LO
CapabilityCognitive automation2023-05-18
GPT-4 gave incorrect information about Singapore's Master Plan and could not give accurate setbacks and heights in an early test, the planning authority found
Urban planneroccupation page →
Event date / reported
2023-05-18
Evidence stage
CapabilityA demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
Tasks this bears on
Preparing plans and policy
Writing local plans, design codes and policies, weighing evidence about housing, transport and environment.
Still human-led≈ Platform inference
Where this applies
Singapore. The planning authority tested GPT-4 on planning and design questions and found it answered general and conceptual planning issues well, but was unable to give accurate answers on the Master Plan's land uses and planning processes, gave incorrect information in some instances, and could not give accurate urban design parameters such as setbacks and heights. It concluded that human planners should be held responsible to fact-check. A test of one 2023 model; later models may do better.
What this means
An early general model could not be trusted with the details of a statutory plan.
What it does not yet show
A test of a 2023 model by one authority.
What you can check
Open URA's article 'Early exploration of GPT-4's abilities in the context of urban planning and design in Singapore' (18 May 2023) and find "provided incorrect information".
Does it change the assessment?
No — and this stage does not move it either. A "Capability" record is real evidence, but it does not upgrade a task judgement on its own. The 1 linked judgement above stand where they were.
Source
Urban Redevelopment Authority — Early exploration of GPT-4's abilities in the context of urban planning and design in Singapore (18 May 2023) · verified 2026-10-01 · Claude (VOLO agent) · interpreted 2026-10-01 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.