Podcast
The most important question nobody's asking about AI.
- Within 20 years, 99% of the workforce across military, government, and private sectors is projected to be operated by AI, while current LLMs are not yet utilized in mission-critical roles.
- AI is expected to evolve into robot armies for military applications, superhuman advisors for political and corporate leaders, and a replacement for police functions, eventually underpinning civilization through AI labor.
- Supply chain restrictions could force major tech firms like Amazon, NVIDIA, Google, and Palantir to sever ties with Anthropic regarding Pentagon work, potentially leveraging government contracts to enforce such conditions.
- Prediction markets currently assign a 74% probability to the backtracking of supply chain restrictions, though the federal government holds significant leverage over private companies through antitrust enforcement, power permitting, and existing contracts.
- Mass surveillance capabilities are forecasted to be enabled by top models from Cloud 6 and Gemini 5 by 2027, with open-source models matching prior frontier performance by late 2027 or 2028, potentially allowing the government to process camera feeds using open-source tools.
- Technical costs for surveillance are projected to decline by 10x annually; processing 100 million CCTV cameras at a frame rate of one every 10 seconds would cost $30 billion initially, dropping to $3 billion the following year, $300 million the year after, and becoming cheaper than remodeling the White House by 2030.
- Structural intrinsic advantages in AI technology favor mass surveillance and population control, with the government's monopoly on violence potentially augmented by technically aligned, obedient employees.
- While mass surveillance is legally permissible in certain forums involving third-party data (excluding Fourth Amendment protections), it is currently impractical to enforce, though the primary future barrier is expected to be political expectation rather than technical capacity.
- Technical progression is predicted to be multipolar, with competitive companies at every supply chain layer and all entities capable of training frontier-level models within 12 months, preventing any single private company from monopolizing capabilities like robot armies or hacking.
- Anthropic is considered capable of surviving current designations, but future dynamics may see big tech companies prioritizing AI development over government contracts, which represent a tiny fraction of their revenue.
- The diffusion of AI and the integration of technology into all product lifecycles may eventually make it impossible for major tech firms to isolate cloud services from government work, despite current efforts to cordon them off.