Richard Ngo Exits OpenAI to Forecast AI's Psychological Future
Is this a scandal?
No longer — the story has resolved. Noise 1/100, cooling down, across 0 sources.
More senior safety researchers are likely to transition into independent advocacy or creative fields to bypass corporate non-disclosure constraints. This trend will increase public pressure on AI labs to provide more transparent governance frameworks as internal experts express skepticism about current safety protocols.
Noise 1/100 — louder than 88% of tracked AI controversies.
Why it matters
Metadata contamination in major conference archives undermines citation integrity and complicates automated literature reviews for responsible AI researchers.
Key points
- Recovered CVPR 2023 metadata contains four distinct entries spanning computer vision, land praxis, RAI tools, and psychological well-being.
- Three of the four recovered abstracts bear 2026 timestamps despite belonging to a 2023 conference archive.
- Richard P. Wildes is correctly identified as the source of the legitimate CVPR 2023 table of contents message.
- The anomalous bundling suggests metadata corruption or indexing errors affecting academic search reliability.
- No official statement from CVPR organizers confirms or denies the provenance of the recovered non-vision abstracts.
The story
Recovered metadata from the CVPR 2023 proceedings index lists four disparate entries ranging from computer vision to ancestral land healing, suggesting potential archival data contamination. The recovered dataset includes a legitimate conference table of contents message from General Chair Richard P. Wildes alongside three unrelated abstracts on land-based praxis, industrial RAI tools, and AI chatbot well-being simulations dated between January and July 2026. Researchers attempting to verify CVPR 2023 content may encounter these misaligned records when querying academic databases or institutional repositories. The inclusion of future-dated 2026 timestamps within a 2023 conference record indicates either systematic indexing errors or unauthorized database modifications. This anomaly raises concerns about the reliability of automated metadata harvesting for AI ethics and safety literature. Conference organizers have not publicly addressed the discrepancy in the recovered files.
Who's involved
Argues that current AI labs are failing to account for extreme tail risks and the psychological impact of AGI on humanity.
Maintains a focus on safety through their Governance and Alignment teams despite high-profile departures.
Hosted the discussion exploring the divide between AI skeptics and doomers on the Manifold podcast.
Noise Level
The timeline
Manifold Podcast Episode 109 Released
Ngo discusses his departure from OpenAI and his concerns regarding 'Machine God' tail risks and lab governance.
Ngo Publishes 'The Gentle Romance'
Richard Ngo releases a collection of 22 science fiction stories exploring AI and humanity's future.
The forecast
More senior safety researchers are likely to transition into independent advocacy or creative fields to bypass corporate non-disclosure constraints. This trend will increase public pressure on AI labs to provide more transparent governance frameworks as internal experts express skepticism about current safety protocols.
Forecast, not fact — an editorial estimate we score when this resolves.
That's the complete picture as of — nothing more to know right now. We'll update this page the moment it changes.
Join the Discussion
Discuss this story
Community comments coming in a future update
Be the first to share your perspective. Subscribe to comment.