AI summary by TwinMind
What Everyone Gets Wrong about AI
Sabine Hossenfelder
8m 36s
Political Misunderstanding of AI
- Politicians have cognitive dissonance about AI, fundamentally misunderstanding what's happening
- Examples of misguided political rhetoric:
- UK PM Keir Starmer: "global race of our lives" - countries will be "AI makers or AI takers"
- EU Commission President Ursula von der Leyen: AI boosts productivity, first movers rewarded, Europe must adopt AI faster
- President Biden: AI reshapes economies/governments/national security, US and allies must lead
Core Misunderstanding
- Politicians think it's about countries leading AI development
- Reality: It's specific companies registered in America controlling frontier models
- OpenAI's GPT, Meta's Llama, other frontier models
- These models aren't property of America - they're property of very rich individuals
- These people developing systems that will be more intelligent than everything else on planet
The Real Race: Power, Not Profit
- Politicians mistake this as race for profit/prosperity
- Actually: race for power and world domination
- Whoever gets superhuman AI first will rule the world
- Company registration location won't matter - they'll have multiple backups to avoid nationalization
- By time governments realize they're not in control, it'll be too late
Futility of Current Government Strategies
- Small startups UK/EU funding have no chance catching up to frontier models
- Most small apps being built using frontier models anyway
- Government money just reinforces power of frontier models
- Few niche applications may be exceptions, but vast majority will use frontier models
- Reasons: A) easier, B) cheaper, C) more user-friendly long-term
Frontier Models as New Operating System
- Technically not operating systems, but practically how we'll use them
- Users will sign up to one AI to do everything on devices
- AI will handle emails, bills, procrastination, etc.
- Governments/companies will become dependent quickly
- Without AI access in 5 years = like giving up internet today
- Access controlled by company owners
Palantir Example
- Biggest AI profit increase in past year (not Google or Meta)
- Company owned by Peter Thiel
- Provides multimodal AI solutions for data integration/management
- Won contract to overhaul British NHS data management system
- Likely involved with intelligence/military organizations
- Name from Lord of the Rings - "seeing stones" to watch other people
Motivations of AI Leaders
- Elon Musk suing OpenAI and building own AI despite having enough money/projects
- Sam Altman (OpenAI CEO) paradox: declares superintelligent AI super dangerous but only way to assess danger is to build it
- Doesn't worry because he'll be the one in power
- "Rest of us will be living in metaverse mining Bitcoin to pay his energy"
Government Response Analysis
- Only China understands this is fight for world dominance
- Europeans and Americans fail to see enormous power concentration coming
- Concentration happening in next few years in just a few companies
Proposed Solution
- If speaker was "queen of Europe": take money earmarked for bigger particle collider (few dozen billion dollars)
- Pour it into publicly owned frontier model "like yesterday"
- Otherwise Europe will be in bigger trouble in 5 years than now
