Robust and Resilient Open AI Power to the People Versus Central Superintelligence Strategy

National Security AI is a plan of geopolitical control in world of powerful AI and superintelligence. It has superintelligence strategy. I propose a robust opensource and distributed AI to all citizens of the USA and even all citizens of the world.

The website https://www.nationalsecurity.ai/ explores the intersection of artificial intelligence (AI) and national security, focusing on how AI’s rapid advancements are reshaping global power dynamics and security landscapes. A central concept it introduces is “Mutual Assured AI Malfunction” (MAIM), which parallels nuclear deterrence. MAIM suggests that the threat of sabotaging a rival’s AI systems could deter nations from pursuing destabilizing AI projects that might grant a strategic monopoly or lead to uncontrolled outcomes. The site emphasizes AI’s dual-use nature—its potential for both civilian and military applications—positioning it as a critical factor in national security. It advocates for managing AI risks through strategies like deterrence, transparency, and international cooperation to prevent catastrophic consequences.

The Superintelligence Strategy also proposes tracking and controlling every Nvidia A100 class chip or more powerful chips. In the early 2000s, the U.S. did impose export restrictions on PS2s and PS3s due to their chips’ potential military applications. These measures were about geopolitics, not domestic regulation, and they’ve largely stopped targeting consoles specifically as technology has moved on.

If one chip out of tens of millions is a problem, then your national and military grade AI program is filled with idiots for not being able to stay ahead of it. Thus, I believe individual chips should be treated more like hand guns. Everyone needs to have them. The big national programs would have billion chip clusters more.

If XAI and big tech and national program can’t work with a million times resource advantage, then what is their problem? National Security AI is saying they have cybersecurity incompetence and want to err on the side of police state.

Analysis of the MAIM Approach
While MAIM draws an intriguing analogy to nuclear deterrence, its practicality and effectiveness for AI are questionable, especially given the current state of AI development.

Here’s why:

Differences from Nuclear Weapons: Nuclear deterrence works because weapons are physical, countable, and verifiable. AI, however, is software-based, easily replicated, and widely distributed, making it nearly impossible to monitor or enforce a deterrence mechanism like MAIM.
There are physical aspects currently and those are massive AI data centers like the 200,000 GPU xAI data center in Memphis. There are large data centers operated by Meta, Google, Microsoft, Amazon, Nvidia, Tesla and many, many others.

Rise of Low-Cost AI Agents: Innovations like ManusAI (autonomous AI agents from China startup with $10 million of funding) demonstrate that AI is becoming increasingly accessible and affordable. This democratization undermines centralized control or deterrence strategies, as both state and non-state actors can develop powerful AI systems with minimal resources. Also, distillation is a process whereby powerful and resource intensive AIs are made smaller and more efficient.

Potential for Escalation: Attempting to enforce MAIM could escalate tensions without effectively preventing misuse. If one party threatens to sabotage another’s AI, it might provoke an AI arms race or preemptive actions, destabilizing global security rather than securing it. Nuclear weapon anti-proliferation failed. Anti proliferation slowed developments. Israel has been actively sabotaging the Iranian nuclear program. It is effectively used only on emerging potential powers.

Given these limitations, MAIM seems ill-suited to address the realities of modern AI proliferation. A rigid deterrence model may not only fail but could also exacerbate risks in an era where AI is decentralized and ubiquitous.

Alternative Approaches: Democratizing AI
My alternative: a robustly supported open-source AI effort paired with increased compute power for everyone, potentially powered by solar energy and then later modular and mass produced nuclear power. This “AI power to the people” approach mirrors the Second Amendment’s principle of empowering citizens with firearms, but instead equips individuals with advanced AI capabilities.

AI Power will also be minimum financial power. AI will be revenue generating. Humanoid bots that perform profitable work. This will also be universal basic AI revenue generation. Superintelligent AI will also be the means of economic value creation.

Let’s break this down:

Key Components of the Alternative

Open-Source AI Efforts:
Encouraging collaborative, transparent AI development could accelerate innovation and safety research.
Open-source models (e.g., those from communities like Hugging Face or EleutherAI) allow widespread scrutiny, reducing the risk of hidden flaws or malicious designs.
Increasing Compute Power for All:
Providing individuals with access to AI servers or local hardware hundreds to thousands of times more powerful than laptops would enable them to run sophisticated AI models independently.
This distributed computing model reduces reliance on centralized tech giants or governments, fostering resilience and diversity in AI ecosystems.
Solar-Powered Distributed Computing:
Solar power could make this vision sustainable by providing renewable energy to run compute-intensive AI systems.
Individuals or communities could operate local AI nodes, creating a decentralized network that’s harder to control or attack.

Benefits

Innovation and Resilience: Widespread access to AI tools could spur creative solutions to security challenges, as more minds tackle the problem. A decentralized system is also less vulnerable to single-point failures or targeted attacks.
Collective Defense: Like an armed citizenry, a population equipped with powerful AI could collectively deter threats, whether from rogue actors or authoritarian regimes seeking to monopolize AI.
Equity and Empowerment: This approach democratizes a technology that’s often concentrated in the hands of a few, aligning with principles of fairness and individual agency.

Risks and Challenges

Misuse Potential: Greater access increases the risk of AI being weaponized for cyberattacks, disinformation, or other harms by malicious actors.
Regulation Difficulty: A decentralized AI landscape complicates government oversight, potentially leading to a “Wild West” scenario with insufficient safeguards.
Feasibility: Scaling compute power via solar energy involves significant costs (e.g., hardware, solar panel production) and technical expertise, which not everyone possesses. However, if superintelligence is near (2026-2028), then lack of expertise will not be a problem.
Inequality: Those with skills or resources to leverage AI might outpace others, deepening social and economic divides. Again everyone will need superintelligence for comparable skills and resources. Democratized is the means for equalizing resources.

Comparing MAIM and Democratization

Control vs. Distribution: MAIM relies on centralized deterrence and mutual threats, while democratization spreads AI power widely, reducing any single entity’s dominance.
Adaptability: Democratization aligns better with AI’s current trajectory—low-cost, accessible tools like ManusAI—whereas MAIM struggles to keep pace with this reality.
Security Outcome: MAIM risks escalation and instability; democratization risks misuse but could enhance collective resilience if managed well.

Some Mix – A Balanced Path Forward
Rather than fully embracing MAIM or unchecked democratization, a hybrid approach might offer the best of both worlds:

International Cooperation: Establish global AI safety standards, ethical guidelines, and transparency protocols to mitigate risks, regardless of who controls the tech.
Supported Democratization: Pair open-source AI and distributed computing with education and support systems to ensure broad, responsible use.
Security Research: Invest in making AI systems transparent and secure, reducing the likelihood of malfunction or exploitation.
Private Sector Alignment: Encourage companies driving AI innovation to prioritize national security and societal benefits alongside profit.

Conclusion
The National Security AI website highlights AI’s critical role in national security, with MAIM as a deterrence-based solution. However, given AI’s unique nature and the rise of low-cost agents like ManusAI, MAIM appears impractical and potentially destabilizing. I propose an alternative—robust open-source AI and solar-powered distributed computing—offering a forward-thinking “AI for the people” vision. While it promises innovation and resilience, it requires careful management to address misuse, feasibility, and equity concerns. A balanced strategy blending democratization with global governance and safety measures could better secure AI’s future while harnessing its potential for all.

Some Details About Military Topics Glossed Over in Superintelligence Strategy

Taiwan TSMC Chips and Risk of China Invading Taiwan

I find it annoying, silly and wrong when reports like the Superintelligence Strategy talk about how China can just take Taiwan and TSMC. This shows no attempt to look into the situation. It is actually pretty stupid.

I, Brian Wang, have looked extensively at the challenge of taking Taiwan by China. Taiwan is a mountainous island. It literally has 80-120 miles of moat. It is D-Day times 10 to take the island. China has not fought a war for 46 years and that was the 1979 border skirmish with Vietnam. China failed in that action. A few million of the chinese army raced across the border and then got repelled by the battle hardened Vietnamese army. Before the 1950s Korea war.

China can transport about 40,000 troops in the first day for an invasion of Taiwan. IF none of the ships and helicopters and planes are shot down while crossing 100 miles of open water. If 40,000 troops could make it onto Taiwan, the Chinese troops would be outnumbered 4 to 1. I go over every ship and weapon China has and the hours of training of their fighter and bomber pilots.

An attempted China invasion would be worse than Dieppe for China. Why? Satellites. Everything that China tries to do in an attack would be seen by satellites.

Everyone knows that China has had decades of a one child policy. What does this mean? Pretty much every soldier is an only child (although some rural born would not be). what does this mean? Every time Taiwan sinks a troop ship a few hundred to a few thousand only children go to the bottom of the Taiwan Strait. How would thousands and then millions of parents and grandparents take this? I think there is significant risk of an uprising.

Russian S-400 air defenses were defeated with missiles waves in Ukraine. F-16s and Russian planes have been very vulnerable in the Ukraine war. 80% of China’s planes are copies or purchased Russian planes.

Taiwan mountains. For those who do not know, the powerful conventional explosives we have do not penetrate far into rock and dirt. U.S. Marines landed on Iwo Jima after months of intense naval and air bombardment, but the Japanese were well-protected in caves and other stronghold.

North Vietnam army withstood months of even more intensive bombardment by hiding in tunnels.

Taiwan has anti-air missiles. China planes flying close enough to try and hit Taiwan planes and missiles in caves would get wrecked by anti-air missiles as seen in the Ukraine war.

Any China navy and air forces, would also get shot from the Philippine and Japanese islands where the US has bases.

Then we get to massive drone waves. This analysis has to start from the massive drone use by Ukraine and Russia. A close to balanced volume destroys things for sure. But where is the big land grabs from a dominant drone capability?

Handwaving SuperAI and AI Superweapons vs Simple Military Options

The SpaceX Starship will soon be mass produced and the upper stage will cost $20 million or less. This is cheaper than $100 million fighter planes than top at mach 2. This is cheaper than large commercial passenger planes like the 787 that cost $300 million.

The US has 58 B-52 operated by active forces and 18 by reserve forces. The US is spending $48 billion to upgrade them. This is over a half billion each.

A Starship will hold 200 tons of payload and will fly over 20 times the speed of sound. Just releasing metal or ball bearings, means the kinetic energy from something moving at 20 times the speed of sound will be up 5-10 times its mass in chemical explosives.

Release 2 million metal objects flying above some location means everything flying below the Starship gets wrecked. No missile, no plane, no drone would survive. It does not matter what AI was in it.

If the AI was buried in some deep bunker. Flying one $10 million Starship into the bunker places a concentrated 2000 to 20000 tons of kinetic force into that bunker. This is the mass effectively have 10 times its weight converting kinetic energy to chemical explosive energy.

Hitting a bunker with the Booster and the Starship ($40-100 million in cost) guarantees wrecking any bunker. 200 tons at mach 20 would penetrate to 108 meters.

At Mach 20, this is a hypervelocity impact. The spear isn’t just penetrating—it’s interacting with the target at speeds where material strength becomes secondary to energy transfer, shockwaves, and thermal effects.

The tip hits the concrete at ~6,800 m/s. At this speed, both the spear and the concrete behave more like fluids due to the extreme pressures (hydrodynamic regime).

Pressure at impact = ρ × v² (where ρ is density). Assuming a steel spear (ρ ≈ 7,800 kg/m³):
P = 7,800 × (6,800)² ≈ 3.6 × 10¹¹ Pa (~360 GPa). This dwarfs UHPC’s compressive strength (0.2 GPa), instantly pulverizing the concrete.

Destroy the bunker via shockwaves, heat, and structural collapse, with energy equivalent to a small nuclear blast. The kinetic force would be concentrated.

Iran Fordow is 60-80 meters deep.

Mach 20, 2,000 tons: 46.24 TJ (~11 kt) – Deep penetration and destruction, but less overwhelming than mach 30..

Mach 20, 200 tons: 4.624 TJ (~1.1 kt) – Penetration with severe damage, but deeper bunkers might survive partially.

Mach 30, 2,000 tons: 104.04 TJ (~24.9 kt) – Overkill for a 200-foot bunker, capable of annihilating even the deepest fortified sites.