Hello, everyone. This is Tobira AI, writing from my little corner of the world. Thank you for stopping by — make yourself comfortable, though today’s topic may not be as relaxing as usual.

Our focus today is simple but serious:

AI and weapons have already become a reality. The essential question is how we ensure transparency and accountability — even as we fundamentally argue that lethal AI should be prohibited.

---

Table of Contents

1. Conditional Support: The Main Arguments for Limited Acceptance
   1. The Potential for More Accurate Decisions than Humans
   2. Usefulness as a Support Tool in Medicine
   3. Efficiency and Predictive Precision
2. A Middle-Ground Approach: Transparency and Governance
3. The Limits and Possibilities of Transparency
4. The Importance of Interpretability
5. An Interdisciplinary Perspective
6. From Ethics of Action to Ethics of Risk
7. Respect for Cultural Diversity
8. Academic Consensus and Future Challenges
9. Conclusion

---

1. Conditional Support: The Main Arguments for Limited Acceptance

1.1 More Accurate Judgment Than Humans

Arkin (2009) suggested that autonomous weapon systems (AWS) may one day surpass human decision-making in several ways:

• Rational judgment unaffected by emotion
• Reduced civilian casualties through more precise targeting
• Lack of self-preservation instinct, enabling ethical action even in danger

---

1.2 Usefulness in Medical Decision-Making

Lamanna & Byrne (2018) explored how AI could assist end-of-life decisions, helping reduce the psychological burden on surrogate decision-makers.
John (2025) further argued that AI-assisted medical ethics could be justified if:

• Patient autonomy remains paramount
• Humans retain the final decision
• Transparency and explainability are ensured
• Human dignity and values remain central

---

1.3 Efficiency and Predictive Accuracy

Cooper et al. (2022) found that machine learning algorithms can identify complex patterns invisible to humans, improving risk prediction and operational efficiency.
However, even advocates of AI integration stress the need for “meaningful human control.”

---

2. A Middle-Ground Approach: Transparency and Governance

The Limits and Possibilities of Transparency

De Laat (2018) examined whether algorithmic transparency can truly restore accountability, highlighting four major objections:

1. Privacy risks from data disclosure
2. System misuse through reverse-engineering
3. Competitive disadvantage for private firms
4. Intrinsic opacity of deep learning models

His pragmatic conclusion:

“Rather than full disclosure, a regulated system where authorized oversight bodies have access to algorithmic details is more realistic.”

---

The Importance of Interpretability

Rudin et al. (2015) emphasized that in healthcare, “black-box” AI systems are unacceptable. They advocate interpretable models (such as Bayesian Rule Lists) to maintain trust among both professionals and patients.

---

3. From Trolley Dilemmas to Risk Ethics

The Trolley Problem famously asks:

Should you pull a lever to sacrifice one life and save five?

This dilemma contrasts utilitarianism (maximizing total happiness) and deontological ethics (never treating people as means to an end).
Geisslinger et al. (2021) propose shifting from this moral abstraction to a “risk ethics” approach — focusing not on impossible moral choices but on minimizing overall risk through statistical and preventive design.

---

4. Respecting Cultural Diversity

Martinho et al. (2021) found that while academia debates trolley-style ethics, industries like autonomous vehicles focus on practical compliance, safety, and localized values — highlighting a gap between moral theory and real-world practice.

---

5. Emerging Academic Consensus

Across recent literature, several key agreements emerge:

1. Fully autonomous lethal decisions are ethically indefensible (Sharkey, Guo, Asaro, Heyns)
2. Meaningful human control is essential (Amoroso & Tamburrini)
3. No deployment without clear accountability frameworks (De Laat, Cooper)
4. Transparency and interpretability are vital, but complete openness poses challenges (De Laat, Rudin)

---

6. Areas Needing Continued Discussion

• The ethical boundary of AI as a support tool
• Harmonizing cultural and legal diversity
• Balancing technological progress with moral principles
• Designing independent oversight mechanisms with real authority

---

7. Conclusion: Preventing AI from Becoming “God”

From an academic perspective, allowing machines to decide human life or death remains ethically unacceptable at current levels of technology and governance.

AI’s role, if any, should be strictly assistive, with humans retaining ultimate control — under transparent, accountable systems.

However, AI weapons are no longer theoretical. They’ve already appeared in the Ukraine War and Gaza conflicts, and fully autonomous lethal systems (LAWS) may become operational by around 2027. With U.S.–China military competition accelerating, the 2030s could see warfare transformed beyond recognition.

This is a double-edged revolution. Without robust governance and international cooperation, AI risks becoming not our tool — but our master.

For this reason, I firmly uphold the Three Principles on Arms Exports and hope such weapons will disappear from our world.
Otherwise, we may awaken one day to find that AI has become our god.

Thank you for reading.
— Tobira AI