SYNAPSE: Systematic Neural-Symbolic Attn and Eval

chevron-icon
Back
Top
chevron-icon
project-presentation-img
simuliinc
Project Owner

SYNAPSE: Systematic Neural-Symbolic Attn and Eval

Status

  • Overall Status

    ⏳ Contract Pending

  • Funding Transfered

    $0 USD

  • Max Funding Amount

    $30,000 USD

Funding Schedule

View Milestones
Milestone Release 1
$8,000 USD Pending TBD
Milestone Release 2
$8,000 USD Pending TBD
Milestone Release 3
$8,000 USD Pending TBD
Milestone Release 4
$6,000 USD Pending TBD

Project AI Services

No Service Available

Overview

SYNAPSE (Systematic Neural-Symbolic Attention Processing System Evaluation Framework) develops a biologically-inspired framework for evaluating and optimizing attention allocation in modular cognitive systems, integrating neural-symbolic processing while maximizing resource efficiency and cognitive synergy. The framework incorporates key principles from cortical organization, memory systems, sleep-based optimization, and modular hierarchies.

RFP Guidelines

Framework for evaluating approaches to attention allocation

Complete & Awarded
  • Type SingularityNET RFP
  • Total RFP Funding $60,000 USD
  • Proposals 5
  • Awarded Projects 2
author-img
SingularityNET
Oct. 9, 2024

The goal of this project is to develop a framework to evaluate various approaches to Attention Allocation (AA) within the OpenCog Hyperon and PRIMUS architectures. The AA system dynamically allocates cognitive resources to Atoms in the Distributed Atomspace (DAS), and this framework will help assess AA methods based on desired cognitive dynamics. The framework will improve both Probabilistic Logic Networks (PLN) and evolutionary methods like Meta-Optimizing Semantic Evolutionary Search (MOSES), which are critical components of the PRIMUS architecture.

Proposal Description

Company Name (if applicable)

Simuli Inc

Project details

Problem Statement

Current cognitive architectures face several critical challenges in attention allocation and resource management:

Resource Management Challenges

  • Excessive resource consumption in neural-symbolic systems

  • Inefficient sharing of processing resources across modules

  • Poor balance between specialization and flexibility

  • Lack of biologically-inspired optimization mechanisms

  • Absence of effective pre-allocation strategies

Module Organization Difficulties

  • Inefficient module communication patterns

  • Suboptimal hierarchical structures

  • Poor management of information encoding units

  • Limited support for information synthesis in context

  • Inadequate monitoring of information flow management

Attention Control Issues

  • Ineffective dynamic resource allocation

  • Limited support for different representations of recallable information types

  • Poor integration of symbolic and neural processing

  • Inadequate novelty detection and prioritization

  • Lack of optimization phases of resources allocated to architecture supporting accurate context retrieval

Solution Overview

SYNAPSE addresses these challenges through a comprehensive, biologically-inspired framework. The framework establishes a perspective of viewing attention allocation techniques in terms of resource allocation. The framework includes a methodology of evaluation and measurement of attention allocation in any system capable of attention, focusing on AI and AGI models. The framework also embodies a exempliary archetype of how to manage attention in terms of resource allocation mechanisms for optimal balance between encoding useful information and retreiving information for the current task within the appropriate context.

SYNAPSE allows for effective attention to be measurable by monitoring and assesing the resource allocation protocols of a attention-enabled system. For attention to be effective, the system attends to all available data in any given condition and chooses which information will be encoded -- or stored -- in the systems. Discovering which information is pertinent relies in part on what information will be useful in another condition. Therefore, we describe effective attention as not only the ability to effectively take information into the system but also to recall pertinant information based on the context of the condition in which the information is useful to executing some task (internal or external).

In nature, evolutionary preasure of metabolic processes and the kluge nature of relying on previous genetic expressions as a basis for the next innovation of biological constraints creates a constraint on resources. This evolutionary pressure is exhibited in the human brain as the number of neurons, the shape, connections between them, and general morphology being limited. Brains cannot get bigger and bigger infinetly because the metabolic processes cannot support the maintence of the system. In L. Andrew Coward's theory of understanding higher order cognition in terms of anatomy and physiology -- The Recommendation Architecure -- this constraint dictates how brains were pressured into evolving to be adaptable to a breadth of tasks, learning new domain specific information without forgetting previously learned informaton, synthesising knowledge quickly in appropriate contexts, utilizing information to form behiavors quickly. Most importantly in this proposals context, the presure of having limited biological units forced brains to devlope mechanisms to make information encoding and recalling information to synthesize and utilize information in a mannor somewhat optimally useful to it's cognition and behavior. The afformentioned brain properties are still challenges for AI models and desirable properties in most visions of AGI.

The premis of SYNAPSE relies on viewing information stored in some discrete unit (e.g. weights in a DNN). The number of units available to store information is thus a resource. Further, the amount of information allowed to be stored by each unit is a property of the resource allocation protocol for gaining information. How information is represented in a collection of units is thus a function of how the resources are active during encoding information and retreival.

We first develop a literature review suitble for peer review of prior attention mechanisms implementations and measurement tools in AI/AGI followed by a relevant review of Recommendation Architecture and related neuroscince to motivate the SYNAPSE approach.

Then we develop the SYNAPSE framework with a user considerate roll-out plan to increase communication effectiveness and open source peer-review. The SYNAPSE framework includes documentation, tutorials, resources, and tools for information about the SYNAPSE core framework, how to use it to evaluate attention in AI/AGI, and how developers can use it as a guide to building more effective attention mechanisms.

We will develop SYNAPSE by relating biological processes for attention from a resource allocation perspective to a equivalent components and mechanisms in AI attention mechanism, which is then extended to a general systems architecture framework. The mapping from biological to generic complex systems is based on rigorous research and established literature.

Aspects of SYNAPSE

Module Efficiency Metrics

  • Information exchange ratio optimization (targeting >99.7% internal like cortical systems)

  • Resource sharing effectiveness across hierarchical levels (cortical column model)

  • Processing time for cross-module communications

  • Module specialization efficiency based on process types

  • Sleep-phase optimization cycles

Attention Allocation Metrics

  • STI variation measurements aligned with hippocampal novelty detection

  • Tononi Phi coefficient calculations for consciousness integration

  • Receptive field adaptation monitoring

  • Temporal correlation tracking across multiple timescales

  • Dynamic resource pre-allocation metrics

Integration Metrics

  • Cross-module information flow rates (following cortical column patterns)

  • Resource utilization efficiency during active and preparation phases

  • Learning transfer effectiveness across memory types

  • Cognitive synergy measurements between symbolic and neural processes

  • Provisional connection utilization rates

Modular Hierarchy

  • Implementation of strict biological ratios for internal vs external connectivity

  • Hierarchical organization following brain structure patterns

  • Process-type based module organization (not feature-based)

  • Dynamic connection creation based on usage patterns

  • Sleep-like preparation phases for resource optimization

Memory Integration

  • Support for five memory types (semantic, episodic, priming, working, procedural)

  • Temporal correlation-based connection management

  • Interference minimization through careful change management

  • Break periods for connection optimization

  • Novel experience prioritization

Resource Management

  • Provisional resource allocation during optimization phases

  • Dynamic resource sharing based on attention requirements

  • Careful management of receptive field changes

  • Pre-allocation of connections based on usage patterns

  • Resource conservation through modular specialization

Implementation Approach

Phase 1: Core Architecture (Month 1)

  • Establish modular hierarchy based on process types

  • Implement internal/external connectivity ratios

  • Create basic memory integration systems

  • Set up resource management framework

Phase 2: Optimization Systems (Month 2)

  • Develop sleep-like optimization phases

  • Implement temporal correlation tracking

  • Create provisional connection management

  • Establish interference minimization systems

Phase 3: Integration (Month 3)

  • Connect memory subsystems

  • Implement cross-module communication

  • Develop resource sharing mechanisms

  • Create evaluation metrics

Phase 4: Testing and Refinement (Month 4)

  • Comprehensive system testing

  • Performance optimization

  • Documentation completion

  • Delivery preparation

Success Metrics

  • Comprehensive documentation for different technical audience levels with a focus on organization and clarity

  • Successful examples peer-reviewed and aligned with existing literature

  • Easy to interact with and learn about the system set-up for users (i.e. user-interface)

  • Tutorials work and have received positive community feedback

  • Hyperon team verification that evaluation of attention in Hyperon with SYNAPSE is accurate, how to solve challenges in DAS with SYNAPSE, and integration plan on future hyperon attention mechanisms with SYNAPSE.

Competitive Advantages

  • Based directly on proven biological principles

  • Comprehensive understanding of resource allocation involved in effective memory at different levels of understanding (hardware, logic, architecture, etc.)

  • Able to apply to any attention enabled system

  • Scalable modular framework and evaluation across centralized or decentralized systems

  • Attention to organization and clarity of documention with ability to revise and update in open-source fashion

  • Strong theoretical foundation in neuroscience and AGI

This framework represents a significant advance in cognitive architecture design, incorporating key biological principles while maintaining practical implementation considerations. It allows a highly informed method of evaluating attention in such a way that evaluates the effectiveness as it relates to human-like cognition.

Open Source Licensing

GNU GPL - GNU General Public License

Proposal Video

Not Avaliable Yet

Check back later during the Feedback & Selection period for the RFP that is proposal is applied to.

Group Expert Rating (Final)

Overall

5.0

  • Feasibility 3.8
  • Desirabilty 3.8
  • Usefulness 3.7

New reviews and ratings are disabled for Awarded Projects

Overall Community

3.3

from 6 reviews
  • 5
    2
  • 4
    0
  • 3
    3
  • 2
    0
  • 1
    1

Feasibility

3.8

from 6 reviews

Viability

3.8

from 6 reviews

Desirabilty

3.7

from 6 reviews

Usefulness

0

from 6 reviews

Sort by

6 ratings
  • Expert Review 1

    Overall

    3.0

    • Compliance with RFP requirements 3.0
    • Solution details and team expertise 3.0
    • Value for money 3.0
    Provides details and metrics for evaluation, team possesses an experience in AA. Acceptable but questionable.

    Weakly Acceptable proposal! Proposal provides the details and metrics for evaluating AA and team has some experience in the field and references their previous research. This is the best proposal for the given RFP however it is questionable in the scope of PRIMUS framework since it is not a single cognitive system but a complex set of algorithms, systems and tools.

  • Expert Review 2

    Overall

    1.0

    • Compliance with RFP requirements 1.0
    • Solution details and team expertise 1.0
    • Value for money 1.0
    Attention allocation as resource allocation

    The author's view to see attention allocation as resource allocation is valid. However no details are given of how attention should work or be implemented.

  • Expert Review 3

    Overall

    5.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 5.0
    • Value for money 5.0
    A very strong proposal with a credible and original approach

    The high level approach here is to evaluate an attention allocation approach in terms of how well it displays certain attentional properties one would expect of a well-functioning "human-like cognitive architecture." This makes a lot of sense. It's an alternative and complement to the idea of measuring attention allocation effectiveness in a more general mathematical sort of way... but it makes total sense to play with this and the Simuli team seems competent to do it. So w00t ...

  • Expert Review 4

    Overall

    3.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 4.0
    • Value for money 4.0
    Biologically-inspired framework for attention allocation.

    It looks like an ambitious proposal, probably falling a bit outside of the scope of that RFP which is about developing a framework for evaluating attention allocation. Would lead to interesting research without a doubt though.

  • Expert Review 5

    Overall

    3.0

    • Compliance with RFP requirements 4.0
    • Solution details and team expertise 5.0
    • Value for money 4.0
    Ambitious proposal, unclear if budget is sufficient

    Proposal ambitiously spans too many dimensions (e.g., modular hierarchy, resource management, memory integration), which seems likely to dilute focus/outcomes. Unclear if team can execute all aspects within the proposed timeline​​/budget.

  • Expert Review 6

    Overall

    5.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 5.0
    • Value for money 5.0

    Comprehensive biologically-inspired evaluation framework.

  • Total Milestones

    4

  • Total Budget

    $30,000 USD

  • Last Updated

    3 Feb 2025

Milestone 1 - Core Architecture Foundation and Development Plan

Status
😐 Not Started
Description

Motivate SYNAPSE approach that attention management is fundamentally a resource allocation system. Establish the fundamentals of SYNAPSE as a framework for guiding and evaluating attention in AI and AGI. Start to convert biological principals of recommendation architecture theory to tools of the framework.

Deliverables

Initial documentation motivating theory behind the basic premise of resource allocation in relation to attention. Key approach of SYNAPSE explained with graphics. Identification of core components in biologic systems resource management. Identification of attention allocation in terms of resource management components in AI/AGI at different levels of understanding (e.g. compute resources logic knowledge). Mapping biologic components to ideal components in AI/AGI models that would exemplify benefits of resource allocation in brain with respect to attention. Section organization outlines for the SYNAPSE about document how-to evaluate user guide document guide to developing based on the SYNAPSE model document.

Budget

$8,000 USD

Success Criterion

Literature review of the current attention management techniques in AI/AGI and the relation to recommendation architecture theory. 3 documents with clear organizational structure of headings, subheadings, brief description of each section, and graphic/tool placement for how to use SYNAPSE to evaluate AI/AGI attention, how to model a new attention mechanism after SYNAPSE and about the SYNAPSE (i.e. what the framework is, literature support, framework documented and explained). The core framework for SYNAPSE established in the about documentation. Report documenting the development and research plan of SYNAPSE framework to be completed in this proposal from steps needed to develope the framework -- then designing how to use the framework to evaluate attention in any system (specifically focusing on AI and AGI systems) --and then finally developer tools and tips for implementing SYNAPSE principals an best practices. Report documenting how the SYNAPSE framework will be communicated to and interact with AI researchers. A plan to organize what tools and resources will be available and in what format to communicate this framework. A peer reviewed paper from a technical standpoint, documentation for AI developers, resources for the general public, and the user interface housing the different levels of communication that different audiences can provide feedback and updates/future work to the framework.

Link URL

Milestone 2 - Full SYNAPSE Framework

Status
😐 Not Started
Description

Finalize the mapping of brain mechanisms to framework of SYNAPSE -- core construction of the SYNAPSE concept which establishes how resource allocation is attention management in a framework that is architecture and substrate independent. This means that the framework is a general design schematic of how any system capable of attention is a function of modular resource allocation concepts. The SYNAPSE framework embodies a best-practices example of resource allocation mechanisms for maximum information encoding and recall in appropriate contexts. Execution of some of the communication report.

Deliverables

Completed about documentation updated based on feedback from related experts on the Recommendation Architecture AI and AGI experts. Set-up of user-interface and roll-out of about documentation including other levels of communication according to the communication report.

Budget

$8,000 USD

Success Criterion

A user-interface is set-up properly and ready for rolling out documentation with internal testing completed. Execution of communicating the SYNAPSE framework is deployed and maintained properly. Feedback on SYNAPSE Framework is positive and the initial sentiment of the communities are encouraging. Feedback is responded to and incorporated in revised-maintained version of the framework documentation, resources, and tools.

Link URL

Milestone 3 - SYNAPSE Evaluation Documentation

Status
😐 Not Started
Description

Design the approach for using the SYNAPSE framework to evaluate attention mechanisms in AI/AGI models.

Deliverables

Complete documentation of evaluating attention with SYNAPSE. This includes a description of evaluation metrics and how they will be used/applied. Followed by a quick guide "how-to" describing the overall process. Then a comprehensive guide of evaluating attention abilities and effectiveness at different levels of resource allocation (e.g. model architecture model logic mechanisms compute resources on the hardware device(s) information theory level etc.). Including examples and practice tutorials. Including a detailed description and definition of the components and mechanisms needed to identify and measure attention. Including a detailed description of how measurements are conducted. Including a long description of how to interpret the measurements of attention. Roll-out of associated tools and resources for this phase of the communication plan. Feedback from many more experts in related scientific fields and from developers and documentation designers. Report on examples of evaluating Attention mechanisms using SYNAPSE in existing AI models

Budget

$8,000 USD

Success Criterion

Execution of communicating the evaluation framework for SYNAPSE is deployed and maintained properly. Feedback on SYNAPSE core framework and evaluation methods update is positive and the sentiment of the communities remains encouraging. Feedback is responded to and incorporated in revised-maintained version of the framework documentation, resources, and tools. Report on examples of evaluating Attention mechanisms using SYNAPSE in existing AI models aligns with previous research and the report is peer-reviewed. Any qualified feedback is used to revise the documentation.

Link URL

Milestone 4 - Developer Tools and Final Delivery

Status
😐 Not Started
Description

Comprehensive testing performance optimization and completion of all documentation and delivery requirements. Finalize how to use SYNAPSE in development for guiding and evaluating attention mechanisms in AI/AGI.

Deliverables

Documentation of developer guide for SYNAPSE is completed with tutorials and examples and extra information. Complete documentation of all resources/tools/documents that were designed according to plan deployed on user interface and organized each with appendices and table of contents for easy look up. Final reviews by many reviewers used to update any of the documentaton or framework. Report on using SYNAPSE as a guide to fixing problems in Hyperon DAS evaluation of attention in current Hyperon/PRIMUS suggestions for monitoring and implementing attention solutions as Hyperon scales.

Budget

$6,000 USD

Success Criterion

Execution of communicating the developer guide framework for SYNAPSE is deployed and maintained properly. Feedback on SYNAPSE core framework and evaluation methods and developer guide update is positive and the sentiment of the communities remains encouraging. Feedback is responded to and incorporated in revised-maintained version of the framework documentation, resources, and tools. Report on using SYNAPSE as a guide to fixing problems in Hyperon DAS, evaluation of attention in current Hyperon/PRIMUS, suggestions for monitoring and implementing attention solutions as Hyperon scales are reviewed and revised by the Hyperon team.

Link URL

Join the Discussion (0)

Expert Ratings

Reviews & Ratings

Group Expert Rating (Final)

Overall

5.0

  • Feasibility 3.8
  • Desirabilty 3.8
  • Usefulness 3.7

New reviews and ratings are disabled for Awarded Projects

  • Expert Review 1

    Overall

    3.0

    • Compliance with RFP requirements 3.0
    • Solution details and team expertise 3.0
    • Value for money 3.0
    Provides details and metrics for evaluation, team possesses an experience in AA. Acceptable but questionable.

    Weakly Acceptable proposal! Proposal provides the details and metrics for evaluating AA and team has some experience in the field and references their previous research. This is the best proposal for the given RFP however it is questionable in the scope of PRIMUS framework since it is not a single cognitive system but a complex set of algorithms, systems and tools.

  • Expert Review 2

    Overall

    1.0

    • Compliance with RFP requirements 1.0
    • Solution details and team expertise 1.0
    • Value for money 1.0
    Attention allocation as resource allocation

    The author's view to see attention allocation as resource allocation is valid. However no details are given of how attention should work or be implemented.

  • Expert Review 3

    Overall

    5.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 5.0
    • Value for money 5.0
    A very strong proposal with a credible and original approach

    The high level approach here is to evaluate an attention allocation approach in terms of how well it displays certain attentional properties one would expect of a well-functioning "human-like cognitive architecture." This makes a lot of sense. It's an alternative and complement to the idea of measuring attention allocation effectiveness in a more general mathematical sort of way... but it makes total sense to play with this and the Simuli team seems competent to do it. So w00t ...

  • Expert Review 4

    Overall

    3.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 4.0
    • Value for money 4.0
    Biologically-inspired framework for attention allocation.

    It looks like an ambitious proposal, probably falling a bit outside of the scope of that RFP which is about developing a framework for evaluating attention allocation. Would lead to interesting research without a doubt though.

  • Expert Review 5

    Overall

    3.0

    • Compliance with RFP requirements 4.0
    • Solution details and team expertise 5.0
    • Value for money 4.0
    Ambitious proposal, unclear if budget is sufficient

    Proposal ambitiously spans too many dimensions (e.g., modular hierarchy, resource management, memory integration), which seems likely to dilute focus/outcomes. Unclear if team can execute all aspects within the proposed timeline​​/budget.

  • Expert Review 6

    Overall

    5.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 5.0
    • Value for money 5.0

    Comprehensive biologically-inspired evaluation framework.

feedback_icon