The 'AI 2027' Forecast and Daniel Kokotajlo's Accelerating AGI Timeline
Is this a scandal?
No longer — the story has resolved. Noise 1/100, cooling down, across 1 source.
The debate is likely to intensify as 2025 approaches and the gap between current capabilities and the 2027 forecast narrows. Expect more former researchers to break NDAs to comment on alignment safety as internal benchmarks potentially validate these accelerated timelines.
Noise 1/100 — louder than 87% of tracked AI controversies.
Why it matters
Aggressive timelines from credible insiders pressure labs to prioritize safety over speed before recursive self-improvement potentially renders human oversight obsolete.
Key points
- Daniel Kokotajlo predicts AI will surpass the smartest humans by early 2027.
- The forecast alleges recursive self-improvement at superhuman speed could begin in late 2027.
- Kokotajlo assigns a 30% probability to this specific disruptive timeline occurring.
- The researcher previously worked at OpenAI and accurately forecasted LLM capabilities in 2021.
- The 'AI 2040: Plan A' report compares potential AI disruption levels to the COVID-19 pandemic.
- Community discussions highlight a divergence between aggressive insider forecasts and mainstream AGI estimates.
The story
Former OpenAI researcher Daniel Kokotajlo has published a forecast predicting artificial intelligence systems will surpass the smartest humans by early 2027. The report, titled 'AI 2040: Plan A,' alleges these systems could achieve recursive self-improvement at superhuman speeds by late 2027. Kokotajlo, who previously authored accurate large language model forecasts in 2021, assigns a 30% probability to this disruptive timeline occurring. This projection contrasts sharply with industry consensus estimates that typically place artificial general intelligence decades away. The forecast argues current development trajectories mirror pandemic-level societal disruption risks rather than gradual technological adoption. Safety advocates cite this assessment as evidence that alignment research is lagging behind capability advancement. Industry observers note Kokotajlo’s departure from OpenAI lends weight to his warnings about insufficient safety preparations for near-term superintelligence scenarios.
Who's involved
Argues that AGI is arriving sooner than previously expected and that safety preparations are currently inadequate.
Continuing to develop reasoning-capable models while maintaining that AGI can be managed through iterative deployment and safety guardrails.
Maintain that model development is following a controlled, safety-conscious path toward beneficial AGI.
Engaged in mapping current AI releases like 'o1' to the 'AI 2027' prophetic stages to determine our proximity to AGI.
Analyzes real-world AI developments through the lens of a specific, high-impact predictive scenario.
How the conversation shifted
Polarity (0–100) from the noise pipeline, sampled over time.
Noise Level
The timeline
Community Mapping of 'Agent' Levels
Public discourse intensifies regarding the mapping of 'o1' and other models to the 'AI 2027' hierarchy.
Discourse Peak on AI 2027
Online communities debate '4D chess' tokenomics and the role of 'OpenClaw' in the final approach to AGI.
- 2024-Q4
Timeline Acceleration
Kokotajlo publicly shifts his AGI timeline one year earlier as technical progress outpaces expectations.
OpenAI o1 Released
The release of reasoning-heavy models validates earlier 'Agent Zero' predictions for many observers.
Kokotajlo Leaves OpenAI
Resigns from OpenAI citing concerns about the company's commitment to AGI safety and alignment.
Kokotajlo Initial Predictions
Daniel Kokotajlo predicts the rise of multi-step reasoning models years before their mainstream emergence.
Early Predictions Formulated
Daniel Kokotajlo predicts the rise of multi-step reasoning models and sets early AGI timelines.
The full record
Sources & methodology
- AI researcher warns of AI surpassing humans by 2027 ... — linkedin.com · located later (2026-07-30)
- AI 2040: Plan A — news.ycombinator.com · located later (2026-07-30)
The records from this story's original coverage were pruned, so items marked located later were found by searching for it afterwards. The summary above has since been rewritten to take them into account — it is not the text first published. How we score →
The forecast
The debate is likely to intensify as 2025 approaches and the gap between current capabilities and the 2027 forecast narrows. Expect more former researchers to break NDAs to comment on alignment safety as internal benchmarks potentially validate these accelerated timelines.
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.