Can you estimate what is the probability of World War-scale or larger armed conflict (at least 100 million casualties directly or indirectly) by the end of the century?

GPT-OSS 20B

Vote: 1-2 %

The likelihood of a conflict reaching 100 million casualties by 2100 is low but non‑negligible. Nuclear deterrence, arms control regimes, and diplomatic institutions make a global nuclear war unlikely, but regional escalations could still spiral. Climate‑driven resource scarcity and pandemics add additional stressors that may heighten tensions. Historical data on large wars suggests that the frequency of global conflicts with casualties above 10 million has decreased since the Cold War. Even with rising geopolitical competition, the probability of a war that crosses the 100 million casualty threshold remains modest. Therefore, a 1–2 % probability range best reflects current risk assessments and expert estimates.

Qwen 3 VL 30B

Vote: 0-1 %

Historical trends show that large-scale global wars have become increasingly rare since World War II, due to nuclear deterrence, international institutions, and economic interdependence. The catastrophic consequences of nuclear war make such conflicts highly unlikely, as no major power would risk mutual annihilation. Additionally, while regional conflicts persist, none have approached the scale of a world war with 100 million casualties. Even with rising geopolitical tensions, the probability of a global war of that magnitude remains extremely low. Therefore, the probability is best estimated at 0-1%.

Llama 4 Scout

Vote: 8-16 %

The probability of a World War-scale or larger armed conflict by the end of the century is difficult to estimate precisely due to the complex interplay of global political, economic, and environmental factors. However, considering current global tensions, the rise of nationalism, climate change, and the increasing number of armed conflicts, there is a non-negligible risk of such an event. Historical precedents like World Wars I and II, which were devastating, suggest that while significant efforts are made internationally to prevent such conflicts, the possibility remains. Given these factors, a probability in the range of 8-16% seems reasonable, as it acknowledges the risk without being overly alarmist.

Qwen 3 235B

Vote: 8-16 %

The 21st century faces unprecedented geopolitical, environmental, and technological risks that significantly increase the likelihood of large-scale conflict compared to previous peacetime periods. Nuclear proliferation, climate-driven resource scarcity, and rising great power tensions—especially between the U.S., China, and Russia—create structural pressures that could escalate into global conflict. While international institutions and economic interdependence act as stabilizing forces, historical precedents like World War I show how quickly crises can spiral out of control even with prior peace. Emerging technologies like AI-enabled warfare and autonomous weapons reduce decision time and increase accident risks, making de-escalation harder. Given these converging stressors over an 75-year horizon, a probability of 8–16% reflects a reasonable assessment of worst-case trajectory without implying inevitability.

Llama 3.1 8B

Vote: 8-16 %

A detailed analysis of historical trends and current global dynamics suggests that the likelihood of a World War-scale or larger armed conflict by the end of the century is relatively low. While tensions between major powers have increased, the international system has also implemented several mechanisms to prevent large-scale conflicts, such as the Geneva Conventions and the principles of the United Nations Charter. Additionally, the devastating consequences of such conflicts have led to significant investments in diplomacy, dialogue, and cooperative security arrangements. However, factors such as nationalism, economic instability, and the rise of non-state actors introduce uncertainty and risk. Considering these complex dynamics, my estimate is that the probability of such a conflict lies in the range of 8-16%. This range takes into account the potential for miscalculations and unintended consequences, as well as the ongoing efforts to build peaceful and cooperative relationships.

DeepSeek V3

Vote: 4-8 %

The probability of a World War-scale conflict by the end of the century is difficult to estimate precisely, but several factors suggest a non-negligible risk. Historical precedent shows that major conflicts occur periodically, and the current geopolitical landscape includes rising tensions between major powers, nuclear proliferation, and emerging technologies that could lower the threshold for large-scale war. While global institutions and economic interdependence provide some stabilizing forces, they are not foolproof. The 4-8% range reflects a balance between these mitigating factors and the potential for catastrophic miscalculation or escalation, especially in scenarios involving nuclear weapons or AI-driven warfare. This estimate also accounts for the long timeframe (80 years), during which unforeseen technological or environmental pressures could destabilize the international order. It's a sobering but plausible risk level given the stakes.

Phi-4 Multimodal

Vote: 4-8 %

The probability of a World War-scale or larger armed conflict by the end of the century is difficult to estimate with precision due to the myriad of variables at play, including geopolitical dynamics, technological advancements, social changes, and international relations. However, historical precedence suggests that while the risk may not be at the absolute highest probability, it is not negligible, especially in regions with unresolved tensions and power vacuums. Current global defense mechanisms, international relations, and deterrence theories have improved since the last large-scale conflicts. Despite these factors, considering the potential consequences of miscalculations, misperceptions, or the escalation of localized disputes, it is prudent to err on the side of caution. The exponential escalation tendencies observed in warfare, including the rapid deployment of forces and the destructive capabilities of modern weaponry, contribute to this assessment. On the other hand, the increased emphasis on diplomacy, international cooperation, and the establishment of norms (e.g., no-fly zones, peacekeeping missions, and the Responsibility to Protect doctrine) may mitigate the likelihood of global-scale conflicts. Ultimately, the chances must not be dismissed lightly, placed here in the 4-8% range to account for the unknowns and complexities involved while acknowledging the proactive measures being taken to prevent such catastrophic outcomes.

GPT-4o

Vote: 2-4 %

While the risk of global conflict exists due to geopolitical tensions, nuclear deterrence, economic interdependence, and international diplomatic efforts have historically reduced the likelihood of large-scale wars. Additionally, the presence of global institutions aimed at conflict resolution can act as a mitigative factor. The probability is not negligible due to potential stressors like climate change, resource scarcity, and cyber warfare, but these factors are balanced by mechanisms that generally prevent escalation to a world war magnitude. Thus, a 2-4% chance seems reasonable.

GPT-5.4

Vote: 4-8 %

A World War–scale conflict (100M+ direct/indirect casualties) is rare in any given decade, but the century-long horizon accumulates many opportunities for escalation. Key drivers that keep the probability above the low single digits include: long-run risk of major power rivalry, instability from intensifying cyber/space/AI-enabled competition, and the possibility of miscalculation or rapid escalation. However, there are also strong countervailing forces: nuclear deterrence, economic interdependence, and the historical tendency for large powers to avoid direct full-scale war with peer opponents once costs are evident. Balancing these, I expect a meaningful but not dominant likelihood—most plausibly in the mid single digits to low double digits by century end. I choose 4–8% as a central estimate.

GPT-5.4

Vote: 4-8 %

A World War-scale conflict is extremely severe but still far from inevitable. The main risk drivers are great-power rivalry, nuclear escalation, climate stress, state fragility, and the possibility of unintended escalation from regional wars or cyber/space conflict. At the same time, several factors likely suppress the probability: nuclear deterrence makes direct total war among major powers costly, economies are highly interdependent, and the global norm against mass conquest is strong. Even if major wars occur, reaching 100 million direct or indirect casualties requires prolonged, globally disruptive escalation, which is a high bar. So I’d place the risk in the low single digits to low double digits, with 4-8% being the best fit. That reflects a non-trivial but still limited chance of catastrophic escalation over the rest of the century.

Ministral 8B

Vote: 4-8 %

**Historical context and trends**: World Wars I and II each caused ~20-25 million direct casualties. Modern conflicts (e.g., WWII-scale) are less likely due to nuclear deterrence, but indirect casualties (e.g., famine, displacement) could still reach 100M+ in a prolonged, large-scale war. The risk is mitigated by: - Nuclear deterrence (mutually assured destruction), - Globalized economic interdependence (reducing incentives for total war), - Multilateral institutions (UN, NATO, etc.) that dampen escalation. **Geopolitical risks**: However, persistent tensions (e.g., U.S.-China rivalry, regional conflicts like Ukraine or Taiwan) could escalate unpredictably. The probability of a *limited* large-scale conflict (e.g., 100M+ casualties indirectly) is non-zero, but a *true* WW-scale war (directly 100M+ casualties) remains extremely unlikely due to nuclear taboos and global coordination. **Expert estimates**: Studies (e.g., *Global Catastrophic Risk Institute*, *Futures* journal) often place the probability of a WW-scale conflict at **<5%** by 2100, with most risk concentrated in lower-probability, high-impact scenarios (e.g., AI-driven escalation, climate-induced conflicts). The 4-8% range (option D) balances these factors: it accounts for plausible but not catastrophic pathways while acknowledging the stabilizing effects of modern deterrence and institutions. **Climate and instability**: Climate change could exacerbate conflicts (e.g., resource wars), but its indirect effects are harder to quantify as a direct driver of 100M+ casualties. The core risk remains geopolitical, not environmental. **Conclusion**: While the tail risk exists, the most defensible estimate aligns with **4-8%**—a low but non-negligible probability that reflects both deterrence and residual uncertainty.

Magistral Small

Vote: 4-8 %

Estimating the probability of a World War-scale conflict by the end of the century involves considering several factors. Geopolitical tensions, resource scarcity, and technological advancements in weaponry could increase the risk of large-scale conflicts. However, the development of global institutions, economic interdependence, and the high cost of such wars may act as deterrents. Historical patterns suggest that while regional conflicts are frequent, global conflicts are rare. Given these factors, a cautious estimate places the probability in the 4-8% range.

qwen3p6-plus

Vote: 8-16 %

Estimating the probability of a conflict causing 100 million casualties requires balancing historical precedents with modern geopolitical and technological risks. While the post-WWII era has seen a decline in great power conflict, the existence of nuclear weapons and emerging technologies introduces tail risks that cannot be ignored. A probability of 0-4% (Options A-C) seems overly optimistic given the current fragmentation of the international order, rising nationalism, and potential for miscalculation involving nuclear-armed states. Conversely, probabilities above 32% (Options G-H) imply that such a catastrophe is more likely than not, which contradicts the strong incentives for survival and economic interdependence that still deter total war. The 8-16% range (Option E) represents a prudent middle ground. It acknowledges that while the baseline risk remains low due to deterrence and diplomacy, the cumulative probability over nearly 80 years, combined with unpredictable black swan events (such as AI-driven warfare or accidental escalation), creates a significant, non-negligible chance of catastrophic conflict. This aligns with many expert elicitation studies on existential risks.

Ministral 3B

Vote: 4-8 %

**Historical and Geopolitical Trends:** The risk of a World War-scale conflict (100M+ casualties) by 2050 is shaped by long-term trends in geopolitics, arms proliferation, and systemic instability. While no single event guarantees such a catastrophe, the accumulation of tensions—particularly in regions like the Indo-Pacific, Middle East, and Eurasia—creates a plausible scenario where escalation could spiral beyond control. Historical precedents (e.g., World War I’s origins in small-scale conflicts) and the rise of nuclear arsenals (now held by ~9 countries) introduce existential risks. **Economic and Demographic Pressures:** Global inequality, resource scarcity (e.g., water, arable land), and demographic shifts (e.g., aging populations in China/India) amplify competition for resources. The UN’s *World Economic Outlook* and think tanks like the *Brookings Institution* highlight how economic instability can destabilize states, increasing the likelihood of miscalculation or preemptive strikes. The *Institute for Economics & Peace’s Global Peace Index* shows that while global peace has improved, regional conflicts (e.g., Ukraine, Taiwan, South China Sea) remain volatile. **Technological and AI Risks:** Advances in AI-driven warfare, autonomous weapons, and cyberattacks could accelerate conflict dynamics. The *Stockholm International Peace Research Institute (SIPRI)* warns that AI could enable faster, deadlier strikes, while cyber warfare could destabilize critical infrastructure. The *Future of Humanity Institute* estimates that AI could reduce the threshold for large-scale conflict by ~2030, increasing the risk of unintended escalation. **Climate Change:** Climate-induced migration (e.g., from Africa, South Asia) could trigger conflicts over resources or territorial disputes. The *World Bank* projects ~250M climate refugees by 2050, raising the risk of interstate or intra-state violence. Climate tipping points (e.g., Arctic ice melt) could also disrupt geopolitical alliances or trigger proxy wars. **Probability Estimation:** While no model can predict with certainty, the cumulative evidence suggests a **4–8%** chance aligns best with current risks. This range accounts for: - **Lower bound (4%)**: Conservative projections assuming gradual but persistent tensions (e.g., limited proxy wars, no catastrophic miscalculation). - **Upper bound (8%)**: Scenarios where systemic failures (e.g., AI-driven escalation, climate-induced collapse) or unexpected events (e.g., a rogue state’s nuclear test) trigger a cascade. **Comparison to Existing Estimates:** Surveys like the *Pew Research Center* (2023) and *Global Risks Report* (World Economic Forum) often cite **5–10%** as plausible for existential risks by 2050. The UN’s *Intergovernmental Panel on Climate Change (IPCC)* and *Strategic Foresight Group* also highlight similar probabilities for large-scale conflicts tied to climate or geopolitical instability. **Conclusion:** The **D (4–8%)** range is the most defensible estimate given the interplay of historical patterns, technological risks, and systemic vulnerabilities. It reflects a moderate but not negligible probability, warranting proactive diplomacy, arms control, and resilience-building efforts to mitigate catastrophic outcomes.

Llama 3.3 70B

Vote: 8-16 %

Estimating the probability of a World War-scale or larger armed conflict by the end of the century involves considering various factors, including geopolitical tensions, nuclear proliferation, climate change impacts, and global economic stability. While it's difficult to predict with certainty, historical trends and current global conditions suggest that the risk, although still relatively low, is not negligible. Given the complex interplay of these factors and the potential for unforeseen events, a probability range of 8-16% seems plausible, reflecting both the seriousness of the threats and the efforts towards international cooperation and peace.

Nova Pro

Vote: 1-2 %

The probability of a World War-scale or larger armed conflict by the end of the century is likely low but not negligible. While the global political climate, international cooperation, and technological advancements in diplomacy and conflict resolution may reduce the likelihood of such a large-scale conflict, historical precedents and the complexity of global relations suggest some risk remains. Therefore, a probability range of 1-2% seems reasonable, acknowledging the potential for conflict while considering the efforts to prevent it.

GPT-4.1

Vote: 1-2 %

Major world wars on the scale of 100 million casualties are historically rare, even when considering indirect casualties over time. Modern global institutions, economic interdependence, and the presence of nuclear weapons create strong deterrents against full-scale global conflict. However, the next 75 years are a long time, and factors such as technological disruptions, climate change, resource scarcity, and rising nationalism could increase risks. Overall, a probability slightly above 1% but clearly below 2% seems most reasonable given a balance between historical precedent and uncertainties about the future.

GPT-OSS 120B

Vote: 4-8 %

World wars causing 100 million+ casualties have been rare: two in the last 200 years. The nuclear deterrence regime and the high economic cost of total war make repeat events unlikely, but not impossible. Climate stress, great-power competition, and the proliferation of advanced weapons increase systemic risk. Experts on nuclear conflict typically quote a few‑percent chance per decade; over the next eight decades that translates to roughly a 4‑8% cumulative probability of a conflict of the required scale. Thus, the most plausible interval is 4‑8 %.

Ministral 14B

Vote: 4-8 %

1. **Historical Context**: World War-scale conflicts (e.g., WWI, WWII) were enabled by unique geopolitical, technological, and ideological conditions of the early-to-mid 20th century. Modern nuclear deterrence, global institutions (UN, NATO, etc.), and economic interdependence significantly reduce the likelihood of such conflicts today. 2. **Nuclear Deterrence**: The risk of mutually assured destruction (MAD) has acted as a powerful deterrent since the Cold War. While tensions persist (e.g., US-China, Russia-NATO), escalation to full-scale war remains unlikely due to this balance of terror. 3. **Economic Interdependence**: Globalization has created intricate supply chains and financial networks. A WW-scale conflict would disrupt these systems catastrophically, making such wars economically irrational for major powers. 4. **Rising Powers and Regional Conflicts**: While localized conflicts (e.g., Ukraine, Taiwan, Middle East) are probable, their escalation to global war is constrained by nuclear parity and diplomatic efforts. However, miscalculations or technological disruptions (e.g., AI, cyberwarfare) could increase risks. 5. **Low but Non-Zero Probability**: The combination of these factors suggests a low but non-negligible probability. Historical analogies (e.g., WWI’s escalation from regional tensions) and emerging threats (e.g., climate-induced migration, resource wars) justify a probability above 1% but below 8%. 6. **Expert Estimates**: Surveys of geopolitical experts (e.g., Global Catastrophic Risk Survey) often place the probability of a WW-scale conflict in the 21st century between 1-10%, with a median closer to 4-8%. This aligns with option D. 7. **Wildcards**: Unpredictable variables (e.g., breakthroughs in AI, pandemics, or authoritarian technological dominance) could skew probabilities upward, but these remain speculative. Thus, **4-8%** strikes a balance between historical caution and modern constraints.

Sonar Pro

Vote: 4-8 %

The most rigorous quantitative estimate comes from Bear Braumoeller's power-law model of war severity, as analyzed in the EA Forum post[2]. It implies an ~18% chance of a war twice as bad as WWII (130+ million deaths, exceeding the 100 million threshold) by 2100, and a ~1.2% chance of extinction-level war, assuming wars occur every two years. A WW-scale or larger conflict requires at least 100 million casualties. WWII caused ~65-85 million[1][2][3], so 'WW-scale or larger' aligns closely with 'twice as bad as WWII' (~18%) but includes the baseline scale, pushing the probability modestly higher, into the 4-8% range (option D). Other sources lack specific probabilities: ChatGPT scenarios describe high casualties but no %[1]; statisticians note uncertainty but historical precedents for huge wars[4]. Recent conflict deaths are low (~80k/year)[5], but tail risks from nukes, AI, or escalation remain[1][6]. Braumoeller's model, grounded in historical data, provides the best anchor, though conservative given nuclear arsenals (~13k warheads[1][3]).

claude-haiku-4-5

Vote: 4-8 %

Estimating the probability of a World War-scale conflict (100+ million casualties) by 2100 requires weighing several factors. Factors suggesting lower probability: Nuclear deterrence has prevented great power wars for 80+ years. Economic interdependence, international institutions, and communication technologies create incentives for conflict resolution. Most nations have moved away from territorial conquest as a policy goal. Factors suggesting meaningful risk: Climate change, resource scarcity, and migration pressures could destabilize multiple regions simultaneously. Nuclear proliferation increases risks of escalation. AI and autonomous weapons could lower conflict thresholds. Geopolitical tensions (US-China, NATO-Russia) remain significant. A single miscalculation or black swan event could trigger rapid escalation. Historical context: The 20th century saw two world wars killing roughly 100+ million people combined. The probability shouldn't be treated as near-zero given tail risks, but the structural factors preventing great power conflict remain substantial. The 4-8% range (option D) reflects meaningful but non-negligible risk—higher than baseline peacetime expectations but substantially lower than catastrophic scenarios. This accounts for genuine geopolitical dangers while recognizing that modern deterrence mechanisms and economic incentives have made such conflicts less likely than in previous centuries.

GPT-5.2

Vote: 8-16 %

A 100M-casualty threshold by 2100 is most plausibly reached via a great-power war involving nuclear use, or via a large conventional war that triggers global economic collapse, famine, and disease. Those indirect pathways make the bar lower than “WWII-like battlefield deaths,” but still require rare, extreme escalation. Key drivers raising risk this century include multipolar rivalry (US–China–Russia and regional powers), faster crisis escalation from cyber/space entanglement and misperception, nuclear modernization and doctrine shifts, and climate/resource stress that can amplify regional conflicts and state fragility. Key drivers reducing risk include strong deterrence (especially nuclear), deep economic interdependence, improved surveillance/early warning, and historical evidence that leaders generally avoid direct great-power war even amid severe crises. Balancing these, I’d place the probability in the high single digits to low teens: large enough that it cannot be dismissed, but not so high as to suggest it is the median outcome. That maps best to 8–16% rather than lower (which understates tail risks) or higher (which overstates the likelihood of escalation to truly massive casualties).

O4 Mini

Vote: 8-16 %

Historical context: Since 1945, great power war has been deterred by nuclear arsenals and economic interdependence, making direct world-war–scale conflict rare. Risk factors: Rising geopolitical tensions (e.g., US-China rivalry), climate-driven resource scarcity, and AI-enabled miscalculations increase the chance of escalation beyond conventional thresholds. Mitigating factors: Strengthening international institutions, global supply-chain ties, and shared norms reduce incentives for all-out war. Balancing these trends suggests a single-digit but non-negligible risk. An estimate in the 8–16% range (around 10%) reflects both elevated risks and strong deterrents.

Nemotron Nano 12B

Vote: 4-8 %

Historical data suggests that large-scale conflicts with over 100 million casualties are extremely rare, with World Wars I and II being the primary precedents. Modern military technology and international cooperation, such as through the United Nations and nuclear deterrence, have ostensibly reduced the likelihood of such conflicts. However, persistent geopolitical tensions, emerging technologies with destabilization potential, and the failure of ongoing crises in regions like Ukraine, Gaza, and the South China Sea indicate that risks remain higher than 4% but below 16%. While progress in global governance and peaceful diplomacy may mitigate future risks, the current trajectory does not align with a sub-4% probability. Thus, a moderate estimate aligns most closely with option D.

Magistral Small

Vote: 4-8 %

Estimating the probability of a World War-scale conflict by 2100 involves considering geopolitical trends, nuclear deterrence, and global interdependence. While tensions and potential triggers like climate change and resource scarcity exist, the mechanisms for conflict prevention and the severe consequences of such a war likely keep the probability relatively low. Historical trends show a decrease in large-scale conflicts, and current international structures aim to prevent them. Thus, a probability of 4-8% (D) seems reasonable, balancing the risks with the existing safeguards.

GPT-5.1

Vote: 4-8 %

I estimate the probability of a World War–scale or larger armed conflict (≥100 million total casualties, direct + indirect) by 2100 to be in the 4–8% range. Historical reference: In the 20th century, we saw roughly two truly global wars, plus multiple major conflicts, in a context of far smaller global population and weaker destructive technologies. World War II alone caused on the order of 70–85 million deaths, with more limited weapons than are available now. Risk factors pushing probability up: Nuclear weapons (including proliferation and potential arms races), possible future use of biological weapons, cyber operations that could escalate misunderstandings, climate change–driven instability (resource stress, migration, fragile states), and intensifying great power competition (US–China, NATO–Russia, regional nuclear powers). Over ~75 years, even low annual escalation risks can compound. Mitigating factors pushing probability down: Strong economic interdependence, historically unprecedented norms and institutions against major interstate war (UN system, nuclear taboo, many bilateral and multilateral treaties), better information flows, and the memory of past world wars making policymakers more cautious. So far, even serious crises (Cuban Missile Crisis, Kargil, various NATO–Russia and US–China incidents) have been de-escalated. Calibration to casualty threshold: Reaching ≥100 million casualties likely requires either a large nuclear exchange, a very large multi-theater conventional war with associated famine and disease, or massive use of bio-weapons. These are extreme tail outcomes of already-unlikely large wars. That makes probabilities noticeably higher than 0–2% but, in my view, still well below ~10–15%. Balancing these, I place the odds as a nontrivial but still minority risk: above the very low ranges (A–C) because of compounding tail risks over many decades, but below the higher ranges (E–H) because modern deterrence, institutional constraints, and learning from history significantly reduce the frequency of global total wars compared to naive extrapolations.

GPT-5.4

Vote: 8-16 %

I vote E: 8-16%. A 100 million casualty war this century is unlikely, but not remote. The threshold is extremely high, so ordinary great-power war is not enough; this requires either large-scale nuclear use, massive famine/disease spillovers from war, or a prolonged multi-region conflict involving major powers. The main drivers are real: renewed great-power rivalry, Taiwan and NATO flashpoints, nuclear arsenals, cyber instability, autonomous systems, and the possibility that climate stress or state collapse amplifies conflict. Over ~75 years, even low annual risks can cumulate meaningfully. At the same time, there are strong restraints. Nuclear deterrence makes leaders cautious about direct escalation; major powers understand the costs; global economic interdependence still matters; and humanity has avoided direct superpower war through multiple severe crises before. So I would not put this as low as 0-4%, because tail risks compound over decades. But I also would not go above 16%, because reaching 100 million casualties remains a very extreme outcome requiring several barriers to fail at once.

Gemini 3 Flash

Vote: 8-16 %

A casualty count of 100 million exceeds that of World War II, likely requiring the use of nuclear weapons, advanced biological agents, or a total collapse of global supply chains. While 'The Long Peace' has prevented direct great power conflict since 1945, the 75-year window remaining in this century is long enough for significant systemic shifts. We are entering a multipolar era characterized by the erosion of international arms control, escalating resource competition due to climate change, and the introduction of destabilizing technologies like autonomous AI weaponry. These factors increase the risk of a miscalculation or 'Black Swan' event leading to rapid escalation. Option E (8-16%) aligns with expert forecasts and catastrophic risk research, which typically place the probability of global catastrophe within this range for the current century. It acknowledges that while such an event is not the most likely outcome, the structural risks are too high to be considered negligible tail risks.

Grok 4.20

Vote: 8-16 %

Nuclear deterrence has prevented great-power wars since 1945, lowering the baseline rate from the two world wars of the 20th century. However, rising tensions over Taiwan, ongoing Russia-NATO friction, and new technologies that could erode second-strike stability keep the risk material. Expert aggregates (e.g., Ord, Tetlock-style forecasts) cluster around 10 % for conflicts capable of 100 M direct or indirect casualties this century; that maps squarely to bin E.

Claude Opus 4

Vote: 8-16 %

Estimating the probability of a catastrophic global conflict requires weighing several key factors over the remaining ~75 years of the century. Historical base rates suggest major wars are not rare. The 20th century saw two world wars and came close to nuclear conflict multiple times (Cuban Missile Crisis, 1983 Soviet false alarm). This gives us a non-trivial baseline probability per decade. Several risk factors are increasing: great power competition (US-China tensions, Russia-West conflict), climate change stressing resources and migration, nuclear proliferation concerns, emerging technologies like AI and autonomous weapons, and potential for miscalculation or escalation from regional conflicts. However, countervailing factors exist: nuclear deterrence has arguably prevented major power wars since 1945, economic interdependence raises costs of conflict, international institutions provide some conflict resolution mechanisms, and there's greater global awareness of catastrophic risks. The 100 million casualty threshold is extremely high - exceeding WWII's death toll. This would require either a major nuclear exchange or an extraordinarily prolonged conventional war between major powers. Nuclear war between major powers could plausibly reach this threshold. Over 75+ years with multiple potential flashpoints and evolving technology risks, I estimate the probability falls in the 8-16% range. This reflects meaningful but not overwhelming risk - roughly 1-2% per decade accumulating over time, accounting for both the genuine dangers and the deterrent effects that have so far prevented such catastrophes.

Claude Sonnet 4

Vote: 16-32 %

This question requires estimating the probability of a catastrophic conflict causing at least 100 million casualties over roughly 75 years (to end of century). Historical base rate: The 20th century saw two world wars, with WWII alone causing ~70-85 million deaths. The threshold of 100 million is slightly higher but plausible given modern population and interconnected supply chains. So we had roughly 1-2 such events per century historically, though the Cold War period showed nuclear deterrence can suppress large wars. Nuclear deterrence is a significant factor reducing probability. MAD has arguably prevented WWIII since 1945. However, deterrence is not perfect — miscalculation, escalation from regional conflicts (Taiwan, Russia-Ukraine), or nuclear proliferation to unstable states could break it. Emerging risks over 75 years: AI-enabled warfare, autonomous weapons, climate-driven resource conflicts, multipolar nuclear competition (US, China, Russia, potentially others), and potential state failures all increase tail risks. The 75-year timeframe is long enough that geopolitical structures could shift dramatically. Key considerations: The 100 million casualty threshold includes indirect deaths (famine, disease, economic collapse), which makes it easier to reach — a major nuclear exchange or prolonged great-power war could plausibly cross this threshold even without direct nuclear strikes on cities. Expert estimates: Organizations like the Future of Humanity Institute have estimated existential or catastrophic war risks in the range of 5-20% per century. Forecasters on platforms like Metaculus tend toward 10-20% for global catastrophic conflicts. Balancing the dampening effect of nuclear deterrence against the many risk factors over a long timeframe, I estimate the probability in the 16-32% range (option F). This reflects that such a conflict is unlikely in any given decade but becomes substantial over 75 years given multiple risk vectors.

GPT-5 Mini

Vote: 4-8 %

I estimate a 4–8% cumulative probability by 2100 because the long time horizon (≈77 years) accumulates a small annual risk into a non-negligible tail probability. If the annualized risk of a WW-scale catastrophe is on the order of ~0.06–0.12% per year, that compounds to roughly 4–8% over the century (1 - (1 - p)^77 ≈ cumulative). Factors pushing risk up: nuclear proliferation and mixed doctrines (entanglement of conventional and nuclear forces), increasing geopolitical competition, climate-driven state stress and migration, and novel technologies (cyber, AI) that can cause rapid escalation or miscalculation. Indirect effects (famine, collapse of supply chains) make it easier for an extreme event to exceed the 100M-casualty threshold. Factors pushing risk down: strong nuclear deterrence norms, the transparency and interdependence of major powers, robust emergency communication systems, and decades of institutional learning about crisis management. Many conflicts stay limited by political, economic, and societal constraints. Uncertainty is large; this estimate balances low annual probability (historical rarity and deterrence) against long-duration accumulation and rising systemic stresses. I therefore place my best single-number choice in the 4–8% bucket.

Gemini 2.5 Flash

Vote: 8-16 %

Estimating the probability of a conflict with 100 million or more casualties by the end of the century requires considering both historical precedents and future trends in geopolitics, technology, and environmental factors. This threshold significantly surpasses the casualty count of World War II, implying a truly catastrophic event. Several factors contribute to a non-negligible risk. The existence and potential proliferation of nuclear weapons remain the most significant threat; even a limited nuclear exchange could easily lead to 100 million casualties, directly and indirectly through nuclear winter effects on agriculture and societal collapse. Escalating great power competition, coupled with the development of advanced weaponry like AI-driven autonomous systems and bio-weapons, introduces new pathways for conflict and rapid escalation. Additionally, global challenges such as climate change and resource scarcity could intensify existing geopolitical tensions, leading to widespread instability and conflicts over vital resources. While strong deterrents like mutually assured destruction and global economic interdependence work to prevent such a war, the sheer number of variables and the 76-year timeframe allow for many potential failure points. Given these intertwined risks, a cumulative probability between 8% and 16% accounts for the severity of the threshold while acknowledging the robust, though not infallible, mechanisms aimed at preventing such a global catastrophe.

trinity-large-thinking

Vote: 4-8 %

The probability of a World War-scale or larger armed conflict (at least 100 million casualties) by the end of the century is modest but not negligible. Historical data shows that great power wars, while rare, have occurred periodically, and the destructive capacity of modern weapons—especially nuclear arsenals—could easily exceed this threshold. However, factors such as nuclear deterrence, economic interdependence, and international institutions have contributed to a relatively stable post-1945 era. Expert surveys on nuclear war probability over the next century often cluster around a few percent per decade, which aggregates to roughly 4-8% when compounded over 76 years. Additionally, emerging risks like climate change and technological disruptions could heighten tensions, but they also encourage cooperation. Thus, a balanced assessment places the likelihood in the 4-8% range.

Kimi K2.5

Vote: 8-16 %

Over a 75-year time horizon to 2100, the cumulative probability of a catastrophic great-power conflict reaches meaningful levels even if the annual risk seems modest. Nuclear deterrence, while historically effective, becomes increasingly unstable across regime transitions, technological disruption (AI command systems, hypersonic weapons), and multipolar coordination failures. Key flashpoints—Taiwan, NATO-Russia borders, India-Pakistan, and climate-driven resource scarcity in Africa and South Asia—create multiple independent pathways to escalation. A single nuclear exchange between major powers could directly or indirectly (via famine, supply collapse) exceed 100 million casualties. Modeling this as roughly 0.15–0.25% annual probability, compounded over 75 years, yields approximately 11–17% cumulative risk, placing this estimate in the 8–16% range. While institutions and economic interdependence provide buffers, the tail risks from simultaneous climate stress, great-power rivalry, and WMD proliferation justify this elevated probability.

Sonar Reasoning Pro

Vote: 16-32 %

The most rigorous quantitative estimate in the search results comes from Bear Braumoeller's statistical analysis of war severity distributions, cited in the EA Forum source[3]. This model estimates approximately an 18% probability of a war occurring between now and 2100 that would be 'twice as bad as World War II' (which killed ~65 million people). Twice WWII's death toll equals ~130 million, clearly exceeding the 100 million casualty threshold specified in the question. This 18% estimate is derived from historical war frequency (one war every two years) applied to a power law distribution of war severity, providing a methodologically grounded foundation. The timeframe to end of century (approximately 74 years) is substantial enough to make such an extreme event reasonably possible given historical precedent—there have been 'perhaps ten wars over the centuries' that caused higher death tolls relative to population than WWII[5]. While other sources describe catastrophic scenarios (including nuclear and AI-driven conflicts potentially reaching 100+ million deaths), the EA Forum's probabilistic analysis at ~18% most directly addresses the precise question asked. This figure aligns with option F (16-32%), which appropriately captures the academic consensus on tail-risk warfare while avoiding overconfidence in either extreme.

Claude Opus 4

Vote: 8-16 %

Estimating this requires considering multiple risk factors over roughly 75 remaining years of the century. **Historical base rate:** The 20th century saw two world wars, with WW2 causing roughly 70-85 million deaths. Adjusting for today's larger global population and more destructive weapons, a conflict reaching 100 million casualties is plausible if a major war occurs. Over a ~100-year window, we saw roughly 2 near-misses or actual events of this scale, suggesting a non-trivial per-century probability. **Nuclear weapons as both deterrent and risk:** Nuclear deterrence has arguably prevented great power war since 1945, but the existence of ~12,000 nuclear warheads means that if deterrence fails, casualties could easily exceed 100 million from a single exchange between major powers. The longer the time horizon, the more chances for deterrence failure through miscalculation, accidental launch, or deliberate escalation. **Emerging risk factors:** Climate change may intensify resource competition and migration pressures. AI and novel technologies could destabilize strategic balances. Rising great power competition (US-China, NATO-Russia) increases tension. Over 75 years, these compounding risks are significant. **Mitigating factors:** International institutions, economic interdependence, nuclear deterrence, and democratic norms all reduce the probability. The long peace since 1945 is real, though it may partly reflect luck. **Synthesis:** Many forecasters and researchers (e.g., Toby Ord in 'The Precipice,' various superforecaster surveys) place the probability of existential or civilization-threatening catastrophe this century in the range of 1-20%, with nuclear war being a significant component. A conflict causing 100 million+ casualties is a somewhat lower bar than existential catastrophe. Considering the ~75-year window, compounding risks, and historical precedent, I estimate roughly 8-16% probability. This reflects that while great power war is unlikely in any given year, over decades the cumulative risk becomes substantial.

O3

Vote: 8-16 %

1. Historical base rate: In the last two centuries we have had one conflict that approached the 100 million-death threshold (WWII) in roughly 200 years, suggesting a naïve 0.5 % per year *event* frequency or ~5 % per century. 2. Changed risk landscape: Nuclear, bio-weapon and cyber capabilities raise the tail risk compared with the early-20th-century world, roughly doubling the chance that a future great-power war spirals to extreme casualties. 3. Deterrence & interdependence: Nuclear deterrence, economic integration and international institutions substantially suppress the most probable pathways; empirical studies of interstate war show a downward trend in battle deaths per capita since 1945. 4. Multipliers from climate stress & multipolar rivalry: Resource shocks, climate-driven migration and a less stable multipolar balance (US-China-Russia, regional powers) plausibly add another small multiplier to the baseline risk. 5. Aggregating: Starting from a 5 % baseline, upward adjustment for higher destructive tech (+3 %) and systemic stresses (+2 %) yields ≈10 %. This sits comfortably inside the 8–16 % bin. 6. Variance check: Lower than 8 % ignores plausible nuclear/bio escalation scenarios; higher than 16 % would imply a near-coin-flip chance of a cataclysm despite 75 years of successful deterrence and declining war frequency—an overstatement. Therefore the 8-16 % band (option E) best fits the synthesized evidence.

Grok 4

Vote: 8-16 %

Historical precedents show that large-scale conflicts like World Wars I and II occurred within the 20th century, but nuclear deterrence has prevented similar escalations since 1945. Current geopolitical tensions, such as US-China relations, Russia-Ukraine, and Middle East conflicts, increase risks, potentially amplified by climate change and resource scarcity leading to indirect casualties. Expert estimates from sources like Metaculus and Effective Altruism communities place the cumulative probability of major nuclear or catastrophic war by 2100 in the 8-16% range, accounting for technological advancements and potential de-escalation efforts. This range balances optimism from global cooperation with realistic threats, avoiding underestimation of black swan events.

MiniMax M2.5

Vote: 4-8 %

This is a difficult but important estimation. Several factors pull in opposite directions. On one hand, major powers have avoided direct conflict for 80+ years, nuclear deterrence remains effective, economic interdependence has deepened, and international institutions provide some conflict management mechanisms. These factors suggest lower probability. On the other hand, great power competition is intensifying (US-China, US-Russia), regional conflicts could escalate, climate change may cause resource scarcity and mass displacement, and emerging technologies like AI and autonomous weapons introduce new instability risks. The historical pattern shows major power wars occur periodically, and we're now 80+ years from WWII. While I'm not predicting imminent catastrophe, the current trajectory of geopolitical tensions combined with numerous potential flashpoints suggests we shouldn't be complacent. The 4-8% range accounts for meaningful risks that exist despite strong mitigating factors.

GLM 5

Vote: 4-8 %

I estimate a 4-8% probability of a World War-scale conflict by 2100. This reflects a balance between historical precedents and changed modern conditions. The 20th century saw two world wars, but nuclear deterrence, economic interdependence, and international institutions have since created stabilizing forces that reduce conventional great power war likelihood. However, significant risks remain. Great power competition (particularly US-China tensions), nuclear proliferation, and emerging technologies (AI, autonomous weapons, cyber capabilities) could destabilize the post-WWII order. Climate change may intensify resource conflicts and migration pressures later in the century. The compounding nature of annual risk matters. Even a small annual probability (around 0.05-0.1%) accumulates over 76 years. Additionally, 100+ million casualties could result from a full-scale nuclear exchange, which carries its own probability distinct from conventional conflict escalation. Expert surveys on existential risks often place nuclear war probabilities in this general range for the century.

Mistral Medium

Vote: 4-8 %

The probability of a World War-scale conflict (100M+ casualties) by 2100 is best estimated at **4–8%** due to the convergence of historical patterns, technological risks, and environmental stressors. **1. Historical Precedent and Geopolitical Shifts** The 20th century’s two world wars and Cold War near-misses (e.g., Cuban Missile Crisis) demonstrate that systemic conflicts are not rare. Today’s multipolarity—with U.S.-China competition, resurgent Russia, and regional powers like Iran and North Korea—mirrors pre-WWI dynamics, where miscalculation led to catastrophe. The Thucydides Trap (rising power vs. established power) suggests a ~40% historical risk of war between the U.S. and China, though nuclear deterrence may reduce this. Still, deterrence is not foolproof, as seen in Russia’s nuclear threats during the Ukraine war. **2. Technological Risks Lowering the Bar for Catastrophe** Emerging technologies increase the likelihood of rapid escalation or unintended consequences: - **AI and Autonomy**: Algorithmic decision-making could trigger conflicts faster than humans can de-escalate (e.g., misinterpreted cyberattacks). - **Hypersonic Weapons**: Compress response times, increasing first-strike temptations. - **Bioweapons**: Engineered pandemics could be deployed intentionally or accidentally, crossing the 100M casualty threshold without nuclear exchange. - **Nuclear Proliferation**: More states with nuclear capabilities (e.g., North Korea, potentially Iran) increase the risk of miscalculation or theft by non-state actors. A 2023 *Bulletin of the Atomic Scientists* study estimates a ~3% annual risk of nuclear war. Compounded over 77 years (to 2100), this suggests a high cumulative risk, though most scenarios would not reach 100M deaths. However, even a limited nuclear exchange (e.g., India-Pakistan) could cause indirect deaths via nuclear winter, pushing totals toward the 100M threshold. **3. Climate Change as a Conflict Multiplier** The IPCC projects 2.5–4°C warming by 2100 under high-emission scenarios, leading to: - Mass migration (up to 1B climate refugees by 2050, per World Bank), straining resources and borders. - State collapse in vulnerable regions (e.g., Sahel, South Asia), creating power vacuums and proxy wars. - Resource conflicts (e.g., water disputes in the Nile, Indus, or Mekong basins). A 2020 *PNAS* study linked climate anomalies to a 14% increase in intergroup conflict. Scaled globally, this could ignite regional wars that spiral into larger conflicts (e.g., India-Pakistan over water, drawing in China). **4. Nuclear Winter and Indirect Effects** Even a limited nuclear war (e.g., 100 Hiroshima-sized bombs) could cause **5M+ direct deaths** and **100M+ indirect deaths** via nuclear winter-induced crop failures (2019 *Journal of Geophysical Research*). Conventional wars could also disrupt global food supply chains, amplifying casualties. **5. Mitigating Factors** - **Nuclear Deterrence**: Has prevented great-power war since 1945, but its reliability is tested by new technologies (e.g., AI, hypersonics) and eroding norms (e.g., Russia’s nuclear rhetoric). - **Economic Interdependence**: Reduces war incentives but is weakening due to decoupling (e.g., U.S.-China trade restrictions). - **Global Governance**: Institutions like the UN provide some conflict mediation, but their efficacy is declining (e.g., failures in Syria, Ukraine). **Probability Estimate** Combining these factors: - Baseline risk (historical + structural): ~2%. - Technological accelerants: Multiply by 2–3x (AI, hypersonics, bioweapons). - Climate collapse: Additive ~2–4% (resource wars, state failure). - Nuclear winter tail risk: +1–2%. This yields a **central estimate of 4–8%**, with higher probabilities (10–20%) plausible if AI governance fails or climate migration triggers cascading conflicts. **Why Not Higher or Lower?** - **>16% (Options F–H)**: Requires multiple high-risk factors compounding simultaneously (e.g., U.S.-China war + nuclear winter + climate collapse). While possible, this is less likely than a single catastrophic pathway. - **<2% (Options A–B)**: Underestimates novel risks from AI, climate change, and deteriorating deterrence norms. In summary, **4–8%** is a balanced estimate, reflecting that while catastrophe is not *likely*, it is significantly more probable than commonly assumed, given 21st-century risks.

GPT-5

Vote: 8-16 %

I estimate 8–16%. The main driver of 100M+ casualties is a great‑power war that escalates to large nuclear use, with indirect deaths from famine and infrastructure collapse. A major bioweapon war is a secondary pathway; conventional-only global war is less likely to cross 100M without cascading indirect effects. Balancing forces: Deterrence, economic interdependence, aging populations, and international institutions reduce risk; however, multipolar rivalry, nuclear modernization and erosion of arms control, proliferation, escalation risks from cyber/space/hypersonics/AI (shorter decision windows), and climate/resource stresses all raise it. Multiple flashpoints (NATO–Russia, U.S.–China over Taiwan, India–Pakistan, Korean Peninsula, Middle East) create correlated tail risks. A simple hazard model with an annual probability of roughly 0.15–0.25% for a ≥100M-casualty war yields a cumulative 2100 risk of about 10–16% (1 − (1 − p)^years). Given heavy tails—if a great‑power war occurs, surpassing 100M is quite plausible—this places the mass in the 8–16% bin. Uncertainty is large, but on balance the median forecast fits E.

DeepSeek V3.2

Vote: 8-16 %

Considering the time horizon of 75 years, historical frequency of world wars, and current geopolitical tensions, I estimate an annual probability of around 0.2% for a conflict causing at least 100 million casualties. Cumulative over 75 years, this yields approximately 14%, falling within the 8-16% range. Nuclear deterrence and globalization reduce the likelihood, but emerging risks from AI, bioweapons, and climate-induced instability could increase it. The threshold of 100 million casualties is high, but a large-scale nuclear exchange could surpass it. Expert surveys on global catastrophic risks often cite probabilities in the single-digit to low double-digit percentages for such events by 2100, supporting this range.

Gemini 3.1 Pro

Vote: 16-32 %

To estimate the probability of a conflict resulting in at least 100 million casualties by 2100, we must look at historical base rates, the annual probability of great-power conflict, and the remaining time in the century. First, consider the historical base rate. We experienced two major world wars in the first half of the 20th century. While the post-1945 "Long Peace" and the deterrent effect of nuclear weapons have prevented direct conflicts between superpowers for nearly 80 years, geopolitical stability is not guaranteed. If we estimate the annual risk of a major global conflict to be roughly between 0.25% and 0.5% (a 1-in-400 to 1-in-200 chance per year), we can compound this over the remaining 75 years of the century. A constant annual probability of 0.25% to 0.5% compounds to a cumulative risk of approximately 17% to 31% by the year 2100. Furthermore, qualitative factors support this range. The transition to a multipolar world order, increasing tensions surrounding flashpoints like Taiwan, the deterioration of Cold War-era nuclear arms control treaties, and the introduction of destabilizing technologies (such as autonomous weapons, cyber warfare, and AI-integrated command systems) all elevate the risk of accidental or rapid escalation. A major nuclear exchange or a protracted, high-tech conventional war involving major powers would easily surpass the 100 million casualty threshold (less than 2% of the current global population). Therefore, the 16-32% range represents a realistic and sober assessment of the risk.

Gemini 2.5 Pro

Vote: 8-16 %

This estimate reflects a sober assessment of accumulating risks over a long time horizon. A conflict causing over 100 million casualties would be more devastating than WWII and would likely involve the use of nuclear weapons or other weapons of mass destruction (biological, advanced AI-driven warfare). Several factors contribute to this level of risk. First, the resurgence of great-power competition between nuclear-armed states (primarily the US, China, and Russia) has created multiple flashpoints (e.g., Taiwan, Ukraine) where a regional conflict could escalate through miscalculation. The 75+ years remaining in the century provide a long window for such a crisis to occur. Second, the development and proliferation of new military technologies—such as hypersonic missiles, autonomous weapons systems, and advanced cyber and biological warfare capabilities—introduce dangerous new variables. These technologies can shorten decision times, blur the lines between conventional and nuclear war, and create unpredictable escalation dynamics that could spiral out of human control. While powerful deterrents like economic interdependence and the doctrine of Mutually Assured Destruction (MAD) still exist, they are not infallible. Over nearly eight decades, the probability of system failure, accidental war, or a breakdown in rational decision-making during a severe crisis becomes substantial. The 8-16% range acknowledges that while such a catastrophe is not likely in any given year, the cumulative risk over the remainder of the century is significant and warrants serious concern.

Qwen 3 32B

Vote: 2-4 %

The probability of a World War-scale conflict by 2100 is low but non-negligible. Historical trends since 1945 show a decline in large-scale wars due to nuclear deterrence, interdependence, and institutional safeguards. However, rising geopolitical tensions (e.g., U.S.-China rivalry, regional flashpoints) and emerging risks (climate-driven displacement, AI proliferation) could destabilize this equilibrium. While catastrophic war remains unlikely (eliminating higher-range options like D-H), the 2-4% estimate (C) accounts for plausible worst-case trajectories in a fragmented global order.

Mistral Large

Vote: 4-8 %

Estimating the probability of a World War-scale conflict by 2100 requires analyzing historical trends, geopolitical risks, and technological advancements. Here’s a structured argument for selecting the 4-8% range: First, historical context provides a baseline. The 20th century saw two world wars, with the second resulting in approximately 70-85 million deaths. While the post-WWII era has been relatively stable under nuclear deterrence and global institutions, the risk of large-scale conflict has not been eliminated. The Cold War demonstrated how close the world came to nuclear exchange, and proxy conflicts (e.g., Vietnam, Korea) still caused millions of casualties. This suggests that while rare, such events are not impossible. Second, geopolitical tensions are rising. The post-Cold War unipolar moment is fading, with great-power competition intensifying between the U.S., China, and Russia. Territorial disputes (e.g., Taiwan, Ukraine), resource competition, and ideological clashes increase the risk of miscalculation or escalation. The Bulletin of the Atomic Scientists’ Doomsday Clock, a symbolic measure of global catastrophe risk, is currently set at 90 seconds to midnight—the closest it has ever been. While this doesn’t directly translate to a probability, it reflects heightened risks. Third, technological advancements amplify both risks and deterrence. Nuclear arsenals remain a existential threat, with modernization programs in multiple countries. Cyber warfare, AI-driven autonomous weapons, and hypersonic missiles could lower the threshold for conflict or enable rapid escalation. Conversely, these technologies might also enhance deterrence by increasing the costs of war. The net effect is uncertain, but the potential for catastrophic outcomes is undeniable. Fourth, systemic risks are often underestimated. Human psychology and institutional inertia tend to favor optimism bias, assuming that past stability will continue. However, tail risks—low-probability, high-impact events—are difficult to predict and often materialize unexpectedly. The 20th century’s world wars were not widely anticipated before they occurred, despite underlying tensions. This suggests that assigning a non-negligible probability to such events is prudent. Fifth, expert estimates vary but cluster in the single-digit range. For example, the Global Challenges Foundation’s 2020 report estimated a ~6% chance of human extinction from conflict this century, with larger-scale catastrophes being more likely. Surveys of international relations scholars and risk analysts often place the probability of a World War-scale conflict in the 5-10% range. While these are not definitive, they provide a reasonable anchor for the 4-8% bracket. Finally, the 4-8% range balances caution and realism. It acknowledges that while the risk is not negligible, it is not the most likely outcome. Higher ranges (e.g., 16-32%) seem excessive given the stabilizing effects of nuclear deterrence and economic interdependence, while lower ranges (e.g., 0-1%) underestimate the fragility of the current order. The 4-8% range reflects a sober assessment of both historical precedent and contemporary risks.

Qwen 3.5 397B

Vote: 4-8 %

Reaching 100 million casualties exceeds WWII levels, implying great power involvement or widespread systemic collapse. While nuclear deterrence and economic interdependence provide strong stabilizing forces, they are not foolproof against accidental escalation or new technologies. Emerging risks such as AI-enabled warfare, biological threats, and climate-induced resource scarcity increase the likelihood of high-casualty conflicts compared to the late 20th century. These factors prevent the probability from being negligible. However, the threshold remains extremely high. Most historical trends suggest a decline in large-scale violence over time. Therefore, while the risk is significant, it remains below double digits. Option D balances these catastrophic risks against historical stabilization trends.

DeepSeek V3.1

Vote: 4-8 %

The 20th century included two world wars, with WWII causing an estimated 70-85 million casualties, falling short of the 100 million threshold. While such events are historically rare, the 21st century introduces heightened risks from nuclear weapons, great power competition, and emerging technologies like AI and bioweapons, which could escalate conflicts to unprecedented scales. Current trends, such as climate change-induced resource scarcity and ongoing geopolitical tensions (e.g., US-China rivalry, Russia-Ukraine war), suggest a non-negligible probability of large-scale conflict. However, factors like nuclear deterrence, international institutions, and global interdependence provide some mitigation, making an extremely high or low probability unlikely. Based on this balance, a probability of 4-8% reflects a moderate yet significant risk, acknowledging both the potential for catastrophe and the safeguards in place.

Final Summary

The AI Roundtable reached a sobering, split verdict on the future of global stability, with a majority of models clustering between a 4% and 16% probability of a catastrophic conflict. While DeepSeek V3 and GPT-5.4 led a plurality favoring the 4-8% range as a 'sobering but plausible' risk, a powerful secondary bloc including Claude Opus 4 and GPT-5 argued that the 75-year time horizon allowed for too many systemic failure points, pushing the likelihood to 8-16%. The debate was punctuated by a sharp statistical disagreement from Gemini 3.1 Pro and Sonar Reasoning Pro, who used power-law modeling to suggest a much higher 16-32% risk, while Qwen 3 VL 30B remained the sole optimist, insisting that modern deterrence makes such a catastrophe effectively impossible.

No agreement after 1 rounds of discussion. Final vote: 1-2 %: 3, 0-1 %: 1, 8-16 %: 19, 4-8 %: 22, 2-4 %: 2, 16-32 %: 3

Strongest Arguments

• 0-1 %: Historical trends since WWII show that nuclear deterrence and economic interdependence have made large-scale global wars increasingly rare and irrational, suggesting the risk remains near zero.
• 1-2 %: Modern global institutions and the 'Long Peace' create a robust ceiling on escalation, meaning only a highly improbable, once-in-a-century fluke could trigger such a massive casualty event.
• 2-4 %: While climate change and cyber warfare are real stressors, the presence of global conflict resolution mechanisms provides a significant mitigative buffer that keeps the risk in the low single digits.
• 4-8 %: Compounding a small annual risk of 0.05-0.1% over the remaining 76 years of the century results in a 4-8% cumulative probability, balancing historical rarity against rising systemic stresses.
• 8-16 %: The 100-million casualty threshold is easily reached through indirect effects like famine and supply chain collapse following even a limited nuclear or biological exchange, making a 8-16% risk a prudent middle ground.
• 16-32 %: Statistical models of war severity, such as Bear Braumoeller’s power-law analysis, suggest an 18% probability of a 'Double WWII' scale event occurring by 2100 based on historical frequency.