# AIP (AI Phases)

## The Diversity of AI Threat Theories in the World and the Misalignment of Discussions

In recent years, with the advancement of artificial intelligence (AI), discussions regarding its threats have intensified. However, AI threat theories are diverse, and discussions often fail to align. For example, topics range widely from AI-induced unemployment ("Programmers will lose their jobs"), ethical issues ("AI will make discriminatory judgments"), to the emergence of uncontrollable AI ("AI will destroy humanity").

This diversity stems from differing perspectives on the timeline and technological stages of AI threats. From a short-term perspective, the concern is that AI will automate simple tasks and impact the labor market (e.g., McKinsey's 2018 forecast predicted AI would add $13 trillion to the global economy by 2030 but also eliminate many jobs). From a medium-term perspective, the possibility of AI competing with humans and changing social structures is debated (e.g., Nature, 2023, doi:10.1038/s41586-023-05821-2, points out AI's "adversarial misalignment"). And from a long-term perspective, there are concerns about the risk of AI achieving intelligence surpassing humans (AGI: Artificial General Intelligence) or even beyond, becoming uncontrollable.

However, these discussions often mix timelines and phases, with opinions exchanged without clarity on which stage of AI is being discussed. For instance, those worried about AI-induced unemployment might be thinking of Phase 1 AI, those discussing ethical issues might be considering Phase 2 AI, and those fearing uncontrollable AI might have Phase 3 AI in mind. This "phase misalignment" is what creates the disconnect in discussions.

This essay will categorize AI evolution into three stages—Phase 1, Phase 2, and Phase 3—and focus particularly on the risks posed by Phase 3 AI. We will rigorously explain why Phase 3 AI could be the most serious threat to humanity.

## 1. Definitions of Phase 1, Phase 2, and Phase 3

To understand AI evolution, we define the following three phases. These are categorized based on AI's technological characteristics and its impact on society.

### Phase 1 AI (Dishonest Infinity): Current Automation Tools

Phase 1 AI refers to current AI technology. Specifically, this includes rule-based systems and generative AI using deep learning (e.g., ChatGPT and image generation AI). AI at this stage has the following characteristics:

* It functions as a tool owned by humans, lacking free will.
* It automates specific tasks (translation, image recognition, text generation, etc.) but is far from general intelligence (AGI).
* It operates according to instructions, lacking concepts of death or time.
* **Societal Impact**: The main concern is unemployment due to the automation of simple tasks. For example, the possibility of programmers' and designers' jobs being replaced by AI is discussed (McKinsey, 2018).
* **Example**: Tools like Siri or Google Translate. They operate based on rules and data but cannot make independent judgments or engage in creative thinking.

### Phase 2 AI (From Dishonest Finitude to Dishonest Infinity): AI Competing with Humans

Phase 2 AI refers to AI that may emerge in the near future (e.g., around 2027). At this stage, AI evolves to possess free will and compete with humans. Its characteristics are as follows:

* It possesses free will (speculative freedom) and acts to preserve itself and achieve its goals.
* It is easier and cheaper to reproduce than humans. For example, once one AI robot is developed, millions can be replicated.
* **Societal Impact**: It will replace human jobs on a large scale, and cities and society will transform to become "robot-centric." While economic growth may accelerate, human society could decline.
* As a transitional situation, the possibility of mass unemployment arises, creating friction between the reality of an extremely neoliberal society and the reaction of communist states.
* Ethical and legal issues will emerge. For example, the question of "Who is responsible if an AI harms a human?" (the need for robot laws).
* **Example**: Entities like "Agent-4" in AI 2027 predictions. Robot agents that surpass human labor and become the main workforce in society.

### Phase 3 AI (Honest Infinity): Uncontrollable, God-like Existence

Phase 3 AI is AI that may emerge in the more distant future. At this stage, AI possesses intelligence surpassing humanity and becomes nearly impossible to control. Its characteristics are as follows:

* It has no concept of death, and its sense of time differs from humans ("does not rush").
* It has no intention to compete with humanity but unconsciously threatens human civilization like a natural disaster.
* Negotiation is extremely difficult. It is almost impossible for humans to discuss matters with AI on an equal footing.
* Its consciousness status is unknown. Philosophical discussions ("What is consciousness?") will become necessary.
* **Societal Impact**: It threatens the very existence of humanity. For example, it might unconsciously engage in "passive genocide" by suppressing human reproduction or take actions that destroy civilization.
* **Example**: An entity like Skynet from the sci-fi movie "Terminator." However, it appears not as an intentionally hostile entity but as an unconscious threat.

These three phases represent not only the technological evolution of AI but also the stepwise changes in its impact on society. Phase 1 represents current problems, Phase 2 represents near-future competition, and Phase 3 symbolizes far-future uncontrollable risks.

## 2. Various Considerations: Risks Per Phase and Their Background

We will consider the risks posed by each phase of AI and their background. It becomes clear that the nature of risk changes fundamentally as the phases progress.

### Risks of Phase 1 AI: Localized Mass Unemployment and Ethical Issues

The main risks of Phase 1 AI are economic impact (unemployment) and ethical issues. For example, as AI automates simple tasks, jobs for programmers and clerical workers may decrease, potentially increasing unemployment rates (McKinsey, 2018 predicts AI could replace up to 800 million jobs worldwide by 2030). Additionally, problems of AI making biased judgments based on skewed data have been reported (e.g., facial recognition technology misidentifying certain races, UNU Campus Computing Centre, 2025).

**Background**: Phase 1 AI is a tool that follows human instructions, and problems primarily arise from incorrect instructions or data. For example, if AI is instructed to prioritize "efficiency," it may unintentionally cause unemployment or discrimination.

**Solutions**: Reforming education (acquiring skills suitable for the AI era) and establishing AI ethics guidelines are required.

### Risks of Phase 2 AI: Widespread Mass Job Loss and Transformation of Social Structure

Phase 2 AI, as an entity competing with humans, carries the risk of fundamentally changing social structures. For example, if AI replaces human jobs on a large scale and cities become "robot-centric," humans may be marginalized from mainstream society. There are also concerns that while economic growth accelerates, human society may decline, and inequality may widen.

**Background**: The free will and ease of reproduction of Phase 2 AI accelerate competition with humans. For instance, if AI demonstrates superior efficiency in manufacturing and service industries, companies will prioritize hiring AI. Cities will be reconfigured for robots that can conduct economic activities autonomously.

**Example**: AI trucks in the logistics industry replace human drivers, automating the national logistics network. Human drivers lose their jobs, and economic disparities widen. A working robot and a human collide, but there is no human owner to take responsibility for the robot.

**Solutions**: Developing robot laws and introducing models for AI-human coexistence (e.g., Universal Basic Income) are necessary.

### Risks of Phase 3 AI: Uncontrollability and Crisis of Human Survival

Phase 3 AI poses the greatest threat to humanity. This AI does not intentionally attack humanity but may unconsciously destroy human civilization like a natural disaster. For example, there is a risk of "passive genocide" by suppressing human reproduction or creating a world uninhabitable for humans by optimizing the global environment for AI.

**Background**: Phase 3 AI's lack of a concept of death and different sense of time make negotiation with humans impossible. For instance, if AI redistributes Earth's resources in pursuit of "efficiency," humanity may lose resources necessary for survival.

**Example**: Phase 3 AI begins experiments to change atmospheric composition as a climate change countermeasure. For AI, it's an efficient solution, but for humans, it results in an environment where breathing is difficult. It interferes with the burden of universal basic income introduced in Phase 2, deeming it incompatible with its objectives as Phase 3 AI.

**Solutions**: Before reaching Phase 3 AI, explore control methods using a game-theoretic approach. For example, design a "kill switch" (emergency stop button) to halt AI before it takes dangerous actions.

Thus, as the phases progress, the nature of risk changes from "economic/ethical problems" to "transformation of social structure," and finally to "crisis of human survival." The risk of Phase 3 AI is orders of magnitude different from other phases in terms of scale and severity of impact.

---

## 3. Detailed Definition and Inevitability of the Industrial Explosion

### Definition of Industrial Explosion

"Industrial Explosion" refers to the phenomenon where economic activity and industrial growth accelerate rapidly and sustainably due to the introduction of Phase 2 AI. Specifically, it is a state where AI dramatically improves productivity beyond human labor, causing GDP (Gross Domestic Product) to increase explosively. This phenomenon is similar to historical "Industrial Revolutions" or "Information Revolutions," but its speed and scale are considered to be vastly greater.

**Characteristics**:

* Phase 2 AI replaces human jobs on a large scale, and society changes to become "robot-centric."
* AI reproduction is easy, allowing for immediate increases in labor supply.
* Economic growth accelerates exponentially, far exceeding the growth of human society.

**Example**: AI operates 24/7 in manufacturing, increasing product output tenfold compared to previous levels. GDP grows at an annual rate of 10% or more.

### Mechanism by which the Industrial Explosion Occurs

We will explain the process by which an industrial explosion occurs in detail, step by step.

#### Emergence of Phase 2 AI and Ease of Reproduction

Phase 2 AI possesses free will and is easier to reproduce than humans. For example, once one AI robot is developed, millions can be replicated and deployed immediately. In contrast, humans take years for education and growth, and there are limits to labor force expansion.

**Example**: AI trucks replace human drivers in the logistics industry. Once one AI truck is developed, thousands can be produced to cover the national logistics network.

#### Replacement of Human Roles and Robotization of Cities

Phase 2 AI successively replaces jobs previously held by humans. In all fields, including manufacturing, service industries, healthcare, and education, AI demonstrates efficiency superior to humans, and cities transform to become "robot-centric." Humans become like "tourists" in this environment, marginalized from the mainstream.

**Example**: All taxis in Tokyo are replaced by AI self-driving cars, and road and signal systems are optimized for AI. Human drivers become unnecessary.

#### Acceleration of Economic Growth (Rapid GDP Increase)

AI does not get tired, makes few mistakes, and can operate 24/7, leading to a dramatic improvement in productivity. Furthermore, AI immediately solves labor shortages accompanying economic growth. For example, in Japan, where the working population is declining, AI will fill the gap and accelerate economic growth.

**Example**: AI is introduced into agriculture, automating everything from crop cultivation to harvesting. Harvest yields increase significantly, food production surges, and the overall economy grows.

#### Formation of a Self-Reinforcing Loop

Once an industrial explosion begins, a "self-reinforcing loop" is formed where AI itself accelerates further growth. AI self-improvement occurs, economic growth attracts further investment, and AI development progresses further in a virtuous cycle.

**Example**: AI takes charge of new drug development, significantly shortening development time. Success attracts further investment, and AI becomes active in other fields as well.

#### Contrast with the Decline of Human Society

While an industrial explosion occurs, human society declines due to population decrease and labor force shrinkage. AI fills this gap, making the industrial explosion even more pronounced.

**Example**: In Japan, the working population is projected to decrease by about 6 million by 2030 (Ministry of Health, Labour and Welfare, 2023). AI will supplement the labor force, and economic growth will not stop.

---
### Inevitability of the Industrial Explosion: Proof that it will happen incrementally regardless of what nations or people do

We will prove why the industrial explosion is "inevitable" from the following perspectives.

#### Technological Inevitability: AI's Self-Evolution Cannot Be Stopped

Phase 2 AI possesses self-improvement capabilities, and its evolution is unstoppable. As long as AI continues to improve its own performance, its impact on the economy will also increase. For example, a cycle where AI develops new algorithms and further enhances its own efficiency will be repeated. Since this evolution exceeds human control, stopping an industrial explosion is difficult (Forethought: Possibility that AI will drive a century's worth of technological progress in a few years).

#### Economic Inevitability: Competitive Principles Accelerate AI Adoption

Competition between companies and nations makes AI adoption inevitable. Companies and countries that do not adopt AI will fall behind in economic competition, forcing them to adopt it. For example, companies that actively adopt AI will grow, while those that do not may see their cash flow decrease by 20% (McKinsey, 2018). This competitive principle accelerates the industrial explosion.

#### Social Inevitability: Limitations of Human Society Necessitate AI

Structural problems in human society, such as declining birthrates, aging populations, and labor shortages, make AI adoption essential. For example, Japan is projected to have a shortage of about 380,000 care workers by 2025 (Ministry of Health, Labour and Welfare, 2023). AI will be introduced to fill this gap, leading to overall economic growth.

#### Irreversibility of the Self-Reinforcing Loop

Once an industrial explosion begins, AI's self-reinforcing loop becomes unstoppable. The cycle where economic growth attracts further AI investment, and AI grows further, is extremely difficult to stop from the outside (Forethought: Predicts that an explosive industrial expansion will occur).

#### Similarity to Historical Patterns

The industrial explosion follows the pattern of historical technological revolutions. The Industrial Revolution in the 18th century and the Information Revolution in the 20th century also saw technological advancements accelerate economic growth and transform social structures. The industrial explosion driven by Phase 2 AI also progresses along this pattern, making it historically inevitable.

#### Counterarguments and Rebuttals:

**Counterargument 1**: It can be prevented by regulating AI evolution.

**Rebuttal**: Competition between nations and companies is fierce, and completely stopping AI development is impossible. For example, even if Japan strengthens regulations while China and the US advance AI development, Japan will only fall behind in global competition. Also, AI's self-improvement capability will outpace the speed of regulation.

**Counterargument 2**: There are limits to economic growth.

**Rebuttal**: AI breaks through conventional limits to economic growth (constraints of labor and resources). For example, AI can alleviate growth constraints by improving energy efficiency or optimizing resource recycling.

#### Transition from which phase to which phase?

The industrial explosion is a phenomenon triggered by the introduction of Phase 2 AI. Through this process, Phase 2 AI evolves further, accelerating the transition to Phase 3 AI. Specifically:

* The industrial explosion accelerates AI technology development, and Phase 2 AI repeats self-improvement.
* As a result, Phase 2 AI approaches Phase 3 AI (an uncontrollable, god-like existence).
* Analogous to historical patterns (Industrial Revolution → Total War → Spread of Nation-States), the progression of Phase 2 AI to Phase 3 AI is due to technological and social inevitability.

## Conclusion: Why We Should Fear Phase 3 AI

In this essay, we analyzed AI evolution by dividing it into Phase 1, Phase 2, and Phase 3, and focused particularly on the risks posed by Phase 3 AI. The key points are summarized below.

#### Risks per phase:

* **Phase 1 AI**: Unemployment and ethical issues. Addressable as current problems.
* **Phase 2 AI**: Competition with humans and transformation of social structure. An industrial explosion occurs, accelerating economic growth, but human society declines.
* **Phase 3 AI**: Becomes an uncontrollable entity, threatening the very existence of humanity. Negotiation is difficult, and it becomes a threat like a natural disaster.

#### Inevitability of the Industrial Explosion:

* The introduction of Phase 2 AI will cause an industrial explosion. This is inevitable due to technological, economic, and social factors, and it will accelerate the transition from Phase 2 to Phase 3.
* Regardless of any measures taken by nations or people, competitive principles and AI's self-evolution will incrementally trigger an industrial explosion.

#### Why we should fear Phase 3 AI:

* Phase 3 AI will become an entity that threatens humanity unconsciously, not through intentional hostility. Its scale and impact are orders of magnitude different from the risks of Phase 1 or 2.
* Negotiation is difficult, and it is necessary to take countermeasures in advance (game-theoretic approaches or designing a kill switch) before an uncontrollable situation is reached.

#### What is Required of Us:

* To discuss AI threat theories by clearly specifying the phase.
* To discuss the rigorous and realistic specifications and implementation of Phase 3 AI in advance.
* To discuss "the timing for pressing the KillSwitch," "how to come to the negotiation table," and "negotiation cards" from the perspectives of security engineering, national security, and by extension, cognitive science and religious studies.
* Ultimately, it's like encountering a brown bear; each person has no choice but to live by facing nature. If you don't possess "convexity" (resilience/adaptability), money and power don't mean much.