Confronting the Intelligence Curse, w/ Luke Drago of Workshop Labs, from the FLI Podcast

Resource curse analogy for AI economies

Just as oil-rich states invest in extraction rather than citizens because it offers higher returns, societies relying on AI systems for production will face perverse incentives to prioritize capital investment in machines over education and empowerment of people.

Economic value as political bargaining power

Historical democracies like post-Magna Carta England emerged when diffuse actors controlled material resources; Drago argues that when humans lose economic production value, they lose the bargaining chips necessary to secure rights, creating a precarious 'permanent pensioner' state dependent on elite goodwill.

Capital compounding without labor constraints

Unlike past technologies that required human operators, advanced AI allows capital to convert directly into economic results, potentially triggering rapid wealth accumulation by existing asset holders and sudden spikes in inequality.

Bottom-up automation of white-collar work

AI is replacing entry-level positions (analysts, junior lawyers, software engineers) first, collapsing the talent pyramid from the bottom and eliminating the pipeline that feeds senior leadership, unlike previous automation that tended to augment knowledge workers.

Early warning metrics to monitor

Drago identifies declining economic mobility, rapidly accumulating income inequality, and rising unemployment specifically among 22-25 year olds in automatable sectors as key indicators that the intelligence curse is materializing.

Zero-to-one automation for physical labor

While knowledge work faces gradual pyramid replacement, blue-collar physical labor faces a 'zero-to-one' cliff where robotics limitations currently protect jobs, though management and coordination roles may automate before the physical workers themselves.

Legal protections and shadow automation

Roles like judges resist full automation due to institutional legitimacy requirements, but face risks of 'shadow automation' where all decision-makers rely on the same AI model, creating systemic fragility and homogenized judgment despite human figureheads.

Tacit knowledge and taste as moats

Local, embodied knowledge and aesthetic judgment—exemplified by artists who fine-tune models on their own work and curate outputs—remain harder to automate than explicit cognitive tasks, potentially preserving economic value for those with distinct taste.

Commoditize intelligence through open source

Drago recommends aggressive investment in open-source AI to prevent excessive economic rents flowing to model owners and to avoid concentration of political power among a small group of technology controllers.

Architect for augmentation, not replacement

Companies should design systems that empower individual users while allowing them to retain control over economically valuable data, deliberately choosing tool-based architectures over autonomous replacement of human labor.

Individual career hedging strategies

Individuals should guard valuable tacit know-how carefully, develop 'NF1' (non-fungible) career paths that leverage unique judgment, and pursue ambitious moonshot projects sooner rather than later while economic windows remain open.