Skip to content

GCRI · Artificial Intelligence and Autonomous Systems

Uphando nobungqina

The research behind Artificial Intelligence and Autonomous Systems: the questions it asks, the models and measurements it uses, how uncertainty is stated and how a result is reviewed and reproduced before it is published.

01Indawo yokusebenzela

The workspace for research and evidence.

Evidence register

AI and autonomous systems · Research and evidence · Example interface, not a record

Register currentGCRI · Artificial Intelligence and Autonomous Systems

Izalathisi

unit · cadence · label

Models in registrymodelsweeklyPublished statisticsSP-9867
Evaluations passed at release of suitesper releaseASSESSEDSP-0910
Open AI incidentsincidentsdailyKUXELIWESP-2588
Agent actions needing approvalactions / daydailyKUQWALASELWESP-8895
Compute per training runGPU-hoursper runKUXELIWESP-6682

Source passport

Umthombo
Model registry
Ilayisensi
open · attribution
Umgcini
Incident responder
Ubungakanani bokugubungela
83 pct of places
Known limits
sparse in uplands

Coverage by place

Evidence register.Illustrative interface; not live data.

Artificial Intelligence and Autonomous Systems · Evidence register. Evidence register: example names and illustrative values, not a record.

02Imibuzo

What the research asks.

  1. 01

    Model evaluation

    Validity, calibration and generalisation.

  2. 02

    Agentic systems

    Tool access, delegated authority and safe execution.

  3. 03

    AI cyber security

    Adversarial methods and model exposure.

  4. 04

    AI in science

    Discovery and reproducibility.

  5. 05

    AI and public services

    Public accountability, access and rights.

  6. 06

    Autonomous physical systems

    Robots, sensors and controls.

03Open questions

What AI research cannot yet answer well enough for decisions.

AI risk research has to keep pace with capability. The agenda concentrates on evaluation, agents, uses in public services and science, and the system-level effects that single-model testing misses.

  • Which evaluations predict real-world failure, rather than benchmark performance? See model evaluation.
  • How should agents that act through tools be scoped, monitored and stopped? See agentic systems.
  • What evidence should a public service require before using AI in decisions about people? See AI and public services.
  • How do autonomous physical systems fail in mixed human-machine environments? See autonomous physical systems.
  • What is the energy and water footprint of AI compute, and where does it concentrate? See compute and energy.
04Ubungqina

Every result keeps its label.

Results are labelled as observed, published statistics, estimated, modelled, forecast, scenario or expert judgement. Missing data is shown as missing, never filled silently.

Iindlela zokusebenza

Methods and uncertainty

How results are produced and where they stop

Imithombo

Source and data register

Where data comes from and on what terms

Iimpapasho

Reviewed outputs

Citable work with its editorial status

05Imithombo

Where this page’s references lead.

Amaqumrhu namaxwebhu eli phepha elifundwa ngokuthelekiswa nawo. Imeko neenguqulelo ziyatshintsha; ikhonkco ngalinye liya kwiqumrhu elikhuphayo.

06Amaphepha

Pages in Research and evidence.

07Imibuzo neempendulo

Imibuzo neempendulo.

Can I propose a research question?
Yes. Write to the contact form with the question, the context and who would use the answer. A proposal receives a reference; it does not commit the platform to the work.
IingxeloNikela

Faka ubungqina obuvela kweli phepha.

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
Ubukrelekrele Bokwenziwa Neenkqubo Ezizimeleyo

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.