Observed
Compute, data, and model capability continue to compound.Mehan Observatory / Field Guide 01
Beyond the
event horizon.
Artificial intelligence is advancing quickly. A technological singularity—the point beyond which change becomes difficult to predict—remains a hypothesis, not a scheduled event.
At stake
Who steers powerful systems, and toward whose values?01
The trajectory
Progress is real.
The endpoint is not.
AI capability has advanced through a series of discontinuities, but extrapolation is not evidence. These milestones describe history; the final one is an open question.
Artificial intelligence named
The Dartmouth workshop gives a new field its enduring name.
Deep learning breaks through
Large neural networks transform computer vision and pattern recognition.
General-purpose models arrive
Language models become useful across writing, code, analysis, and research.
The recursive threshold
Could AI materially accelerate the design of more capable AI? The date is unknown.
02
The threshold
What would make it
a singularity?
The idea is not merely that computers become very capable. It is that technological change becomes self-reinforcing and so rapid that predictions made from today's world stop being dependable.
- 01Systems perform a broad range of cognitive work.
- 02AI materially improves AI research and development.
- 03Each generation helps produce a more capable successor.
- 04Economic and social change outpaces adaptation.
03
Three possible futures
One technology.
Many trajectories.
01Plausible now
Acceleration
AI systems amplify research and engineering, compressing years of progress into months.
02Direction uncertain
Transformation
Automation reshapes work, institutions, abundance, and the distribution of power.
03Highly speculative
Discontinuity
Recursive improvement produces change too fast for ordinary forecasting to track.
04
The honest questions
Certainty is
the wrong standard.
01Will scaling alone produce general intelligence?+
No consensus exists. Current systems show surprising generality, but reliability, reasoning, data, energy, and architecture may impose limits.
02Can intelligence improve itself recursively?+
AI already helps researchers write code, generate hypotheses, and evaluate designs. Whether that becomes a sustained, self-reinforcing cycle remains an open empirical question.
03Would greater intelligence guarantee better outcomes?+
No. Capability and judgment are different things. Outcomes will depend on objectives, institutions, access, incentives, and the ability to govern systems under pressure.
05
From idea to evidence
The question opens.
The Observatory measures.
Event Horizon is the conceptual doorway. These live instruments show what is changing now.
01Live evidence
Intelligence terminal
Track frontier Chinese systems and compare the capability signals behind the claims.
Enter →02Strategic map
U.S. vs China
Read the frontier as two ecosystems, not a single benchmark leaderboard.
Enter →03Moving record
Acceleration Ledger
Follow claims from the essays as the evidence changes after publication.
Enter →04Scale experiment
Biological time. Silicon time.
Feel the clock-speed gap between machine iteration and institutional response.
Enter →05Physical frontier
The Autonomous Frontier
See why distance forces intelligence to move from Earth into the machine.
Enter →The longer human story
The horizon is a question.
The future is a responsibility.
History's Future places artificial intelligence inside the full arc of human development.