This innovative approach is superior to the current trend of agent-based softwares that for the sake of simplicity sacrified efficiency and control.
What is the benefit of using HybridAGI over DSPy with Llama-index?
HybridAGI is specifically tailored for building interactive and reasoning agents quickly and effortlessly. The DSL allows for algorithmic flexibility while making it possible to describe every type of system without having to implement it from scratch. Plus, we focus our work on an open-source vector/graph database, allowing people and businesses to maintain control of their data. Moreover HybridAGI have been designed to use knowledge graphs not only to store triplets but also for planning purposes providing an holistic approach to Agent reasoning.
Can HybridAGI be used for tasks other than robotics?
Yes, HybridAGI can be used for a wide range of tasks beyond robotics. The system is designed to be flexible and adaptable, making it suitable for any application that requires complex reasoning and decision-making, such as retrieval-augmented generation (RAG), chatbots, knowledge scrapers, personal assistants, and more generally, any agent-based application.
How does HybridAGI help me regarding the EU AI Act or future regulation on your country for non-EU?
Because this system can only execute actions that are in the graph, businesses can use the graph to classify the behavior of their AI system and document it. However, this is not sufficient, and you should always conduct specific safety tests in accordance with the safety practices of your domain, in particular red-teaming of the model. We plan to release tools to help regarding these aspects.
How can I contribute to HybridAGI?
We encourage you to join our community on Discord to connect with other developers and share your ideas.