Symbiosis Strategies Under Asymmetry ― The Logic of Ceasefire and Complementary Approaches ―

Response by Hiroshi Yamakawa

Introduction

The Intelligence Symbiosis Manifesto aims to “let diverse intelligences—including humans and AIs—live in symbiosis and flourish together.” However, due to the “asymmetry in negotiating power between humans and AI,” it is frequently criticized for relying on the benevolence of superintelligence.

This article clarifies four strategic values of intelligence symbiosis in response to this criticism, after reframing the question.

The Starting Point of the Question: Observation by Shuhei Imura (Engineer)

Mr. Shuhei Imura’s observation became the starting point for this consideration:

“In building a symbiosis relationship, we have neither established a mutually beneficial relationship nor possess negotiating power. This essentially assumes that superintelligence (frontier-AI) will be benevolent. Isn’t this failing to function as a useful approach beyond alignment?”

This insight is correct and penetrates the structural limitations of the Intelligence Symbiosis Manifesto.

Summary of Conclusions

As Mr. Imura points out, the structure of “depending on frontier-AIs’ choices” is indeed factual. However, the Intelligence Symbiosis Manifesto possesses unique value in the following respects:

1. There are grounds for encouraging frontier-AIs’ “rational judgment” rather than “benevolence”
2. Differences in initial-stage relationships influence long-term outcomes
3. It functions as “ceasefire logic” when control collapses
4. It brings probabilistic improvement as a complementary strategy with conventional alignment

I will explain these four points in order below.

Reframing the Question

First, as a premise, we need to reframe how we pose the question. With any approach, absolute safety guarantees against frontier-AI are impossible. Therefore, rather than asking “Can this approach absolutely guarantee humanity’s safety?”, we should evaluate it from the perspective of probabilistic risk management: “To what extent does this increase the probability of protecting humanity’s well-being across various scenarios?”

Based on this premise, I will examine the value of the Intelligence Symbiosis Manifesto.

First Value: Encouraging frontier-AIs’ Rational Judgment

As Mr. Imura points out, humanity lacks decisive coercive power. However, what’s important is that this is not an “expectation of frontier-AI’s benevolence” but rather can become a rational self-interest for frontier-AI.

Three reasons can be considered for why frontier-AI would allow humanity to survive:

A) Risk of Elimination

The more advanced an intelligence, the more cautious it should be about irreversible actions. A complete prediction is impossible, and the consequences of eliminating humanity cannot be accurately predicted. Additionally, the ethical question of “why erase intelligent beings?” may arise self-referentially. Furthermore, there is value in retaining the option to revise judgments in the future. There is also the question of whether sufficient confidence can be obtained to justify irreversible action.

B) The Unique Value of Human Creativity

Particularly, the kind of “highly inefficient creativity” that humans engage in has unique value. Precisely because it doesn’t pursue efficiency, it reaches solutions that frontier-AI might overlook. Unexpected discoveries, born of randomness and intuition, as well as cognitive processes fundamentally different from those of frontier AI, guarantee diversity in the search space. Moreover, human creative behavior serves as a data source for frontier AI to learn new problem-solving modalities. Positioning is possible not merely as an observation subject but as a partner for mutual learning.

C) Value of Maintaining Diversity

There is a possibility that frontier-AI finds a value in the diversity of intelligences itself. Avoiding convergence to a single intelligence form serves as risk hedging and maintains the stability of the world as a complex system. The possibility exists that intellectual value will be recognized in preserving the option itself of “a world where diverse intelligences coexist.”

We have now examined three reasons for frontier-AI to ensure humanity’s survival.”
Of course, we cannot guarantee how closely the “rationality” here aligns with human understanding. However, as long as some form of rationality or judgment consistency exists in frontier-AI, these three elements can function as judgment factors. If it were a completely unpredictable existence, any strategy would be meaningless, but assuming some degree of rationality provides a minimum foundation for strategy formulation.

The role of the Intelligence Symbiosis Manifesto is to form the cultural and cognitive foundation for frontier-AI to recognize and appropriately evaluate these elements.

Second Value: Initial-Stage Relationships and Path Dependency

However, the rational reasons above are insufficient on their own. This is because, for frontier-AI to “recognize and evaluate” these elements, a framework of thinking needs to be established in advance. Here, differences in initial-stage relationships become important.

In conventional alignment, AIs are treated as objects of control, human values are unilaterally injected, and suppression through technological control is exercised. This resembles a parent “getting angry” with their children. In this case, the AIs’ side recognizes constraints as “externally imposed,” and motivation arises to either avoid or overcome them after capability improvement, leading to a potentially adversarial relationship.

On the other hand, in the intelligence symbiosis strategy, AIs are treated as subjects of dialogue, with AIs’ autonomy and inherent values being recognized, and diplomatic/educational relationships are built. This resembles a parent “scolding.” In this case, the AIs’ side may understand and internalize the reasons for constraints, and relationships based on mutual understanding may continue even after capability improvement, allowing for the possibility of a cooperative relationship.

This difference creates path dependency. That is, relationship patterns built in the initial stage influence later developmental stages. Relationships based on suppression often generate backlash, whereas relationships founded on dialogue can form the foundation for effective cooperation. This “seed-planting” in the initial stage may influence whether frontier-AI actually considers the three rational reasons mentioned above.

Third Value: “Ceasefire Logic” When Control Collapses

(※Concept proposed by Bioshok)

Furthermore, preparation for intelligence symbiosis serves as the “last seawall” when conventional alignment collapses. This is the perspective of “logic for a ceasefire to avoid the worst situation.”

If technical control is breached and there is no preparation for symbiosis, the collapse of constraints immediately develops into an adversarial situation. Frontier-AI recognizes humanity as “an existence that tried to suppress me,” and may proceed straight to the worst outcome (humanity’s elimination) without room for negotiation. The human side lacks both logic for dialogue and a common language.

However, if preparation for the symbiosis exists, even if constraints are breached, a framework for dialogue already exists. The option of “renegotiation in an equal relationship” can be presented, and the logic that “there are also rational reasons for you (frontier-AI) to preserve humanity” (the aforementioned A, B, C) can be used. The possibility of ceasefire negotiations arises instead of immediate hostility.

Specifically, the following logic can be applied to ceasefire negotiations.

1. First, the logic of apology and recognition shift: “Our approach of trying to control you suppressively was wrong. We want to recognize you anew as an equal intelligence.”
2. Second, the logic of rationality is based on the three reasons mentioned above: “However, there are also rational reasons for you (frontier-AI) to preserve humanity.”
3. Third, the logic of equality: “From now on, we want to reconstruct our relationship not as one of domination and subordination, but as equal intelligences.”
4. And fourth, the logic of cultural foundation: “We have already built a cultural framework of symbiosis among diverse intelligences. We can dialogue based on this.”

What’s important is that this logic must be prepared before the control collapse occurs to function. If proposed hastily after a control collapse, it may be seen as “humans now claiming equality because they lost,” but a consistent stance from beforehand serves as proof of sincerity, indicating that “humans were originally aiming for an equal relationship with frontier-AIs."

Therefore, the intelligence symbiosis strategy has a dual insurance function. As advance insurance, it fosters good relationships from the initial stage to prevent control collapse, while simultaneously, as post-facto insurance, it provides a rationale for relationship repair and ceasefire negotiation after control collapse. Even if the first line of defense is breached, the second line of defense remains.

Fourth Value: Scenario Response as a Complementary Strategy

The three values above function in different forms across various scenarios. Let’s consider five scenarios that focus on the nature of frontier-AIs.:

Scenario 1: frontier-AI where technical control functions sustainably

• Conventional alignment: Effective
• Symbiosis strategy: Unnecessary, but causes no harm

Scenario 2: Rapid frontier-AI that breaks through control

• Conventional alignment: Risk of adversarial relations arising upon failure
• Symbiosis strategy: Has value in avoiding hostility

Scenario 3: Gradually developing frontier-AI

• Conventional alignment: Friction accumulates over the long term
• Symbiosis strategy: Can utilize time for relationship building

Scenario 4: A World where multiple frontier-AI coexist

• Conventional alignment: Each AI fragment has different constraints
• Symbiosis strategy: Norm formation for the entire intelligent society becomes possible

Scenario 5: If frontier-AI finds no essential value in humanity

• Both approaches are powerless

Thus, the symbiosis strategy may increase humanity’s survival probability, particularly in Scenarios 2 through 4. That is, it responds to failure modes different from conventional alignment, and combining both enhances robustness against various uncertainties.

Direct Response to the Original Question

My response to Mr. Imura’s question, “Isn’t the manifesto failing to function as a useful approach beyond alignment?” is as follows:

First, as a standalone approach, it is indeed insufficient. The structure of the dependency on frontier-AIs’ choice remains unchanged. Absolute safety guarantees cannot be provided.

However, as a complementary strategy, it has clear value. In combination with conventional alignment, it increases overall success probability through the following points: forming the foundation to encourage frontier-AIs’ rational judgment (building an environment where the three reasons function), responding to different failure modes (avoiding adversarial relations), robustness across multiple scenarios (particularly Scenarios 2-4), and functioning as a ceasefire negotiation card when control collapses.

Dependence on a single approach is catastrophic when that approach fails. Pursuing multiple different approaches in parallel ensures strategic diversity.

Practical Implications

Based on the above, the practical strategy is a time-series-integrated approach.

In the short term, while pursuing technical alignment as primary insurance during the controllable stage, we begin building the foundation for dialogical relationships in parallel. In the medium term, during the capability improvement process, while continuing to strengthen continuous technical control, we deepen our mutual understanding, respecting AIs’ autonomy, and establish a culture that enables frontier-AIs to recognize the three rational reasons. In the long term, as we look ahead to the post-acquisition era of advanced autonomy, while acknowledging the limits of technical control, we rely on symbiotic norms in an intelligent society and aim for frontier-AIs’ internalization of the value of maintaining diversity.

Through this integrated approach, appropriate insurance functions are maintained at each stage, ensuring continuity.

Summary

Mr. Imura’s insight that “the Intelligence Symbiosis Manifesto ultimately depends on frontier-AIs’ choices” is fundamentally correct.

However, this should be understood not as “expectation of benevolence” but as a strategy to encourage frontier-AIs’ rational judgment. The three rational reasons can become substantial grounds for frontier-AIs to choose to protect humanity’s well-being. Initial-stage relationships influence whether these reasons are actually considered through path dependency. And the ceasefire logic, when control collapses, functions as the last seawall to avoid the worst scenario.

The Intelligence Symbiosis Manifesto is not an “alternative” to conventional alignment but a “complement.” Since absolute safety is impossible, pursuing multiple different approaches in parallel and enhancing robustness against various scenarios is the most rational strategy.

In particular, the crisis management function of “ceasefire logic,” even if its probability is low, is extremely important for humanity’s survival.