Compute access for public research
Available compute for universities and public institutions is mapped against the needs of training and evaluating models, showing where access limits research.
GCRI · Artificial Intelligence and Autonomous Systems
Computing for AI
High-performance and cloud computing supply the capacity AI needs. In this field computing is studied for capacity, concentration, energy and access: who can train and evaluate models, at what cost, and with what environmental footprint.
01Contents
Computing for AI in five parts.
02Application
What high-performance, cloud and edge computing does here.
High-performance, cloud and edge computing provide the processing and storage capacity for simulation, analytics and digital services. They are both an enabler of risk analysis and a critical dependency for society. In artificial intelligence and data and compute, computing capacity determines how fast and at what resolution risks can be modelled. Its concentration is itself a systemic dependency for finance, health and government services. The canonical description is on the STEM Commons.
Energy measurement of workloads
Energy use of training and inference jobs is measured with hardware counters and facility meters, and reported with the hardware and settings used.
Cloud concentration analysis
Dependence of AI services on a few cloud providers and regions is mapped to show shared points of failure.
03Monitor
Compute workload.
The screen a team would use to follow this method: its measures, states and decisions. Names are examples and values are illustrative.
Compute workload
AI and autonomous systems · Computing for AI · Run R-9546
- Channels reporting95of 100Normal
- Median latency124ms
- Records failing checks2.7per 1,000Watch
- Backlog34min
Gaps on 4 channelsGCRI · Artificial Intelligence and Autonomous Systems
Latency, p95
Events
- 06:13WarnChannel C-75 stale
- 07:51InfoSchema v5 accepted
- 08:36ErrorChecksum mismatch, batch quarantined
- 08:44DoneReplay complete
04Data
What the method consumes.
Datasets and records used by high-performance, cloud and edge computing on this platform, with the role each plays and the label its values carry. Sources are listed in the source register.
| Data | Role | Label |
|---|---|---|
| Facility and job energy logs | Energy accounting | OBSERVED |
| Public compute allocation records | Access | Published statistics |
| Cloud region and provider data | Concentration | DECLARED |
05Methods and evidence
How Computing for AI is modelled, measured and tested.
The model, its assumptions, its data and its uncertainty, in the form a scientist or engineer would check. Structures follow published methods. The values drawn are illustrative, not findings.
Model, equations and assumptions
The model structure used for Computing for AI, the assumptions it makes and how each is tested.
Risk of element i
Ri = f(Hi, Ei, Vi, Ci)hazard, exposure, vulnerability and coping capacityExpected annual loss
EAL = ∑k pk · Lkover events k with annual probability pkCoupled systems
xt+1 = A xt + B ut + εtA carries the couplings, u the shock, ε the error| Assumption | Why it is made | How it is tested |
|---|---|---|
| Hazard and exposure are independent within an event | Keeps the loss integral tractable | Compare with a copula fit on joint records |
| Couplings are linear near the operating point | Lets shocks propagate through A | Perturb beyond the linear range and compare |
| Vulnerability curves transfer between places | Local curves are rarely available | Validate against observed losses where they exist |
- Inputs
- versioned datasets with provenance
- Outputs
- distributions, never single numbers
- Code
- released with the model card and tests
- Reproduction
- same data, code, environment and seed
FAIR guiding principles · GO FAIR and FORCE11W3C PROV · World Wide Web ConsortiumGuidance on consistent treatment of uncertainties · IPCCCEOS Analysis Ready Data · Committee on Earth Observation Satellites
06Validation
How results are checked.
Systems are tested through benchmarks, load testing and failover exercises. Numerical reproducibility is checked across hardware and software configurations.
| Check | How it is done |
|---|---|
| Measurement boundaries | Energy figures state what is included, such as cooling and idle time |
| Reproducible configuration | Hardware and software settings are recorded |
| Provider data verification | Cloud dependency claims are checked against service documentation |
Concentration in few providers or regions creates shared points of failure. Energy, water and supply chain constraints limit capacity growth.
07Process
How the process runs.
Process
Computing for AI: Build, sign and accept
An admitted mission becomes a cycle project; the build is pinned, verified and signed, and the Global Risks Forum (GRF) records acceptance for exercise, not permission to deploy.
01 Request and sources
- 01Admitted mission becomes a cycle projectRecord · Universe to FoundryAn admitted mission becomes a cycle project with a decision case.
- 02Qualify data sources and resourcesStep · the Global Centre for Risk and Innovation (GCRI)GCRI qualifies data sources and resource offers and builds the scenario models. An offer is not an allocation.
02 Build and verify
- 03Source snapshotRecord · NXS-FAC (blueprints, builds and signing)The snapshot pins the commit, dependencies and generator versions.
- 04Build contractRecord · NXS-FAC (blueprints, builds and signing)The build contract declares the artefact kind, environment and reproducibility envelope.
- 05Integrate and buildModel · GCRIGCRI integrates software and data, signs builds, and tests security and interoperability.
- 06Verification suiteGate · NXS-FAC (blueprints, builds and signing)Required and advisory checks stay separate. Reproducibility needs pinned inputs, declared nondeterminism and independent rebuilds.
03 Sign and admit
- 07Sign the exact artefactStep · NXS-FAC (blueprints, builds and signing)NXS-FAC builds, attests and requests a signature via a scoped signer identity. It holds no key material.
- 08Package admissionStep · NXS-NXQ (software, models and capability lifecycle)NXS-NXQ receives the signed package and keeps admission evidence that binds the task, data class, evaluation suite and expiration.
04 Accept
- 09Core Build ReportRecord · GCRIThe Core Build Report is the go or no-go record. Universe assembles the Core and holds the decision for the exercise.
- 10G3 · Build accepted for exerciseGate · The Global Risks ForumGRF records whether the build is accepted for exercise. A signed build or release candidate is not permission to deploy to production.
The programme year · NXS-FAC · UniverseDesigned programme process, not a record of operation. No cycle has run and every programme is at service level R0. A gate is a programme decision, never a public-authority approval, certification or deployment authorisation. The record order follows the Builds and Releases room lists of NXS-FAC.
08Connections
Topics that rely on this method.
09Stewardship
Who maintains the method.
Data and Compute stewards high-performance, cloud and edge computing for the Consortium and reviews how it is applied across platforms.
10Connected
Where this page leads across the Consortium.
11Questions
Questions and answers.
What are the limits of high-performance, cloud and edge computing in this field?
Concentration in few providers or regions creates shared points of failure. Energy, water and supply chain constraints limit capacity growth.
How are results checked?
Systems are tested through benchmarks, load testing and failover exercises. Numerical reproducibility is checked across hardware and software configurations.
Are the values on this page real?
The data names, roles and checks are real descriptions of practice. The canvas shows a design of the method’s use, not a run; no result is reported here.
Where else is this method applied?
The canonical method page lists every platform that applies it.
ReportsContribute
Contribute evidence from this page.
Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
GCRIArtificial Intelligence and Autonomous SystemsExponential TechnologyLabs Evidence
- EmailGateway GW-GCRISend it from this pageEvidence, a correction, a case, a method, a review, a translation or profile input. It arrives tagged with this page, so the editor who owns it receives it.
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- ReportsGCRISee every way to contributeThe contribution types, how each is reviewed, what credit you receive, the licence and what is never accepted.
Never send personal data, sensitive infrastructure detail or anything you have no right to share. A contribution is reviewed before use; sending it confers no authorship or editorial rights, and sponsorship never buys a finding.
Artificial Intelligence and Autonomous Systems
Bring a method, a dataset or a question.
Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.
Legal
Legal documents.
Maps, data and figures on this site are shown for context and general information. They are not advice, forecasts or warnings; the limits of the information are set out in the Disclaimers and Data Notices.