chevron-icon
Active RFP

Framework for evaluating approaches to attention allocation

Top
chevron-icon
SingularityNET
RFP Owner

Framework for evaluating approaches to attention allocation

Create a conceptual framework for evaluating approaches to attention allocation

  • Type SingularityNET RFP
  • Total RFP Funding $60,000 USD
  • Proposals 1
  • Awarded Projects n/a

Overview

  • Est. Complexity

    💪 75/ 100

  • Est. Execution Time

    ⏱️ 4 Months

  • Proposal Winners

    🏆 Multiple

  • Max Funding / Proposal

    $30,000USD

RFP Details

Short summary

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.

Main purpose

The purpose of this RFP is to develop a framework for evaluating Attention Allocation (AA) approaches within the OpenCog Hyperon and PRIMUS architectures. In AGI, AA is vital for efficiently allocating tasks to the most appropriate algorithms, ensuring that reasoning, learning, and problem-solving processes are handled by the best-suited methods. By optimizing task distribution across algorithms like PLN and MOSES, the framework will enhance the system's adaptability, performance, and overall cognitive efficiency​​. This includes examining how attentional processes can be effectively integrated into more complex systems that also involve neural networks.

Long description

Context and background: 

SingularityNET Foundation, in collaboration with other partners such as the OpenCog Foundation and TrueAGI, is working toward a scalable implementation of the Hyperon AGI framework running on decentralized infrastructure, and toward implementation of the PRIMUS cognitive architecture within this framework.

OpenCog Hyperon involves multiple algorithms and components, which need to demonstrate appropriate functionalities both individually and in combination. This RFP aims to address a portion of this overall need, via funding the initial iteration of one significant component of OpenCog Hyperon.

A crucial component of cognitive synergy is a system for allocating attention to Atoms in the Distributed Atomspace (DAS) in order to efficiently allocate system resources. The goal of this project is to create a framework for evaluating different Attention Allocation (AA) approaches by assessing their dynamics within Hyperon. 

In addition to cognitive components like PLN and MOSES, the framework should also consider how attentional processes operate within systems that incorporate neural networks. Given that neural networks introduce sub-symbolic processing dynamics, it will be important to explore how AA can manage resources between symbolic systems (e.g., PLN, MOSES) and neural architectures. Attentional dynamics in hybrid symbolic-neural systems must be considered for their efficiency, adaptability, and relevance in complex problem-solving scenarios.


Possible questions to ask in constructing such a measurement framework could include: 

  • What quantity or quantities should we measure?
  • How should such measures react to environmental changes?
  • What sorts of attention dynamics would be desired?
  • How do we align measure(s) with desired dynamics?
  • Others?

Below we provide information about AA approaches and measurements previously constructed within OpenCog Classic. In this RFP, we are open to new and creative ideas to better assess AA dynamics within Hyperon. In this sense, the RFP is quite wide open. 

The Economic Attention Networks (ECAN) module (https://agi-conf.org/2009/papers/paper_63.pdf) fulfilled the AA role within OpenCog Classic. As demonstrated in (https://www.researchgate.net/publication/304459082_Controlling_Combinatorial_Explosion_in_Inference_via_Synergy_with_Nonlinear-Dynamical_Attention_Allocation) and https://www.researchgate.net/publication/383696770_Using_Nonlinear_Dynamical_Attention_Allocation_to_Focus_Probabilistic_Logical_Inference_Upon_Relevant_Information#fullTextFileContent

ECAN effectively selected relevant premises for the Probabilistic Logic Networks (PLN) uncertain inference system to perform reasoning upon and come to better conclusions, within the context of OpenCog Classic. 

What is needed for OpenCog Hyperon is a similar, but perhaps broader, system able to measure AA cognitive dynamics across the entire PRIMUS cognitive architecture including PLN and evolutionary methods such as the Meta-Optimizing Semantic Evolutionary Search (MOSES) procedural learning algorithm (https://www.cs.york.ac.uk/rts/docs/GECCO_2007/docs/p626.pdf). 

However, in order to develop and evaluate various approaches to creating such a “similar, but perhaps broader” system, it will be valuable to have a systematic framework for evaluating attention allocation approaches in a Hyperon and PRIMUS context.


Collaboration:

This RFP may be followed by subsequent RFPs for applications that leverage Hyperon/PRIMUS to carry out various applications, and that aim to guide Hyperon/PRIMUS systems in cognitive development toward beneficial AGI.

 

RFP Expected Outcomes:

  • Primary outcome would be a framework in which to evaluate various approaches to AA. Some possible ideas to include and questions to address in creation of the framework are:

Functional Requirements

Must Have

  • Comparative Framework: Provide a structured method for comparing and evaluating different AA approaches, based on theoretical pros and cons​​.
  • Hierarchical Orders of Attention: Incorporation of attentional processes that span both symbolic systems (PLN, MOSES) and neural networks, considering mechanisms like Hebbian learning, but not limited to it.
  • Literature Review Support: Reference existing studies and models (e.g., ECAN) to guide evaluation and comparison​​.
  • Custom Evaluation Criteria: Allow researchers to define criteria like efficiency, adaptability, and resource management for evaluating AA approaches​​.
  • Structured Analysis: Facilitate systematic documentation of the strengths and weaknesses of each approach​​.

Should Have

  • Research-Based Recommendations: Offer guidelines based on past research to help researchers assess AA strategies​​.
  • Guided Evaluation Questions: Include prompts to help researchers think through important aspects of AA, like task allocation and cognitive load​​.

Could Have

  • Study Proposal Template: Provide a template for proposing future research or experiments based on findings​.
  • Visualization Tool Integration: Optional integration with external tools for visualizing theoretical AA dynamics​.

Won’t Have Yet

  • No Simulations or Data Output: This framework will not simulate systems or generate data; it is purely for comparative analysis​

Non-functional Requirements

  • Comprehensive Citations and References: The framework must include thorough citations, linking to foundational studies, papers, and models (e.g., ECAN in OpenCog Classic) to support the analysis of different AA approaches. This will ensure that the evaluation is grounded in relevant, high-quality research​​.
  • Support for Theoretical Models: The framework should enable researchers to incorporate and compare theoretical models of AA from existing cognitive architectures (e.g., PLN, MOSES), fostering in-depth evaluation without requiring empirical data​.
  • Structured Analytical Approach: The framework must follow a systematic and structured approach for analyzing AA methods, making it easy to assess theoretical strengths and weaknesses across multiple criteria such as scalability, adaptability, and efficiency​​.
  • Cross-Disciplinary Insights: Should integrate relevant insights from fields like cognitive science and neuroscience to enrich AA evaluation​​.

Main evaluation criteria

Alignment with requirements and objective

  • Does the proposal meet the requirements and advances the objectives of the RFP

Pre-existing R&D

  • Has the team previously done similar or related research or development work in other platforms / languages / contexts?

Team competence

  • Does the team have relevant skills?

Cost

  • Does the proposal offer good value for money?

Timeline

  • Does the proposal include a set of clearly defined milestones?

Other resources

Hyperon and related AI-platforms are quickly evolving! This is a bit of a moving target, but the internal SingularityNET team will be available for help and expert advice, where needed. Also included:

  • SingularityNET technology links
  • Educational materials and resources for learning MeTTa
  • SingularityNET holds MeTTa study group calls every other week. Proposers are welcome to attend for support from our researchers and community.
  • Recurring Hyperon study group calls for community are currently being planned. These will cover MOSES, ECAN, PLN, and other key components of the OpenCog and PRIMUS Hyperon cognitive architectures.
  • Access to the SingularityNET World Mattermost server, with a dedicated channel for discussion and support among the RFP-winning teams and SingularityNET resources.

To provide additional context on the sorts of cognitive processes we are ultimately interested in including in measures of effectiveness for AA, we list below a sampling of additional background resources, mostly from OpenCog Classic. 

  • Guiding Probabilistic Logical Inference with Nonlinear Dynamical Attention Allocation
    • This demonstrates how AA using ECAN effectively helped PLN perform probabilistic inference.
  • Speculative Scientific Inference via Synergetic Combination of Probabilistic Logic and Evolutionary Pattern Recognition"
    • This paper demonstrates using probabilistic logical reasoning via PLN (based on declarative knowledge) to generalize procedural knowledge gained by evolutionary program learning via MOSES.
  • Using Tononi Phi to Measure Consciousness of a Cognitive System While Reading and Conversing
    • In which we conducted computational experiments estimating Giulio Tononi’s Phi coefficient to measure the integrated information within the OpenCog cognitive architecture on two types of tasks: Reading (i.e. parsing and semantically analyzing) short documents, and guiding the Sophia humanoid robot in carrying out a dialogue-based interaction. Qualitative (and preliminary), comparison of the variation of Phi with cognitive system behavior over time reveals sensible patterns.
  • Measuring Sophia Robot's Cognitive Dynamics
    • We conducted a series of experiments within the ECAN system governing the attentional cognitive dynamics for Hanson Robotics Sophia robot. We logged short-term importance values (STI) within Sophia’s ”Economic Attention Network” attentional system and then empirically derived Tononi Phi values from the resulting STI time series. We compared the Phi value time-series with Sophia’s current cognitive actions to determine the degree of connectedness in the system, which has been hypothesized to be a neural correlate of consciousness (NCC.)
  • Nonlinear-Dynamical Attention Allocation via Information Geometry
    • We conduct a preliminary investigation of the application of information-geometry based learning to ECAN (Economic Attention Networks), the component of the integrative OpenCog AGI system concerned with attention allocation and credit assignment. We apply Amari’s ”natural gradient” algorithm for network learning to small example cases of ECAN, demonstrat- ing a dramatic improvement in the effectiveness of attention allocation compared to prior (Hebbian learning like) ECAN methods.
  • Toward a Formal Model of Cognitive Synergy
    • "Cognitive synergy" refers to a dynamic in which multiple cognitive processes, cooperating to control the same cognitive system, assist each other in overcoming bottlenecks encountered during their internal processing.
  • Wiki sites

RFP Status

Submit Proposals for

Days
Hours
Minutes
Seconds
Ends: 1 Dec. 2024 12:00 UTC Submit Proposal
1 proposals
rfp=proposal-img

Sphere Attention Allocation

  • Type SingularityNET RFP
  • Funding Request n/a
  • RFP Guidelines Framework for evaluating approaches to attention allocation
author-img
Tom M
Nov. 2, 2024
0 projects

No Projects Avaliable

Check back later by refreshing the page.

Join the Discussion (0)