If you expect a single runaway agent that outplans humanity, the answer is no. That is a technologically illiterate perspective.
In Bostrom's collective sense (Bostrom, 2014, ch. 2), networks of humans and machines that together outperform any isolated human mind are already an established condition. Those networks already run hiring, medicine, finance, education, and policy. In this framework, that operational structure is defined in our work as Artificial General Intelligence (AGI): the pervasive human–AI cooperation across economy, employment, education, and ecology.
AGI is not a future capability threshold. It is a highly unstable present state, where risks from Transformative AI are already escalating worldwide, with implications from unemployment to warfare.
In this work (Korompilias, 2025, Gyroscopic Global Governance), Artificial Superintelligence (ASI) is defined as the aligned and coordinated state in which those same human–AI systems operate near a specific balance between global coherence and local differentiation (aperture ≈ 2.07%) while maintaining four principles: Traceability, Variety, Accountability, and Integrity.
The central risk is therefore not an autonomous takeover. It is Gradual Disempowerment through Power Concentration displacement: treating AI outputs as if they were human authority, and losing the measurement of ancestry between Direct and Indirect Authority and Agency.
The Gyroscopic Global Governance simulator asks how economy, employment, education, and ecology move toward that high-alignment regime.
The result is direct. Under several coordination strategies, and across 1000 random starting points, the domains converge to that balance and stay there. A separate check with Google Research's TimesFM ranks which of those paths look preferable next, and that ranking agrees closely with the simulator's own outcomes.
In plain terms: aligned Collective Superintelligence is a reachable governance regime, not a fantasy threshold. Reaching it dismantles the shared structure that keeps reproducing poverty, jobless displacement, misinformation, and ecological harm. It makes Unconditional High Income, work as alignment maintenance, and ecological regeneration structurally viable in our already Post-AGI world.
Where We Stand
What the gauge shows
- SI is a 0–100 alignment index for each domain
- SI ≥ 90 is the simulator's high-alignment region (the collective-superintelligence operating band in this model)
- The gap below 90 marks coordination friction: gains lost to misinformation, accountability gaps, and ecological externalisation instead of distributable surplus
How to read the milestones
They are heuristic alignment levels at selected moments: early AI research, AlphaGo, large language model adoption, and the 2025 calibration endpoint.
Convergence Across Coordination Strategies
The chart follows one number over time: aperture A.
What aperture is
Aperture measures a specific balance between global coherence and local differentiation. It is the share of variation that circulates in local cycles rather than fitting one coherent configuration of the four principles.
The target
- A* ≈ 0.0207
- About 97.93% gradient (global coherence)
- About 2.07% cycle (local differentiation)
That number is not a fitted policy knob. It comes from our mathematical physics research on recursive measurement: how information, inference, and intelligence stay coherent when governance closes a loop. The same structure that yields ordinary space and motion also fixes how much openness must remain for distinction. In plain terms, it bridges information theory with AI governance, and the simulator uses that derived balance as its operating target.
What κ means
κ (kappa) is coordination intensity. Lower κ means looser coupling between domains. Higher κ means tighter coupling.
What the runs show
- Seven scenarios plus 1000 random initial aperture triples (other parameters held canonical)
- Well-coupled regimes reach high alignment
- Weak coupling (κ = 0.5) is slower: Economy ends at 91.37 by step 100
- Strong coupling (κ = 2.0) is faster, but ends with the highest Governance Traceability Displacement
- Low-aperture and at-target starts leave Employment below 90 at step 100: aperture near the target is not enough if the stage profile stays imbalanced
Post-AGI cooperation is already here. The simulator shows that the ASI-like equilibrium of that cooperation is dynamically reachable when coordination holds.
Explore the Seven Scenarios
The sandbox lets you compare seven coordination strategies on the path from operational AGI toward collective superintelligence at the target balance.
Ecology note
Ecology SI stays near 100 in every run because the model weights canonical memory at 97.93%. The scenario-sensitive ecological story sits in the displacement measures (GTD, IVD, IAD, IID), not in that SI score.
Forecasting the Same Trajectories
We also asked a separate question: if you only see the early stretch of a trajectory, how well can a forecasting model continue it, and which scenario does that continuation favour?
The model used here is TimesFM 3, a pretrained time-series foundation model from Google Research, run with frozen weights (zero-shot, not retrained on our simulator). It receives the first stretch of each trajectory and predicts the steps that follow. Simple baselines (last-value persistence and drift) provide a transparent check.
Forecast accuracy
Across 105 scenario-channel combinations:
- TimesFM best in 80
- Drift best in 23
- Persistence best in 2
- 10%–90% forecast bands covered 90.5% of held-out values overall
Scenario preference
Each predicted continuation is scored against the framework's alignment objectives, then the seven scenarios are ranked. That ranking matches the ranking from the simulator's own continuations with a mean Spearman correlation of 0.74.
- Leading on average: uniform stage weights and asymmetric initialisation
- Last on average: weak coupling
A path can be easy to forecast and still be a weak governance regime. Weak coupling shows that split clearly.
What Alignment Enables
SI ≥ 90 is the high-alignment region. In this framework, that region is the operational definition of the governance configuration associated with four goals. Separate empirical series are still needed to measure poverty, job quality, literacy, and ecosystems in the field; the simulator itself has no production or fiscal sector.
When that configuration holds:
Poverty resolution. Surplus becomes distributable, and an Unconditional High Income becomes structurally viable, because less is spent on the tax of misalignment.
Meaningful employment. Work is alignment contribution across Governance Management, Information Curation, Inference Interaction, and Intelligence Cooperation.
Epistemic literacy. Education aims at tracing sources, keeping multiple Authority types in view, owning conclusions, and keeping coherence over time.
Ecological regeneration. Ecological strain is the downstream record of upstream governance displacement, so less harm is pushed outward as an external cost.
That is what collective superintelligence looks like when it is governed: not a smarter overlord, but a human–AI arrangement that can actually distribute surplus, organise work, keep knowledge accountable, and stop exporting harm.
Long-Horizon Stability
A 1,000-step run at canonical coordination shows the high-alignment configuration is not a short transient:
- After step 200, SI stays above 98.83 for the remaining 800 steps
- Apertures stay within about 0.00025 of the target
- No degradation appears over that horizon once domains are in the high-alignment region
Evidence from Existing Programmes
Programme-level evidence shows the same pattern at local scale: when allocation chains become shorter and more traceable, health, education, and fiscal outcomes improve.
Unconditional income support
- Mincome (Canada): roughly 8.5% fewer hospitalisations; better educational attainment
- Negative income tax trials (US): limited work reduction; gains in high school completion
- GiveDirectly (East Africa): investment in assets, housing, and enterprises; no rise in alcohol or tobacco spending
Housing First
- Utah: about 74% less chronic homelessness; estimated annual public cost per person from about $16,670 to $11,000
- Dutch city programmes: reported benefit–cost ratios of 2:1 to 3:1 when justice, emergency, and health costs are counted
Direct health interventions
- Deworming (Kenya/India): higher school attendance; lower mortality
- Free insecticide-treated nets: high uptake; even small fees cut uptake sharply
Less is lost to emergencies and administrative friction. That is the local form of what system-wide alignment makes structurally coherent.
What You Can Do
- AI safety and governance: use the four principles (traceability, variety, accountability, integrity) as a checklist for human–AI workflows and institutions
- Policy and social programmes: look for displacement (responsibility and information passed along without clear human ownership), then shorten those chains
- Research: test whether local programmes that reduce displacement show the same structural patterns as high-SI regimes in the simulator
Paper, report, simulator, and forecasting materials are linked below. The aperture target itself is derived in our mathematical physics research on recursive measurement and governance structure.
