c. 2012–present
Artificial Intelligence
A current transition: machines that learn patterns and perform cognitive work—capabilities rising faster than institutions can calmly absorb.
Previous state
Software mostly executes hand-authored rules; cognitive scarcity is a binding constraint on many knowledge tasks.
New state
Learning systems perform broad classes of perception, generation, and decision-support tasks with rapidly improving competence.
New capability
Cheap, scalable pattern-based cognitive labor (uneven, error-prone, but expanding)
Enabling conditions
- Large datasets and scalable compute
- Neural network methods that generalize surprisingly well
- Internet-scale digital exhaust as training fuel
- Intense capital competition among labs and states
Technologies
- Artificial Intelligence
- Digital Minds
- Human-Machine Systems
- Brain-Computer Interfaces
- Robotics
Social changes
- White-collar workflows rewriting in real time
- New arms races in persuasion, cyber, and surveillance
- Public uncertainty about work, authorship, and truth
Major consequences
- Potential acceleration of science and software
- Concentrated geopolitical and corporate leverage
- Novel catastrophic and chronic risk pathways
New bottlenecks
- Evaluation, reliability, and alignment
- Energy and chip supply
- Legitimacy of automated decisions
- Human meaning and responsibility under automation
Competing explanations
- Continuous tool improvement, not a civilizational transition
- True transition comparable to industry/electricity in generality
- Hype cycle masking brittle narrow gains
Notes
Status: current, unfinished
Unlike agriculture or print, this transition is happening around us. That means higher uncertainty and higher rhetoric. Figuora marks claims carefully.
Established: rapid capability gains on many benchmarks and deployments since deep learning scaled. Interpretation: whether this becomes a general civilizational transition. Hypothesis/speculation: timelines to transformative or superhuman systems.
Capability claim (working)
The candidate unlock is cognitive labor becoming more abundant and less tied to individual human training time. Whether that amplifies human capability or displaces it is an open linked question.
What to watch
Bottlenecks migrating to verification, energy, embodiment, and governance would match prior transition logic. If those do not bind, our model needs revision—not apologetics.
Sources
- 1. Artificial Intelligence: A Modern Approach — Stuart Russell and Peter Norvig (2020) · Technical survey; not a civilization theory.
- 2. Human Compatible — Stuart Russell (2019)
- 3. The Alignment Problem — Brian Christian (2020)