Guns, Bots and Steel: Should Lethal Autonomous Weapons be Banned?

The development of lethal autonomous weapons, often referred to as “killer robots,” has sparked intense debate. Proponents argue these AI-powered weapons could make conflict more precise and spare human lives. Opponents counter that removing human oversight crosses a moral threshold and these weapons may be prone to catastrophic failure. So should lethal autonomous weapons be banned? Let’s dive into the complex ethical, technical and strategic considerations.

An Introduction to Lethal Autonomous Weapons

Lethal autonomous weapons systems (LAWS) are robotic weapons that use artificial intelligence to select and engage targets without human intervention. They have sensing, reasoning and decision-making capabilities built into their algorithms. Once deployed, they can operate independently to identify threats and make the decision to use lethal force.

Also known as “killer robots,” fully autonomous weapons do not yet exist. However, militaries around the world are developing weapons with increasing levels of autonomy. These range from drones with human oversight, to automated sentry guns, to theoretical fully independent systems.

The Conversation around LAWS focuses on two key concerns:

• Loss of human control over life-and-death decisions: Removing the human from the loop crosses an ethical line and gives power to algorithms which lack morality.
• Risk of unpredictable or indiscriminate harm: The complexity of war means autonomous systems may behave in unexpected ways, leading to unintended casualties or escalation.

The stakes are high. Supporters believe autonomous capabilities could revolutionize warfare and make conflict more precise. But without regulation, critics warn LAWS could have devastating humanitarian consequences.

So should lethal autonomous weapons be banned? Let’s examine the debate in more detail.

The Case For Banning Lethal Autonomous Weapons

Several arguments have been made for prohibiting the development and use of fully autonomous weapons systems:

1. Removing Meaningful Human Control

Many experts argue lethal force should never be delegated to algorithms alone. Machines lack empathy, intuition and the ability to understand context. Their decisions to kill would be impersonal and amoral.

Taking humans “out-of-the-loop” removes accountability for loss of civilian lives. Opponents consider it morally unacceptable and an affront to human dignity.

2. Risk of Unpredictable Behavior

The rules of war are complex, yet robots cannot morally reason like humans. Autonomous weapons could misinterpret data and behave erratically, leading to unintended engagement with lawful combatants or civilians.

Their systems may interact in unpredictable ways, escalating conflict through cascading effects. Once deployed, they would be difficult to control.

3. Proliferation and Arms Racing Risks

Lethal autonomous weapons would make conflict more abstract, lowering political thresholds for war. The arms race to develop these weapons may be destabilizing.

Their proliferation could be difficult to control. Access to autonomous capabilities would empower non-state actors like terrorist groups.

4. No Accountability When Things Go Wrong

When autonomous systems fail, cause casualties or violate protocols, legal accountability would be unclear. Prosecuting design teams or commanders presents difficulties.

This lack of closure in investigating failures undermines moral norms around protecting civilians. No one could be held responsible.

5. Challenging to Program Ethically

There are no agreed standards for programming autonomous weapons to replicate laws of war and rules of engagement.

Verification techniques cannot guarantee ethical behavior in complex real-world environments. This moral uncertainty surrounding LAWS presents an insurmountable challenge.

The Case Against Banning Lethal Autonomous Weapons

Others argue autonomous weapons systems do not necessarily require a preventative ban and that their threats are manageable:

1. Gradual Technological Progress

Lethal autonomy will emerge slowly and incrementally. As capabilities improve, ethical and legal frameworks can co-evolve to ensure human oversight.

A ban would be premature. We should monitor development and legislate when necessary, not restrict technologies with potential benefits.

2. Could Reduce Risks for Civilians

Proponents argue LAWS may be better than humans at complying with the laws of war. Their autonomous capabilities could lead to more precise engagements.

They do not act out of anger, panic or self-preservation, potentially reducing civilian casualties and war crimes. A ban could remove humanitarian advantages.

3. Could Reduce Risks for Soldiers

By reducing reliance on human warfighters, autonomous capabilities may lower military casualties. Robot soldiers could take on “dull, dirty and dangerous” missions more safely.

Some proponents consider it unethical for societies to send human soldiers into harm’s way if autonomous alternatives exist.

4. Verifiability of Systems

Engineers can develop verifiable patterns of predictable behavior for autonomous systems. Their capabilities, limitations and self-diagnostic abilities can be tested rigorously.

Autonomous systems may be less prone to breakdowns and more easily understood than humans. This verifiability provides confidence.

5. No Reason to Single Out Robotics Specifically

Lethal autonomous weapons do not inherently violate the laws of war. Their autonomy should be regulated similarly to other military technologies.

A specific ban would be discriminatory. The focus should be on ethical human use, not the technologies themselves.

Technical Hurdles to Developing Lethal Autonomous Weapons

The debate often makes assumptions about autonomous weapons capabilities that are unrealistic in the near future. Developing lethal autonomous weapons that are ethical and effective faces big technical obstacles:

• Situational awareness: Obtaining comprehensive knowledge of complex environments like urban warfare remains extremely difficult. Important context will be missed.
• Adversarial environments: Opposing forces will actively try to manipulate and deceive autonomous weapons in unpredictable ways. They will be designed to exploit limitations.
• Coordinating teams: Getting swarms of systems to collaborate smoothly requires solving hard problems like conflict resolution. Individual failures could cascade.
• Unstructured environments: The real world differs vastly from test environments. LAWS must generalize safely to unfamiliar settings.
• Testing and validation: Rigorously proving effectiveness and ethical behavior across scenarios is extremely challenging. Unexpected failures will occur.
• Security vulnerabilities: Autonomous systems may be susceptible to hacking, spoofing and hijacking by adversaries.

Overcoming these challenges could take decades. Effective lethal autonomy may remain out of reach even with significant progress in AI.

The Difficulty of Defining “Meaningful Human Control”

A key debate around autonomous weapons focuses on maintaining “meaningful human control”. But what constitutes sufficient control, and how could it be defined in technical standards?

• Should a human give affirmative authorization for each individual attack? Monitor systems and veto specific targets? Or simply oversee their general use in an area?
• Does control require human judgement to be “in-the-loop” for every target, or is occasional supervision sufficient?
• How much context and information should humans have access to? Should they understand the system’s reasoning?
• If multiple humans are involved, how should responsibilities be divided? Who is ultimately accountable?

These questions have no consensus answers. Requirements like “appropriate levels” of control are subjective. Codifying nuanced concepts like “meaningful” into engineering requirements presents difficulties.

1. Defining human control remains an open technical challenge. Standards will require nuance to satisfy both ethics and capabilities.
2. Controls restricting autonomous functions may undermine military advantages. Tradeoffs will arise.
3. Even meaningful control cannot address underlying ethical objections to delegating lethal authority.

Establishing oversight policies for autonomous weapons that balance these considerations will require extensive debate between ethics, law, policy and engineering. Simple definitions or requirements are unlikely.

The Role of Public Perception

Public discomfort with “killer robots” has influenced the debate and perceptions may shape policy:

• Moral repulsion and fears autonomous weapons undermine human dignity resonate widely. But are based on assumptions about capabilities.
• Concerns over catastrophic scenarios, however unlikely, are compelling. Expectations rarely match reality.
• Lack of trust in military and technology institutions exacerbates skepticism. Transparency can address perceptions.
• Autonomous systems behave differently from humans, making interactions feel strange. Familiarity through positive exposure could build acceptance.
• Framing the debate around “robots” and “terminators” provokes reactions. More neutral terms like “autonomy” may ease fears.

Policymakers should acknowledge public unease and avoid inflammatory language. But perceptions based on science fiction should not drive disproportionate regulations. Engagement and education around actual capabilities of autonomous systems can balance discourse.

Potential Alternatives to Banning Lethal Autonomous Weapons

Rather than pursuing prohibition of lethal autonomous weapons, some experts propose alternatives:

• International conventions: Non-binding norms against inhumane autonomous weapons, like those on blinding lasers. Could stigmatize rogue uses.
• Limited autonomous capabilities: Restrict systems to narrow offensive contexts to limit risks, while retaining benefits where appropriate.
• Strict human oversight policies: Require direct human authorization for any use of force rather than broad autonomy.
• Safety certification regimes: Develop rigorous testing and validation protocols to ensure civilian safety and ethical behavior.
• Monitoring and recall mechanisms: Ensure autonomous systems have in-built tracking and can be deactivated if behaving unexpectedly.
• Forecasting and awareness: Study trajectories of autonomous weapons progress to guide policies. Monitor developments globally.

Banning autonomous weapons may be infeasible politically. Wiser alternatives could allow limited benefits while addressing immediate risks and providing frameworks to evolve governance.

6 Key Questions on Banning Lethal Autonomous Weapons

Should lethal autonomous weapons be preemptively banned?

• Pro ban: Yes, we must retain meaningful human control over lethal force. Autonomous weapons are an unacceptable breach of ethics and human dignity. Their risks outweigh potential benefits.
• Anti-ban: No, autonomous weapons can be developed safely and ethically. Banning them would be an overreaction and close off advantages. We should regulate incrementally as capabilities emerge.
• Qualified: Some weapon functions could be restricted but autonomous systems still assist humans in lawful ways. A nuanced approach is required.

Are fears around autonomous weapons overblown?

• Overblown: Yes, hype exceeds reality. Achieving full lethal autonomy faces huge technical obstacles. Current capabilities remain limited and heavily supervised by humans.
• Valid concerns: No, public worries reflect genuine ethical threats. Removing humans from lethal force decisions is inherently unacceptable. Rapid progress means risks are imminent.
• Partly valid: Concerns have merit but also reflect sci-fi perceptions. With prudent development and governance, autonomous weapons can improve safety and ethics in war.

Could autonomous weapons reduce civilian harm in conflicts?

• Yes: Potentially. Their autonomous functions may enable more precise targeting and improved adherence to laws of war. They don’t act out of fear, anger or recklessness like humans.
• No: Unlikely. Thereal-world complexity of war means autonomous systems will inevitably misinterpret situations and harm civilians in unpredictable ways.
• Uncertain: It depends. Limited autonomous functions could enhance precision but fully autonomous targeting poses too many risks and uncertainties. More evidence is needed.

Are autonomous weapons analogous to other military technologies?

• Yes: Fundamentally their risks reflect how humans choose to use them, like missiles or drones. They do not need singling out through a specific ban.
• No: This technology crosses a moral line by reducing human control over life-and-death decisions. Autonomous weapons have unique risks unlike any previous weapons.
• Partly: They have some similarities but also pose distinct ethical threats from removing human agency. A nuanced governance approach is needed.

Can autonomous systems be trusted to follow legal and ethical rules?

• Yes: Their software can be verified to follow programmed rules more consistently than humans. Engineers can design and test them for adherence to laws of war.
• No: There are too many unknowns. No one can forecast how autonomous weapons will perform in complex real-world environments. Lethal force requires human judgement.
• Uncertain: It is unclear. With transparency, testing and design ethics they could follow rules reliably in limited contexts. But human supervision is still essential.

If lethal autonomous weapons are developed, how can human control be maintained?

• Authorization of individual attacks
• Real-time human monitoring with veto power
• Overall human supervision of autonomous systems
• Responsible command accountability
• Rigorous design and testing requirements
• Operational restrictions to low-risk functions

There are no easy choices. Human control likely requires a combination of technical limits, operational procedures, legal requirements and ethical norms.

Conclusion

The prospect of lethal autonomous weapons evokes apprehension. With mounting AI capabilities, fully robotic warfighting may one day become possible. But that capacity does not inherently make their development inevitable or their use ethical.

Societal aversion to relinquishing human control over life-and-death decisions warrants respect. However, preemptively banning autonomous weapons overlooks potential advantages and risks constraining benign applications.

With prudent governance, autonomous technologies could enhance adherence to legal and ethical norms in warfare in limited contexts, while ensuring meaningful human oversight. But unsupported development of full lethal autonomy would cross a moral line.

There are no straightforward solutions. As we enter this complex and high-stakes debate, wisdom lies in promoting nuanced discourse over reactionary positions, and emphasizing our shared humanity.