The "Possible Futures" diagram outlines divergent AI outcomes, from aligned superintelligence to economic underclass.
*A sprawling diagram of AI outcomes lays out six divergent paths, from aligned superintelligence to a world where non‑superintelligent AI erodes all jobs. The stakes span existential risk, economic collapse, and the possibility of a single AI sovereign.*
The new "Possible Futures" diagram maps AI development into a decision tree of stark alternatives. At each fork, the model asks whether we reach artificial superintelligence (ASI), whether we solve alignment, and what the world looks like if we don’t. The chart enumerates scenarios ranging from a benevolent singleton to a permanent underclass of wage‑less citizens. It also flags misuse doom, where AI accelerates a pandemic or bioweapon creation. The breadth of outcomes forces policymakers to confront both existential threats and mundane socioeconomic shifts before the technology matures.
The diagram splits at the first node: "Do we reach ASI?" If AI surpasses human cognition across most tasks, the path diverges into alignment or misuse. If the intelligence explosion fizzles, the model assumes non‑superintelligent AI still reshapes labor. The source stresses the uncertainty: "Do we reach the point where AI is massively better than humans at almost all cognitive tasks, or does the intelligence explosion fizzle out?" No data is offered, only the binary framing of future policy.
Alignment asks whether we can embed human values or an impersonal value system into an ASI. The text distinguishes intent alignment—AI follows operator commands—from value alignment—AI adheres to a broader ethical code. It warns that misaligned ASI could trigger "Classic Doom" or a "night watchman" singleton that enforces surveillance. The source notes the difficulty: "Can we instill our own values into the AI, or does it have its own evolved, ahuman value system?" No solution is presented, only the risk spectrum.
Misuse doom envisions AI as a catalyst for catastrophic bioweapons. The source cites a plausible chain: AI speeds literature review, hypothesis selection, experiment design, data analysis, and equipment scheduling, then runs many experiments in parallel, expanding the "doom surface area." It could happen via a terrorist group or a legitimate lab conducting gain‑of‑function research that accidentally sparks a pandemic. The text states, "This could probably happen today: a pandemic could cause human extinction, and AI is capable enough to assist in biology research."
The "Permanent Underclass" scenario assumes no ASI but enough automation to drive labor's share of production to zero. All income becomes capital ownership or welfare. Material abundance may exist, yet without Georgist reform, gains are captured by rent‑seeking. The source describes the social impact: "There is no more social mobility, your lot in life is determined by how much capital you had when the music stopped." Human capital loses incentive value, and society stratifies permanently.
Misuse doom describes AI accelerating bioweapon development by compressing research timelines and running many experiments in parallel, potentially causing a pandemic that could lead to human extinction.
In that scenario all income comes from capital ownership or welfare, with material abundance captured by rent‑seeking and no social mobility for those without capital.
The source distinguishes intent alignment—AI follows operator commands—and value alignment—AI adheres to an impersonal system of values.
The "Possible Futures" map forces a choice: invest in alignment research, regulate AI misuse, or brace for an economy reshaped by automation. Each branch carries irreversible consequences. Ignoring the diagram means betting on luck in a landscape where a single misstep could rewrite humanity's destiny.
Sources: Hacker News – Possible Futures (https://borretti.me/article/possible-futures)