ChatGPT image 2.5 botches old RFK stadium render, seemingly fighting its training data
billyjhowell · x · 2026-09-12
A user tested ChatGPT image 2.5 with "visualize the old RFK stadium from above" and got geographically wrong output — the Capitol and Washington Monument appear across the water when they shouldn't be. More interestingly, the render looks like the model is struggling between the old and new stadium designs, suggesting it's fighting conflicting versions in its training data — a classic failure mode for image generation on landmarks with multiple historical forms.
More from Fun
- Ex-Anthropic researcher's AI self-replication warning mocked: where are the 10,000 GB300 rigs? — basedjensen · 2026-09-12
- Everyone hypes every new model, but real use is just 'make this email sound less unhinged' — tech__unicorn · 2026-09-12
- Viral meme: While Anthropic and OpenAI trade doom scenarios, law firms just learned to save a PDF — ATTlKA · 2026-09-12
- User finds AI surprisingly good at pinpointing the climax of songs — jfischoff · 2026-09-12
- Dev launches live Astra improv stream showing off GPT-6 running around the clock — craigsdennis · 2026-09-12
- Redditor shares a goofy AI-generated video — hiddenprofiledude · 2026-09-12