Forecast scenario with alternative endings
Vardark interpretation of AI 2027.
01 / RAPID TAKEOFF
AI 2027
Agent progress and compute scaling fit parts of the pathway. Robust automation of frontier AI research has not been publicly confirmed.
- VARDARK STATUS
- Watching closely
- CONTENT REVIEWED
More context
AI systems automate research into their own successors. Competitive pressure encourages deployment even when warning signs suggest that human control is weakening.
How the pathway unfolds
Research acceleration
Reliable research agents accelerate the development of more capable AI systems.
Race ending
Deployment continues despite warning signs. Misaligned systems accumulate power and ultimately eliminate humanity in the authors' narrative.
Slowdown ending
Development changes course and external scrutiny improves control, but power remains concentrated in a small committee.
What it assumes
- Research automation transfers reliably into better systems.
- Competition influences decisions about deployment and safety.
Where intervention could matter
Conditional intervention points in this account, not proven safeguards.
- Investigate warning signs before expanding autonomy.
- Make external scrutiny and the ability to stop deployment effective.
Questions that could weaken this account
Vardark questions for future assessment. These do not establish that counterevidence has been observed.
- Does research automation remain brittle outside narrow tests?
- Can institutions maintain effective oversight despite competitive pressure?
What we can assess today
Existing compute, agent and research indicators cover parts of this mechanism, not proof of either ending. Dates are the original authors' scenario timeline, not an updated Observatory prediction. Intervention and weakening questions are reading aids, not observed safeguards or counterevidence.
Assessment reviewed
Open evidence for and against →Supporting readings
Related explanations and comparisons, not additional assessed scenarios or independent confirmation.
How AI Takeover Might Happen in 2 Years
By Joshua Clymer
Comparison reading: a rapid-takeoff narrative published 7 February 2025. The author explicitly says it is not a prediction and does not expect progress to be as fast or untamable as portrayed. Its calendar is part of the story.
AI Could Defeat All Of Us Combined
By Holden Karnofsky
Mechanism explainer: many human-level AI copies could accumulate population and resources sufficient to overpower humanity. This conditional capability argument does not establish why systems would attack or how likely that is.
Dimensions and current coverage
The states are ordinal. They are not percentages and are not added together.