From Classical AGI on Quantum HW to Native QAGI

chevron-icon
Back
Top
chevron-icon
project-presentation-img
M T Bennett
Project Owner

From Classical AGI on Quantum HW to Native QAGI

Status

  • Overall Status

    ⏳ Contract Pending

  • Funding Transfered

    $0 USD

  • Max Funding Amount

    $80,000 USD

Funding Schedule

View Milestones
Milestone Release 1
$20,000 USD Pending TBD
Milestone Release 2
$20,000 USD Pending TBD
Milestone Release 3
$20,000 USD Pending TBD
Milestone Release 4
$20,000 USD Pending TBD

Project AI Services

No Service Available

Overview

We are researchers from ANU (PhD) and Stanford (postdoc), with expertise in AGI and quantum computing. We propose a review of AGI, quantum information processing (QIP) and quantum hardware substrates to determine optimal pathways for quantum AGI. We will provide near term practicable, commercialisable insights and a long-term rigorous Theory of Quantum General Intelligence, centred on Hyperon. Via theory and simulations, we will evaluate both how classical AGI algorithms designed for classical computers can take advantage of quantum computing hardware, and conversely how natively quantum AGI can take advantage of quantum computing in ways classical AGI designed for classical hardware cannot.

RFP Guidelines

Review of quantum computing technologies

Complete & Awarded
  • Type SingularityNET RFP
  • Total RFP Funding $80,000 USD
  • Proposals 10
  • Awarded Projects 1
author-img
SingularityNET
Oct. 4, 2024

This RFP seeks to critically evaluate the role of quantum computing in advancing Artificial General Intelligence (AGI). The goal is to distinguish between realistic capabilities and hype, providing clear insights into the practical benefits and limitations of quantum computing for AGI architectures, particularly within the OpenCog Hyperon framework. Part of this should involve interacting with the Hyperon team who've built the existing and in-development MeTTa interpreters.

Proposal Description

Project details

Our Quantum AGI project (QAGI) will produce a report into, empirically-validated simulations relating to and research articles examining practical options for the implementation of artificial general intelligence (AGI) models using quantum information technology, specifically scalable fault-tolerant quantum computers. 

This requires a structured approach that distinguishes between classical and quantum computing and identifies the unique impacts of quantum computing on AGI. 

Current proposals for AGI can be described as classical AGI (based on the tenets of classical computation) implemented according to classical computational paradigms (and on classical hardware). There is also nascent research into the effects of implementing classical AI algorithms on quantum computers or hybrid-quantum devices, examining how such devices, may change the affordances and capabilities of classical AGI models (affecting scale, speed, accuracy or resource use).  

However, rigorous theoretical research on AGI and QIP is yet to be undertaken, and the impact of QAGI extends beyond these two scenarios to ask whether there is a natively quantum paradigm of AGI, whose affordances exceed those of classical AGI and within which classical AGI is a subsidiary theory (akin to classical physics being a limit of quantum physics).  

One can then study the consequences (and limitations) of implementing QAGI theories on both classical and quantum devices – albeit on the assumption of the Extended Turing Thesis, QAGI’s could only be efficiently simulated on quantum computers themselves. These questions can be summarised via our taxonomy:

(1) classical AGI algorithms on classical computers;

(2) classical AGI algorithms on quantum computers;

(3) QAGI algorithms on classical computers; and

(4) QAGI algorithms on quantum computers. 

This follows similar approaches in the literature when comparing quantum and classical forms of algorithm, machine learning, control, measurement and communication. Our focus will be upon (i) classical AGI algorithms on quantum computers and (ii) QAGI algorithms on quantum computers. A solid understanding of the theoretical implications of these two approaches is a necessary condition for understanding to what extent candidate quantum architectures may realise QAGI.

Our study will examine the three sectoral components of QIP: (i) quantum computing, (ii) quantum communication and (iii) quantum sensing, albeit heavily weighted towards quantum computing. The former requires the latter two. For example, for a self organising, multi-agent system like Hyperon to take full advantage of quantum computing, it must account for quantum communication. Quantum communication is essential for distributed quantum systems (for exchange of quantum information and secure post-quantum cryptographic registers). Similarly, in principle something approaching a maximal information-seeking QAGI will interact with the world in part via quantum sensing (indeed a truly QAGI will necessarily be a quantum sensing device, measuring data in ways inaccessible to classical measurement devices). 

 

1. Comparison of AGI theories

We begin with a survey of classical theories of AGI, which will inform our determination of which theories are implementable on a quantum computer, and the dilemmas that may arise. 

We will focus upon Hyperon, its core Atomspace metagraph and MeTTa programming language. Quantum information systems are naturally and necessarily graphical in nature (as seen via tensor network formalism). Any QAGI proposal must engage with the formalism of AGI candidate theories. Our analysis specifically aims to understand such AGI paradigms via (i) their implementation in QIP terms and (ii) what quantum extensions of such formal theories are desirable or required in order to meet the overall objectives of AGI realisation. We will examine a range of AGI theories including those based upon or related to the following:

(1) symbolic vs connectionist vs hybrid neurosymbolic approaches; 

(2) evolutionary algorithms; 

(3) AGI and cognitive architectures, particularly enactive, self organising, distributed models of intelligence (Hyperon and more theoretical models of the multilayer self organisation of biological systems);

(4) probabilistic and graphical models;  

(5) universal, thermodynamic and physicalist models anchored in physical dynamics (rather than abstractions at the information level only). 

The latter includes neuromorphic computing, pancomputational enactivism, UAI and AIXI, self-varying AI and recursive models of improvement and hypercomputational models of AGI (especially relevant to the quantum case below) and non-classical logical approaches to AGI, including touching upon paraconsistent and other models.

We will not engage in a deep-dive with every such theory, but we will filter them for those which we believe align primarily with the theory and practical implementation of AGI on quantum computers (demonstrable via simulations), with a focus on Hyperon. 

 

2. Quantum Information Processing Theories 

In Stage 2, we examine the canonical theory of quantum information processing (QIP) relevant to AGI. Quantum computing is fundamentally based upon the principles and axioms of quantum mechanics. This applies both to the physical computational substrates upon which any AGI systems would be instantiated (e.g. physical qubits) and the logical computational abstractions (e.g. logical qubits). Quantum computing is both epistemologically distinct from classical computing (owing to the existence of distinct ways of processing information) and ontologically distinct, where inherent stochasticity/uncertainty (via superposition and entanglement) is ontological rather than epistemic in nature. These unique features of quantum systems will affect any proposals for QAGI as they have profound consequences for the types of computation that may be undertaken, resource requirements, complexity classes of problems, communication protocols, measurement and sensing, and error correction. The quantum characteristics of computation also have consequences for how such systems are controlled, influenced, verified and distributed in ways that are distinct from classical counterparts. They also give rise to unique differences in how quantum computational systems can be said to learn and so to any generalised quantum theory of intelligence itself. This section of the project will cover: quantum foundations, theory, computation, communication, sensing, open quantum systems, measurement and quantum control, quantum error correction, quantum machine learning, theoretical constraints and physical realisation. Each of the topics is vast, so we will focus concretely on the implications of these features of QIP for key properties of Hyperon-based AGI, such as its feasibility, scalability, generalisation, controllability, reproducibility and other characteristics of merit. 

 

3. Effects of QIP upon AGI 

In Stage 3, we integrate our analyses from 1 and 2 to address (a) the algorithmic consequences for classical AGI on quantum computers and (b) new theories or forms of QAGI which are only realisable in QIP paradigms. This stage will also involve identifying issues regarding quantum (as distinct from conventional classical) AGI theories. We will also set out metrics and a framework according to which efficacy of AGI theories implemented on quantum devices may be assessed. Our groundwork laid in Stages 1 and 2 will enable us to rigorously assess the applicability of quantum computing to key AGI tasks, such as reasoning, pattern recognition, and resource allocation and how such tasks differ (if at all) in the quantum context. We will also highlight specific, realistic scenarios where the exponential computing capabilities of quantum computing and its other affordances may support AGI components with massive workloads. Thus we include specific results regarding:

(1) Consequences of classical AGI given the theory of QIP; 

(2) Discussion of prospective QAGI characteristics that are unique or different from classical AGI;  

This includes analysis of how quantum may assist in overcoming physical barriers to scaling faced by classical hardware e.g. thermodynamic limits. 

 

4. Physical implementation of QAGI

Stage 4 takes the results from 1, 2 and 3 and examines the leading candidate proposals for quantum computing architectures, for the purpose of realising QAGI. To compare each prospective hardware stack, we will develop and apply a taxonomic framework according to which architectures can be ranked and compared in terms of modularity, computational effects, error correction requirements, controllability and verifiability. The architectures we will consider include: superconducting circuits, trapped ions, photonic quantum computing, topological quantum computing, neutral atom architectures, quantum dot-based circuits, silicon-based qubits, cavity-based quantum systems, molecular-based methods and others. We will also examine the utility for AGI of hybrid approaches combining different quantum hardware systems above, including their relation to thermodynamic circuit proposals.

Quantum information processing systems affect the extent to which quantum or classical theories of AGI may be implemented, having a considerable effect on whether, how and in what form such AGI theories are realisable. Thus having analysed each form of quantum architecture above, we will examine their suitability as substrates on which AGI is to be built, identifying their relative strengths and weaknesses regarding controllability, scalability, robustness and economic feasibility.

 

5. Simulations

Building on our expertise in open quantum systems’ simulation for multiple quantum hardware devices, throughout each stage, we will also work with the Hyperon and MeTTa developers to design and implement simulations of multi-qubit and qudit quantum machine learning systems to test the hypotheses regarding QAGI implementation in an empirical way.

 

Open Source Licensing

Custom

Papers will be made available for free on preprint servers, and otherwise published through peer reviewed conferences or journals. Simulation code and documentation (should stakeholders desire them) will be available via a publicly-accessible repository (such as Github) as per our previous quantum simualation projects.

Links and references

E. Perrier, D. Tao, and C. Ferrie. Quantum Geometric Machine Learning for Quantum Circuits and Control. New Journal of Physics 2020

M. Li, E. Perrier, and C. Xu. Deep Hierarchical Graph Convolution for Election Prediction from Geospatial Census Data. AAAI 2020

E. Perrier, A. Youssry, and C. Ferrie. Qdataset. Scientific Data 2021

E. Perrier. Quantum Geometric Machine Learning. 2024

E. Perrier and C. S. Jackson. Solving the KP Problem with the Global Cartan Decomposition. 2024

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 4.7
  • Desirabilty 4.7
  • Usefulness 5.0

New reviews and ratings are disabled for Awarded Projects

Overall Community

4.7

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

Feasibility

4.7

from 3 reviews

Viability

4.7

from 3 reviews

Desirabilty

5

from 3 reviews

Usefulness

0

from 3 reviews

Sort by

3 ratings
  • Expert Review 1

    Overall

    4.0

    • Compliance with RFP requirements 4.0
    • Solution details and team expertise 4.0
    • Value for money 5.0

  • Expert Review 2

    Overall

    5.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 5.0
    • Value for money 5.0
    Outstanding and deeply thought out proposal

    This proposal stands out in terms of going all the way from AGI philosophy and math through to concrete quantum computing particulars. None of the other proposals for this RFP seems to go so far in terms of seriously considering what it means to do QC forAGI rather than just narrow AI -- while also demonstrating competence and knowledge at the many hairy technical details of quantum computing.

  • Expert Review 3

    Overall

    5.0

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

    Great breakdown of tasks, both theoretical and in terms of AGI applications. Very detailed.

  • Total Milestones

    4

  • Total Budget

    $80,000 USD

  • Last Updated

    3 Feb 2025

Milestone 1 - Quantum Paradigms and Research Plan

Status
😐 Not Started
Description

This milestone establishes the theoretical foundation for AGI implementation in QIP and informs subsequent development phases: (1) Identification and analysis of classical theories across symbolic connectionist evolutionary hybrid and cognitive architecture approaches including embodied and enactive cognition Bayesian models probabilistic reasoning and thermodynamic or physicalist frameworks. (2) Exploration of frameworks such as UAI AIXI and recursive self-improvement models alongside computational neuroscience and hypercomputational proposals. (3) Commencement of QIP fundamentals analysis highlighting natural compatibility with graphical frameworks like Hyperon’s Atomspace and MeTTa language. (4) Examination of classical-to-quantum extensions assessing AGI theories implementable on quantum systems and the consequences of such transitions including changes in formal semantics epistemic frameworks and theories of agency and environment interaction. (5) Prioritization of AGI paradigms based on their feasibility for quantum implementation focusing on theories aligning with OpenCog Hyperon models. (6) Comparative study of AGI paradigms in relation to classical machine learning statistical theories and metrics like generalization merit to evaluate their potential effectiveness in QAGI contexts. (7) Analysis of simulation design principles (code stack and doc. plan) for numerical simulations of AGI and QAGI components using multi-qubit and qudit systems.

Deliverables

Research plan: finalization of research plan with stakeholders. This will comprise a comprehensive detailed proposal for assessing QAGI paradigms and implementations identified above. This should include the identification and analyses of several options for realization of quantum computing including photonic superconducting topological entropy-based proposals trapped ion silicon and others and their possible application to AGI systems. We will also include option for running simulations of low-dimensional multi-qubit systems as a means of studying component-wise AGI-related subroutines (and learning protocols) in a quantum context. Literature review: commencement of an in-depth review of existing research on AGI. This will cover each Stage identified above: Stage 1 – Review of theories of AGI; Stage 2 – Review of theories of quantum information processing; Stage 3 – Theoretical analyses of impact of QIP on AGI (classical and quantum); Stage 4 – Analyses of quantum hardware proposals for realisation of classical/quantum AGI The first milestone will focus on commencing Stages 1 and 2 of the review. Simulations: We will also initiate a workstream for co-development with the Hyperon and MeTTa development team of simulation design principles for testing core components of such AGI theories using multi-qubit and qudit systems including with technical laboratories and via online cloud quantum platforms.

Budget

$20,000 USD

Success Criterion

Finalize research plan meeting the above specification, approved by all stakeholders. Commence the in-depth review.

Link URL

Milestone 2 - Demonstration of Progress

Status
😐 Not Started
Description

This milestone establishes a framework for evaluating the feasibility scalability and theoretical grounding of AGI in quantum environments. This includes analysis of: (1) Core quantum mechanics and QIP theories incl. axiomatic treatments of Hilbert spaces quantum evolution measurement superposition and entanglement; (2) Quantum computational principles covering quantum algorithms (e.g. quantum Fourier transform simulation) data encoding techniques and complexity class relationships; (3) Quantum communication protocols including entanglement cryptographic methods (e.g. QKD) and distributed quantum system architectures; (4) Quantum sensing techniques including quantum instruments continuous measurement protocols and their relevance to AGI interaction with environments; (5) Quantum error correction methods (e.g. surface codes stabilizer codes) and their impact on reliable computation; (6) Quantum machine learning (QML) models including hybrid systems (variational circuits quantum neural networks) and constraints like barren plateaus and entanglement effects; (7) Theoretical constraints (e.g. no-cloning thermodynamic limits) and physical QIP realizations including photonic superconducting and trapped ion substrates; (8) Integration of quantum foundational theories (e.g. Bell’s theorems many-worlds hypotheses) and metaphysical concepts (e.g. identity individuation) to frame AGI’s ontological and epistemological basis.

Deliverables

Deliverables for this milestone to include: Literature review: completion of Stages 1 and 2 of the literature review representing a continuation of our in-depth review of existing research on quantum computing technologies. We will also have identified the structural pathway for our Stage 3 analysis including specific theorems we believe are of benefit to prove and candidate hypotheses regarding AGI and quantum integration for testing via numerical simulation. These will explore the implications of the above for AGI- relevant tasks including reasoning pattern recognition and resource allocation. Simulations: We will also look to begin implementation in concern technical simulation of certain AGI-related subroutines on classical simulations of one or two-qubit systems. Classical simulation of open quantum systems rapidly becomes infeasible however it is common in QIP research to simulate simple systems and then theoretically consider how they scale.

Budget

$20,000 USD

Success Criterion

Completion of Stages 1 and 2 as above, which should be sufficient to begin writing papers, to a standard meeting the approval of stakeholders.

Link URL

Milestone 3 - Continued Demonstration of Progress

Status
😐 Not Started
Description

Completion of Stage 3 and commencement of Stage 4 integrating the results of Stages 1 and 2 to evaluate the impact of quantum information processing (QIP) on AGI. This milestone addresses two key issues: (a) the algorithmic consequences for implementing classical AGI on quantum computers and (b) the emergence of new theories or characteristics unique to QAGI that are fundamentally distinct from classical approaches. The analysis includes: (1) A systematic exploration of how QIP principles (e.g. superposition entanglement) influence classical AGI tasks such as reasoning pattern recognition and resource allocation identifying areas where quantum computing provides novel advantages. (2) Development of metrics and a framework to assess the efficacy of AGI theories (focusing on Hyperon) implemented on quantum devices including scalability generalization and task-specific performance. (3) Examination of prospective QAGI characteristics such as unique learning paradigms computational resource distribution and quantum-enhanced decision-making highlighting differences from classical AGI. (4) Identification and simulation of realistic scenarios where quantum computing’s exponential capabilities and its ability to overcome classical scaling barriers such as thermodynamic limits enable AGI components to handle massive workloads. This milestone sets the theoretical groundwork for evaluating the implementation of AGI theories on quantum hardware/implementation.

Deliverables

Our deliverables will include: (1) Literature review: completion of Stage 3 and development of analytical framework for Stage 4 (quantum hardware). This will identify use cases: Highlight specific realistic scenarios where quantum computing may support Hyperon AGI components with massive workloads. (2) Research report and papers: First drafts of research report and drafts of academic papers drawn from our analyses; (3) Simulations: first working simulations testing hypotheses generated via our literature review and other research. At this stage we intend to bed-down our simulations of the theoretical integration of Hyperon AGI and QIP. With this done we then intend to design modifications and environments simulating quantum hardware systems in order to assess how those core components or subcomponents of interest to Hyperon AGI/QAGI fare on different quantum architectures.

Budget

$20,000 USD

Success Criterion

A useful analytical framework for evaluating quantum hardware and first drafts of papers should be completed, and of an appropriate standard to meet the approval of stakeholders.

Link URL

Milestone 4 - Prospective Implementation Pathways for QAGI

Status
😐 Not Started
Description

The final milestone involves completion of Stage 4 (and thus the main imperative of the project) which evaluates practical hardware and systems for QAGI integrating results from Stages 1–3 to assess how quantum computing architectures influence the feasibility and realization of AGI theories with a specific focus on Hyperon and MeTTa protocols. A taxonomic framework will be developed to compare leading quantum architectures including superconducting circuits trapped ions photonic systems topological qubits neutral atoms quantum dots silicon-based qubits diamond NV centers molecular quantum computing and hybrid approaches. The framework will evaluate these architectures on modularity computational effects error correction scalability robustness controllability and economic feasibility. Analysis will address how physical substrates impact AGI implementation exploring their compatibility with classical and QAGI theories. Differences in hardware capabilities—such as scaling beyond thermodynamic limits resistance to noise or suitability for specific AGI tasks like reasoning and pattern recognition—will be highlighted. This milestone provides a comparative assessment of architectures identifying their strengths and weaknesses for AGI and establishing a foundation for realizing QAGI.

Deliverables

(1) Finalisation of research report into QAGI and the implementation of Hyperon-based AGI on quantum hardware platforms; (2) Finalisation of preprint versions of journal articles with submission proposals to relevant journals. We will be aiming to produce academic publications based upon research from the main QAGI report; (3) Simulation development extension in order to simulate (to the extent possible) how classical and quantum AGI protocols vary when implemented on different quantum hardware systems. This involves implementing a simulation of those physical systems and then considering how AGI and QAGI may be implemented via adaptations in order to identify strengths and weaknesses. We also expect to have accompanying code documentation for these simulations available for other researchers (similar to other projects we have run with code bases and repositories). (4) Research recommendations: here we will provide clear recommendations on areas for further study or potential experimentation with quantum computing in AGI. We will focus on practical commercialisable pathways toward QAGI. Our report will contain limited analysis of the economic feasibility and other issues arising that we believe are important when strategically identifying pathways to QAGI.

Budget

$20,000 USD

Success Criterion

Stakeholder approval. Recommendations should be concrete and actionable. Preprints should be of a high standard.

Link URL

Join the Discussion (0)

Expert Ratings

Reviews & Ratings

Group Expert Rating (Final)

Overall

5.0

  • Feasibility 4.7
  • Desirabilty 4.7
  • Usefulness 5.0

New reviews and ratings are disabled for Awarded Projects

  • Expert Review 1

    Overall

    4.0

    • Compliance with RFP requirements 4.0
    • Solution details and team expertise 4.0
    • Value for money 5.0

  • Expert Review 2

    Overall

    5.0

    • Compliance with RFP requirements 5.0
    • Solution details and team expertise 5.0
    • Value for money 5.0
    Outstanding and deeply thought out proposal

    This proposal stands out in terms of going all the way from AGI philosophy and math through to concrete quantum computing particulars. None of the other proposals for this RFP seems to go so far in terms of seriously considering what it means to do QC forAGI rather than just narrow AI -- while also demonstrating competence and knowledge at the many hairy technical details of quantum computing.

  • Expert Review 3

    Overall

    5.0

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

    Great breakdown of tasks, both theoretical and in terms of AGI applications. Very detailed.

feedback_icon