Scenario Planning for an AGI Future—Anton Korinek

Source details

ChatGPT was released in November 2022 and demonstrated human-quality text and code generation, language translation, creative writing, and informative question answering at a previously unseen level.
Foundation models underlying generative AI have been advancing rapidly for more than a decade.
The amount of computational resources (“compute”) used to train the most cutting-edge AI systems has doubled every six months over the past decade.
Modern AI systems can exchange information at speeds that are significantly faster than human senses and language.

Geoffrey Hinton in May 2023 conjectured that artificial general intelligence (AGI) may be realized within a span of 5 to 20 years.
Some AI researchers are skeptical; perspectives diverge on whether progress will be sustained, accelerate, or plateau.
Fundamental uncertainty stems from deep questions about the nature of intelligence and the capabilities of the human brain.
Two competing perspectives on the complexity distribution of work tasks the human brain can perform:
- Panel 1 perspective: human brain capabilities are effectively unbounded; automation will continue to shift workers into more complex tasks.
- Panel 2 perspective: there is an upper bound to human brain task complexity; if true, modern AI systems are catching up fast.

Scenario I (traditional, business as usual):
- Advances in AI boost productivity and automate a range of cognitive work tasks.
- Affected workers create new jobs that are, on average, more productive than those they were displaced from.
- Corresponds to panel 1 of Chart 1.
Scenario II (baseline, AGI in 20 years):
- Over the next 20 years, AI gradually advances to AGI, able to perform all human work tasks by the end of the period, devaluing labor.
- Corresponds to panel 2 of Chart 1 with a 20-year rollout of access to the most complex cognitive tasks.
Scenario III (aggressive, AGI in five years):
- Replicates Scenario II but with AGI reached within five years.
Author’s subjective probability assessment:
- The author estimates that each of these scenarios has a greater than 10 percent probability of materializing.

Chart 2 illustrates model paths for output (left) and competitive market wages (right) across the three scenarios.
Main insights:
- First: In the business-as-usual scenario, growth continues along historical trajectories; in the two AGI scenarios, output growth is much faster because labor scarcity is no longer a constraint.
- Second: Wages initially rise in all three scenarios while labor is scarce, but they plummet as the economy approaches AGI.
- Third: Both the output takeoff and wage collapse in AGI scenarios are driven by substitution of scarce labor by comparatively more abundant machines—implying institutions could be designed to compensate workers and share AGI gains.
Caveats and model limitations:
- The model is cast in an efficient economy with labor earning competitive returns.
- Factors that may slow AGI rollout relative to technological possibility include organizational frictions, regulations, constraints on capital accumulation (such as chip supply chain bottlenecks), and societal choices about AGI implementation.
- Society may choose to retain certain human roles (e.g., priests, judges, lawmakers), creating persistent “nostalgic” jobs.

Policymakers should monitor leading indicators across multiple domains:
- Technological benchmarks: direct measures of AI performance on a wide range of labor tasks.
- Levels of investment: investment in research and development, talent, and computer chips that captures resource flows into AI development.
- Adoption measures: indicators of growing AI deployment across sectors to capture practical, useful use.
- Macroeconomic and labor market trends: productivity statistics and labor market signals that reveal realized economic implications.
Tracking complementary signals allows tailoring policy responses as AI’s effects materialize.

Adopt a portfolio approach: hedge across multiple scenarios rather than planning for a single outcome.
Stress-test existing economic and financial policy frameworks against each scenario and reform where necessary to ensure resilience.
Policy reforms may include:
- Reforming systems of taxation.
- Expanding social safety nets.
- Introducing small basic incomes that can be scaled up when necessary.
Institutional recommendations:
- Charge teams of experts with iterative scenario planning to regularly update probabilities of various scenarios.
- Embrace an adaptable, scenario-based approach to maximize benefits and mitigate risks from AI evolution.

Source: Scenario Planning for an A(G)I Future — Anton Korinek, F&D Magazine, December 2023.

Scenario Planning for An A
- Scenario Planning for An A (Markdown version)
- Scenario Planning for An A (PDF)