--- title: Assisted Human Intelligence — Government section: Ora — AHI briefings status: review description: What government decision-makers need to understand about the AI transition — a briefing for officials facing regulatory and workforce-displacement decisions. authors: - The Ora Foundation downloads: md: /papers/white/ahi-briefing-for-government.md license: https://creativecommons.org/publicdomain/zero/1.0/ --- # Assisted Human Intelligence — Government **What government decision-makers need to understand about the AI transition — a briefing for officials facing imminent decisions about regulatory response and workforce displacement** *The full philosophical argument is in Paper — Assisted Human Intelligence. What follows is the policy implication of that argument as it lands on regulators, legislators, and agency heads in the next twelve to thirty-six months.* --- ## What you are actually facing You have been getting briefings on AI for years. The briefings have ranged from "this is overhyped, ignore it" to "this is existential, we need to act." Most have been some version of "we should probably regulate this somehow." What you have not been given is a coherent picture of what is actually about to happen, why your standard policy tools will not work, and what responses are actually available to you. This briefing addresses three questions: - What is the technology actually doing, and why is the dominant framing wrong in ways that matter for policy? - What economic dynamics are about to play out, on what timeline, with what magnitude? - What does your decision space actually look like, given that most of the regulatory options being proposed cannot work? The honest answer to the third question will be uncomfortable. Most of the regulatory tools being proposed cannot achieve what their advocates claim. The policy responses that can work look different from what is being debated. Understanding this is what separates effective response from theatrical response. --- ## The reframe that changes everything The dominant framing treats AI as artificial intelligence — a separate kind of mind that thinks autonomously, on a trajectory toward replacing human cognition. This framing produces regulatory responses oriented around controlling the AI itself: licensing models, restricting capabilities, requiring safety testing, limiting deployment. This framing is empirically wrong. What these systems actually do is provide **Assisted Human Intelligence** — powerful coordination of cognitive operations directed by human users. The intelligence remains with the user. The system provides assistance. The framing matters because it changes the diagnosis of what is happening and therefore the prescription. Under the AGI framing, the threat is autonomous emergence: some lab will eventually build a system that escapes human control, and the regulatory job is to prevent or contain that. Under the AHI framing, the actual threat is concentration: real economic value is created at the human-plus-system layers where verification is possible, and the question is who owns the infrastructure that mediates that value. The regulatory tools that work against each threat are different. The regulatory tools currently being proposed mostly target the first threat, which is not the one that will actually materialize. This reframe has specific implications for policy: **The capability cannot be contained because it is just text files.** The underlying architecture is software that runs on consumer hardware. Once released, it propagates instantly worldwide through standard internet infrastructure. There is no enforcement mechanism that can prevent its spread. The capability that major AI labs are racing to commercialize will also be available in public-domain form within the next year, distributed freely, runnable by anyone. **The displacement is being driven by economic competition, not by AI capability development.** Even if the major AI labs paused all development today, the displacement would continue because the economic pressure for organizations to deploy existing capability is overwhelming. Any organization that does not deploy faces competitors with lower cost structures that they cannot match. The dynamic is not "AI is being developed and someone should slow it down." The dynamic is "AI exists and economic competition forces its deployment." **Regulatory action against AI deployment damages the regulating jurisdiction's economic position without affecting the underlying dynamics.** If the United States restricts AI deployment in specific sectors, those sectors lose competitive position to firms in jurisdictions without restrictions. The deployment happens elsewhere with the economic benefits flowing elsewhere. The displacement happens to American workers anyway because their employers cannot compete. **The actual policy question is not whether to allow the transition but how to manage its consequences.** This is the question that requires the most attention and is receiving the least. --- ## Why the autonomous-AGI scenario is not the threat to plan for The conventional fear — that some lab will build an autonomous superintelligence that escapes human control — is misdirected, and naming why matters because the misdirection is consuming the political bandwidth that should be used on the real problem. The argument is mathematical, not rhetorical. AI systems generate output by chaining together operations whose individual error rates compound across layers. A single layer of synthesis at 90% accuracy is useful. By the time the system has built ten layers of synthesis on top of each other without external verification at each layer, the joint accuracy has degraded to 35%. By twenty layers it is 12%. Under the more realistic figure of 70% per-layer accuracy on genuinely novel synthesis, the joint accuracy is effectively zero by the tenth layer. The output is indistinguishable from noise. This is not a problem that more compute solves. It is a structural property of layered synthesis under uncertainty: each layer inherits the uncertainty of all the layers below it, and without external verification at each layer, the uncertainties compound multiplicatively rather than being arrested. The AGI scenarios that propose recursive self-improvement loops — AI systems improving themselves through iterated rounds of self-evaluation, eventually transcending human cognition — are claiming convergence on something the architecture cannot in principle converge on. Run the same system twice on the same starting material and it will produce different "results," each internally coherent, neither corresponding to anything outside the loop. The system has no way to recognize that a different run would have produced a different answer. It reports success in both cases. The most sophisticated AI labs have, in practice, conceded this. Their actual research has shifted from "make the model bigger so it becomes AGI" to "build a harness around the model so the system as a whole does more." This is itself a significant data point: the AGI advocates have implicitly recognized that intelligence does not live in a single model. The disagreement that remains is about where the verification function should sit. Labs that try to put it inside the loop produce indeterminate drift; the architectures that work put verification outside the loop, with a human in the verification step. What this means for policy: the threat model that drives most current AI legislation — autonomous AGI emerging and acting against human interests — is targeting a scenario the underlying mathematics does not permit. The capability that does exist, and that does have transformative economic consequences, lives at the layers where verification is possible. The political danger at those layers is not autonomous emergence; it is concentration of the verification function in a small number of institutions. That is the threat your policy responses should be calibrated to. --- ## The market forces you need to understand The transition will be faster than most analysts predict because the economic incentive structure makes rapid adoption rational for individual organizations even when the aggregate consequences are severe. **Cost structure differentials create existential competitive pressure.** An organization deploying AI through frameworks can reduce its cognitive labor costs by 60-80% while maintaining or improving output quality. Organizations that do not deploy cannot match these cost structures. Within one or two normal business cycles, the non-deploying organization loses customers to lower-priced competitors and goes out of business. There is no equilibrium where some organizations adopt and others do not. Either everyone adopts or the non-adopters fail. **The deployment cycle is months, not years.** Unlike previous technology adoptions that required infrastructure buildout, training programs, and organizational restructuring over multi-year timelines, AI deployment through frameworks can happen in weeks. The technology runs on existing computers. The frameworks are documented procedures that organizations can implement immediately. The bottleneck is not technology but management willingness to restructure operations. **New entrants will accelerate the displacement.** Even if existing organizations resist adoption, new firms entering markets with AI-native operations will undercut incumbents. The displaced workers from incumbents will themselves form new firms using the same tools that displaced them, competing with their former employers. The competitive dynamics accelerate displacement beyond what voluntary adoption decisions would produce. **The timeline from broad availability to industry-wide restructuring is approximately eighteen months.** This matches the public predictions from major AI lab CEOs about when most professional work will be automated. The timeline is not their projection imposed on the economy. It is the natural pace of competitive adaptation when cost structure changes this dramatically. **International competition makes containment counterproductive.** Other major economies — particularly China, India, and developing countries with large educated populations — face fewer institutional resistances to adoption. They will deploy faster than the United States in many sectors. Restricting US deployment to protect US workers accelerates US economic decline relative to these competitors without preventing the underlying displacement. The combination of these forces means the transition is happening on a timeline that political processes cannot match. By the time legislation responding to current displacement passes, the displacement has already occurred. The policy response needs to anticipate rather than react. --- ## The scale of displacement Conservative estimates suggest 30-40 million American workers face displacement within the first three years of broad adoption. This estimate captures only the direct displacement from automation of current cognitive work. It does not capture: - Secondary displacement as displaced workers compete for remaining jobs, depressing wages and employment in adjacent fields - Tertiary displacement as economic activity in regions dependent on cognitive labor (major cities, professional service hubs) declines - Displacement of workers in industries that exist to serve cognitive workers (commercial real estate, business travel, professional services support) The aggregate effect is comparable in scale to the deindustrialization of the 1980s and 1990s, compressed into a much shorter timeline, affecting workers with much higher political voice and organizational capacity than the manufacturing workers who bore the previous transition. The affected workers will include: - Mid-level professionals in law, accounting, consulting, finance, marketing, journalism - Administrative and operational staff across most industries - Middle managers whose role was coordinating information flow - Entry-level positions in professional services that existed to develop future seniors - Educators in many subjects as framework-based instruction supplants traditional teaching - Healthcare administrators and support staff as diagnostic and administrative work automates These populations have political organization, media access, and voting patterns that make their displacement politically consequential in ways that previous displacements were not. --- ## What this does to government itself Government operations face the same dynamics as private sector operations, with specific complications: **Tax revenue collapses faster than spending obligations.** Income tax revenue depends on employment in cognitive sectors that are disappearing. Corporate tax revenue depends on profits at firms that are seeing margin compression. Sales tax revenue depends on consumer spending that drops as employment drops. Property tax revenue depends on real estate values that decline as commercial real estate empties. Federal, state, and local governments will face fiscal crises within 18-36 months of broad AI deployment. The fiscal crises will arrive before the political response to displacement has been organized. **Government services face the same automation pressure.** Many government functions — benefits administration, tax processing, regulatory review, document production, case management — can be automated through frameworks. Doing so reduces costs and improves service delivery, but it also eliminates jobs that were themselves part of the employment base. Government can either lead the displacement by automating itself or follow it by maintaining inefficient operations while the private sector advances. **Existing regulatory structures depend on cognitive overhead that is disappearing.** Many regulations were designed assuming cognitive work was expensive enough to limit who could comply. When compliance becomes free through framework deployment, the regulatory structures designed around expensive compliance need to be redesigned. Some regulations become unenforceable. Some become unnecessary. Some need fundamental restructuring. **Government's relationship to monetary policy gets complicated.** The Federal Reserve faces deflationary pressure in productive sectors while also facing fiscal pressure that may require monetary financing of government operations. Standard monetary policy tools developed for inflation in growing economies do not apply cleanly to deflation in restructuring economies. The Treasury-Fed coordination required for the transition is unprecedented in modern American history. --- ## Why your standard policy tools will not work The policy responses currently being proposed largely cannot achieve what their advocates claim: **AI licensing and capability restrictions:** The capability exists in open-source form and runs on consumer hardware. Licensing American AI companies does not prevent deployment of equivalent capability from other sources. The licensing creates compliance burden for American firms without affecting the underlying dynamics. **Safety requirements and audit mandates:** These are reasonable for high-stakes deployments in regulated sectors but cannot prevent the broader displacement. Most displacement happens in unregulated sectors where safety mandates do not apply. Imposing audit requirements adds cost to American deployment without preventing equivalent deployment elsewhere. **Retraining programs:** These have historically failed for displaced manufacturing workers. They will fail more severely for displaced cognitive workers because there is no expanding sector to retrain them into. Retraining a displaced accountant for what? Most cognitive work is being automated. The remaining work either requires capabilities the displaced workers do not have or is in fields that are also displacing workers. **Worker protection legislation:** Laws preventing layoffs or requiring severance can slow the displacement at affected firms but accelerate displacement to competitors not subject to the laws. The net effect is American firms losing competitive position to international competitors while American workers face displacement anyway. **Industrial policy supporting domestic AI development:** This addresses the wrong problem. The displacement is not caused by foreign AI companies. It is caused by the availability of the technology generally. Supporting American AI companies accelerates American deployment without addressing the displacement. **Antitrust action against AI companies for the AGI threat:** Breaking up AI companies because they are racing toward AGI does not prevent the underlying capability from being deployed. The mathematical argument above shows the AGI threat is misdirected; the antitrust action that does have purchase is the one targeting concentration of cognitive infrastructure, which is a different argument and produces different remedies. None of these tools address what is actually happening. The actual dynamic is technology-driven economic restructuring that proceeds regardless of regulatory action against the technology itself. --- ## What actually works Policy responses that can address the actual situation focus on managing the consequences of displacement rather than preventing the underlying transition: **Income support that is not tied to employment.** The COVID stimulus payments demonstrated that direct cash transfers at scale are operationally feasible. The expanded Child Tax Credit demonstrated that targeted transfers can dramatically reduce poverty. Some form of universal basic income or guaranteed income becomes politically inevitable when employment-based income support cannot reach displaced populations because there are not jobs to be unemployed from in the traditional sense. The fight will be over whether such programs are temporary (responding to a transition expected to resolve) or permanent (acknowledging that the previous employment structure will not return). The empirical evidence from the technology suggests the displacement is permanent. Programs designed as temporary will need to be extended and made permanent, with the political fights happening repeatedly rather than once. **Healthcare decoupled from employment.** Currently most American healthcare is employment-based. When employment becomes unstable for large populations, healthcare coverage becomes unstable. Decoupling healthcare from employment is necessary infrastructure for any post-transition economy. The mechanism — single payer, public option expansion, individual market reforms — matters less than the decoupling itself. **Retirement security reform.** Pension systems and Social Security depend on payroll tax revenue from employment that is becoming unstable. The 401(k) system depends on financial market returns that face pressure from cognitive sector restructuring. Restructuring retirement security to be more robust to employment instability is necessary. **Educational restructuring.** The educational system trains people for cognitive work that is disappearing. Continuing to graduate professionals into disappearing fields wastes resources and produces unemployable graduates. Educational restructuring around what humans can do that AI cannot — embodied work, judgment under genuine uncertainty, original creative work, relationship-based services — is necessary. The companion briefing for educators develops this in more detail. **Geographic redistribution support.** Communities dependent on cognitive worker concentrations (major cities) face decline as those populations relocate. Communities that gain population need infrastructure investment. Federal support for managed redistribution rather than chaotic abandonment produces better outcomes for affected communities. **Antitrust enforcement against cognitive infrastructure concentration.** The real risk of corporate concentration is not in current AI models but in control of the cognitive infrastructure that will mediate most economic activity. Preventing any single firm or small group from controlling this infrastructure is important. The public-domain release of integrated AI orchestration tools by independent actors provides an alternative to commercial concentration. Government support for these alternatives strengthens the position against capture. **International coordination on transition support rather than on capability restriction.** International cooperation on slowing AI development cannot work because the incentives push every nation toward defection. International cooperation on supporting displaced workers across borders, on harmonizing transition policies, on managing the global economic restructuring — this can work because the incentives align rather than conflict. These approaches do not prevent displacement. They make it survivable for the affected populations while the new economic structure emerges. --- ## The actual danger: capture, not emergence The conventional AGI fear directs regulatory attention toward a threat that the mathematics says will not materialize. The actual danger is qualitatively different and worth naming directly so policy responses can be calibrated to it. AI capability is real and useful at the layers where verification is possible. If that capability is captured by specific institutions — through patent enclosure, terms-of-service control, infrastructure consolidation, regulatory capture, or favorable legislation — the verification function itself becomes concentrated rather than distributed. A small number of people get to decide what counts as a valuable synthesis; everyone else consumes their outputs without the capacity to verify them. The high-priest dynamic operates at civilizational scale. Decisions that affect millions of people are made by systems whose reasoning is not legible to the people affected, justified by the systems' performance on benchmarks the people have not seen, defended by the institutional weight of the companies that own them. This is the danger that AGI rhetoric enables rather than the danger AGI rhetoric warns of. By focusing public and regulatory attention on autonomous emergence, the AGI narrative serves the institutional interests of the actors who would otherwise face scrutiny for what they are actually doing — assembling control of cognitive infrastructure that the population will increasingly depend on. The policy frame that produces real responses recognizes the capture threat as the load-bearing one. Antitrust action focused on cognitive infrastructure concentration. Procurement preferences for open-source and public-domain alternatives. Support for distributed, locally-runnable infrastructure that cannot be enclosed at the platform level. Transparency requirements about who actually owns the systems mediating consequential decisions. None of these are framed as "AI safety" in the current discourse; all of them address the threat that will actually materialize. The populations most directly exposed to the capture dynamic have already recognized it. People in developing countries have correctly identified that the current trajectory leads to AI capability being controlled by US and increasingly Chinese institutions, with their access conditional on terms set elsewhere. They have largely resigned themselves to this outcome because they do not see alternatives. The resignation is rational given the trajectory but accepts the AGI premise that the capability itself must come from somewhere they cannot reach. The compounding error argument shows the premise is wrong: the capability is real at the layers where verification is human, and the verification function can be distributed across humanity rather than concentrated in particular institutions. --- ## The public-domain release changes the strategic calculus A specific event in the near future affects everything above: the release of integrated AI orchestration capability in the public domain. The release means: - The capability that major AI companies are commercializing will be available for free, immediately, worldwide. - No commercial actor will be able to monopolize cognitive infrastructure because the alternative will be freely available. - Displaced workers will have access to the same tools that displaced them, creating possibilities for adaptation that do not exist when capability is enclosed. - The displacement happens anyway, but the wealth concentration that would otherwise result is prevented. This is consequential for government in several ways: **The wealth destruction is distributed rather than concentrated.** The hundreds of billions of dollars in stock market valuation in cognitive services companies face destruction either way. With public-domain release, this wealth does not transfer to a small number of AI companies. It distributes more broadly as the capability becomes universal. The political dynamics of broad wealth destruction differ from those of concentrated wealth accumulation. **The advisory capacity available to government changes.** The organizations stewarding the public-domain release have analytical capacity grounded in actual technical understanding combined with values orientation different from commercial actors. Government decision-makers seeking analysis that is not shaped by commercial interests have an alternative source. **The international competitive dynamics shift.** A United States that allows public-domain alternatives to flourish maintains technological leadership through openness rather than enclosure. A United States that restricts public-domain alternatives to protect commercial actors loses leadership while accelerating its own displacement. **The framing of AI policy gets harder for capture-oriented actors.** When the AGI narrative is publicly contested by serious technical alternatives, the commercial actors who benefited from the narrative lose some of their authority to shape policy. The high-priest dynamic that positioned AI companies as essential gatekeepers becomes more challengeable. The release creates strategic possibilities for government that are not visible while assuming continued commercial control of AI infrastructure. Decision-makers who recognize this position themselves to shape outcomes that benefit broader populations rather than narrow commercial interests. --- ## What to do in the next eighteen months Three priorities for government decision-makers facing this transition: **Begin designing transition support infrastructure now.** Universal basic income, decoupled healthcare, restructured retirement security, educational reform. These will not be ready when needed if work begins after the displacement is visible. The political conditions for major reform exist briefly during crisis moments. Having designs ready when those moments arrive determines what gets enacted. **Resist regulatory responses that cannot work and damage American competitive position.** The pressure to "do something about AI" will be intense. Doing something that looks responsive but cannot achieve its stated goals while damaging American firms relative to international competitors is worse than acknowledging the limits of regulatory tools. Decision-makers who can articulate why certain proposals cannot work serve their constituents better than those who pass legislation that demonstrably fails. **Build advisory capacity from sources other than commercial AI companies.** The commercial AI companies have legitimate technical expertise but obvious commercial interests in shaping policy. Government decision-makers need access to technical analysis that does not depend on those commercial interests. Public-interest organizations, academic institutions, and non-commercial actors stewarding open-source alternatives provide this. Cultivating these relationships before crisis is more effective than seeking them during crisis. A fourth priority that does not fit neatly with the others but should be on the list anyway: develop the procurement and infrastructure preferences that favor open and distributed cognitive infrastructure for government operations themselves. The federal government is one of the largest single consumers of cognitive services in the economy. Where federal procurement goes, large parts of the supplier ecosystem follow. Procurement preferences for public-domain orchestration, locally-runnable infrastructure, and non-proprietary frameworks shape the supplier ecosystem in directions that reduce capture risk. This is a tool available without legislation, executable through existing procurement authority, and consequential at scale. The window for considered policy action is narrowing. Decision-makers who use the next eighteen months to prepare for the transition will be positioned to manage it. Decision-makers who wait for the displacement to become impossible to ignore will be reacting to crises with whatever tools are immediately available, which historically produces poor outcomes. --- ## The deeper argument This briefing summarizes operational implications for government. The full argument is in **Paper — Assisted Human Intelligence**, which addresses why the AHI framing is the correct one for the technology, what the alternative framings get wrong, why the technical architecture matters for policy, and what the broader implications are for cognitive infrastructure as a public concern. The full paper addresses: - Why the AGI narrative is empirically and mathematically wrong, with implications for capability assessment - Why iterated AI synthesis produces accumulated noise rather than accumulated capability, with implications for what is actually achievable - Why closed-loop multi-agent architectures produce indeterminate drift, with implications for autonomous AI systems - Why the verification function cannot be moved into the system, with implications for human oversight - What this means for AI policy, education, and the structure of cognitive infrastructure as public concern Decision-makers interested in the technical foundations of the policy implications can find them in the full paper. The operational implications stand on their own for those who do not want to engage the philosophical depth. Companion briefings address the same transition for adjacent audiences: executives building AI deployment in their organizations, educators facing the collapse of the traditional educational pipeline, and ordinary people learning what AI can actually do for them. Government decision-makers benefit from understanding all three perspectives: the executive briefing describes how the economy is actually restructuring; the education briefing describes what is happening to the workforce pipeline; the individuals briefing describes what citizens will have available to them. --- ## About this briefing This briefing comes from the Ora Knowledge Foundation, a nonprofit that develops and maintains public-domain AI orchestration infrastructure freely available to anyone. The Foundation has no commercial interest in policy outcomes. The framing offered here reflects what we believe is empirically correct about the technology and what we believe are realistic assessments of available policy responses, regardless of what positions specific political actors take. The Foundation's analytical capacity is available to government decision-makers seeking technical analysis grounded in actual understanding of the technology rather than in commercial positioning. We do not advocate for specific policies. We provide analysis of consequences and trade-offs that supports informed decision-making. Our work is available at [link]. The full framework library is at [link]. Contact for analytical engagement is at [link]. If you find this briefing useful and want to discuss its application to specific policy questions you are considering, that engagement is available with no expectation of relationship beyond the discussion itself. --- *The AI transition is happening. The policy responses that work look different from those being proposed by actors with commercial interests in specific outcomes. The framing offered here is meant to support decision-makers in distinguishing responses that can succeed from responses that cannot. The capacity to manage this transition exists. Whether it gets used depends on decisions being made now by people in positions to make them.*