"Superintelligence Is 'Completely Stupid' – Eli the Computer Guy
Summary
Will Gaia and Eli the Computer Guy discuss the Trump administration’s rebranding of AI to "superintelligence," arguing it is purely political branding with no technical substance. They critique the lack of meaningful benchmarks for new models like Google’s Argon, noting that synthetic tests often fail to reflect real-world utility. The conversation also covers the practical value of AI in optimizing data centers, the challenges of independent security scrutiny, and the rising costs and pricing strategies of major AI providers.
Topics discussed
Sponsors: Zen Nicotine Pouches and Perk
Intro: The Hollywood presidency and AI hype
Trump rebrands AI to 'Super Intelligence'
Branding vs. Reality: What changes with the new name?
Defining Super Intelligence and the confusion of terms
The .si domain rush and government procurement implications
What would a system need to do to be 'super intelligent'?
The AI Czar: Should the role be held by an engineer?
AI security incidents and the need for trusted oversight
Self-regulation, voluntary accords, and corporate incentives
Sponsors: Zen, Perk, and London Business School
The AI race with China: Hype or reality?
The 120-day task force report and government AI strategy
Gemini 4 Argon: Benchmarks vs. real-world performance
Google's 300TB memory savings and practical AI uses
Cyber defenders first: Caution or scrutiny avoidance?
Sponsors: Zen, Perk, and London Business School
Testing new AI models: How long should it take?
Pricing strategies and the sustainability of AI costs
Measuring success: The difficulty of general intelligence
Outro and final sponsor reads
Listen ad-free on Castria