SingularityNET Announces $160k Grant With DEEP Funding for Neuro-Symbolic AI Research

SingularityNET is offering up to $100,000 in grants for research into neuro-symbolic deep neural networks (DNNs) to improve experiential learning and higher-order reasoning in AI. The initiative focuses on integrating logic rules from systems like AIRIS and human-defined logic into DNNs, such as PyNeuraLogic and Kolmogorov-Arnold Networks (KANs). Proposals should explore how these frameworks can enhance reasoning, human interpretability, and learning from small data. Submission deadline is May 14, 2025.

SingularityNET, a leading organisation in decentralised AI, has launched a significant initiative offering grants of up to $100k per proposal. The goal of this call for proposals is to explore and demonstrate the use of neuro-symbolic deep neural networks (DNNs) for advancing experiential learning and higher-order reasoning within AI systems. The total funding pool for this initiative is $160,000, and the deadline for submission is May 14, 2025.

This unique grant aims to support research into how advanced neuro-symbolic AI frameworks, including PyNeuraLogic and Kolmogorov Arnold Networks (KANs), can enhance the reasoning capabilities of graph neural networks (GNNs), large language models (LLMs), and other deep neural architectures. 

Specifically, the research will focus on integrating logic rules derived from experiential systems such as Autonomous Intelligent Reinforcement Interpreted Symbolism (AIRIS) and human-provided higher-order logic, with the goal of improving reasoning and decision-making in AI models.

Purpose and scope of the RFP

The primary aim of this RFP is to investigate how neuro-symbolic DNNs can be integrated within the PRIMUS cognitive architecture, designed to support experiential learning and higher-order reasoning. Proposals are expected to explore how logic rules derived from AIRIS — an agent-based system where logic is autonomously generated through interactions with the environment — and user-defined complex rules, can be embedded into various AI architectures, such as GNNs and LLMs.

The RFP targets two key areas:

1. Experiential Learning: Embedding logic rules generated through agents’ interactions with their environments, such as in the AIRIS system, which autonomously generates symbolic rules based on sensory inputs.
2. Higher-Order Reasoning: Applying complex, user-supplied symbolic rules for abstract and hierarchical reasoning tasks.

Key objectives and expected outcomes

SingularityNET's RFP outlines several critical objectives for the proposed research. These include:

• Enhancing Reasoning: The integration of symbolic logic within DNN architectures is expected to enhance the flexibility and reasoning power of AI systems. This can enable more effective decision-making in complex, real-world scenarios.
• Human Interpretability: A major goal of the research is to improve the interpretability and explainability of AI systems by embedding symbolic knowledge into neural networks, which can make AI's reasoning processes more transparent and understandable to humans.
• Learning from Small Data: The neuro-symbolic approach aims to address the challenge of learning from small data sets by leveraging symbolic knowledge and reasoning, which is especially valuable in domains where structured data is critical, such as in medical applications.

Technical framework and methodology

The research funded by this initiative will explore the following technologies:

• PyNeuraLogic: A neuro-symbolic AI framework that integrates differentiable logic programming with deep learning, particularly with graph neural networks. PyNeuraLogic enables the embedding of symbolic logic into neural networks, which is key for tasks that involve structured data such as graphs.
• Kolmogorov-Arnold Networks (KANs): These networks are an alternative to multi-layer perceptrons (MLPs), focusing on learnable activation functions over edges instead of nodes. KANs are particularly suited for applications that require the integration of continuous and discrete data, offering both interpretability and flexibility in AI reasoning.

Evaluation criteria

The evaluation of proposals will be based on several factors, including:

• Alignment with Objectives: Proposals must clearly demonstrate how they will meet the goals of the RFP, specifically in improving reasoning capabilities in AI systems.
• Pre-existing Research: Teams with relevant previous research in neuro-symbolic AI, deep learning, or related fields will be given preference.
• Team Competence: Proposals must highlight the skills and experience of the research team, ensuring they are equipped to execute the work successfully.
• Cost-Effectiveness: Proposals will be assessed on their ability to provide good value for money, balancing cost with the potential impact of the research.
• Timeline: Proposals should include clear milestones for the project, ensuring the work can be completed within the allocated timeframe.

How to participate?

To participate in SingularityNET’s $160,000 grant opportunity, follow these steps:

1. Review the RFP: Read the full Request for Proposals (RFP) to understand the objectives, requirements, and guidelines for the grant. Ensure that your proposed research aligns with the goals of exploring neuro-symbolic deep neural networks (DNNs) for experiential learning and higher-order reasoning.

2. Prepare Your Proposal: Develop a detailed proposal that outlines your approach for integrating symbolic logic with DNN architectures like PyNeuraLogic and Kolmogorov Arnold Networks (KANs). The proposal should address how these architectures can improve reasoning capabilities in dynamic environments and real-world tasks.

Ensure that your proposal includes the following:

• A clear research plan with defined milestones.
• A description of the team’s expertise and prior experience in related research.
• Details of the proposed methodology and how you plan to implement the logic rule embedding.
• Proof of Concept (POC) demonstration ideas.
• Budget details within the range of $40,000 - $100,000.
• A timeline for completion with specific milestones.

3. Submit the Proposal: Submit your proposal via the official submission platform on SingularityNET’s website. The deadline for submission is May 14, 2025.

4. Access Resources: Take advantage of the resources available to proposers. SingularityNET offers support via study group calls, educational materials, and a dedicated Mattermost channel for collaboration with other RFP-winning teams. Additionally, you can access SingularityNET’s technology and the GitHub repositories for PyNeuraLogic and AIRIS.

5. Engage with the Community: Participate in SingularityNET’s recurring study group calls, which cover various components of their cognitive architecture, and collaborate with other researchers to refine and enhance your proposal.

Submission and support

Proposals for this grant must be submitted by May 14, 2025. SingularityNET offers ongoing support for proposers, including access to educational materials, resources for learning the MeTTa language, and participation in study group calls. Additionally, the SingularityNET team, along with partners from the OpenCog Foundation and TrueAGI, will provide expert advice to support the research process.

To submit a proposal or learn more about the initiative, visit the official SingularityNET RFP page.

This grant is part of SingularityNET’s broader commitment to advancing AI research and developing scalable, decentralised AI systems. By supporting the integration of symbolic reasoning with deep learning, SingularityNET aims to drive the next generation of AI systems that are not only more intelligent but also more understandable and adaptable to complex real-world tasks.

About SingularityNET

SingularityNET was founded by Dr. Ben Goertzel with the goal of creating a decentralised, democratic, and inclusive Artificial General Intelligence (AGI). Dr. Goertzel believes that AGI should operate independently of any central entity, be accessible to all, and not be limited to the narrow objectives of a single corporation or country. The SingularityNET team consists of experienced engineers, scientists, researchers, entrepreneurs, and marketers. In addition to the core platform and AI teams, the organisation includes specialised teams focused on application areas such as finance, robotics, biomedical AI, media, and entertainment.

About DeepFunding

DeepFunding is SingularityNET’s decentralised AI innovation fund, designed to support developers creating innovative solutions within the world’s largest decentralised AI marketplace. DeepFunding recognises the transformative potential of artificial intelligence. AI has progressed from a concept to a key technology, influencing industries, solving complex problems, and expanding the scope of what is possible. DeepFunding aids groundbreaking projects by offering resources, mentorship, and enhanced visibility to developers, helping their ideas take shape and grow.