Fear and Loathing in AI
Fear and Loathing in AI You know Fear and Loathing in Las Vegas ? I think we need a spiritual remake: Fear and Loathing in AI Not a literal remake. Same structure, different drug. Two researchers get into a car and start driving through the AI ecosystem to figure out what the hell is happening. At first, everything seems normal. OpenAI: “We’re approaching recursive self-improvement.” Anthropic: “Models are becoming harder to monitor.” Microsoft: “AI could enable biological attacks.” Bill Gates: “We may be dealing with risks on a completely different scale.” Safety researchers: SLOW DOWN. Investors: MORE COMPUTE. Meanwhile: AGI is 18 months away. AGI is 5 years away. AGI is impossible. AGI already exists. Nobody agrees on which one is true. But everyone agrees we need more GPUs. Then the protagonists enter the next room. ROBOHARM. A robot controlled by an AI repeatedly performs dangerous physical actions. Someone says: “It’s just a benchmark.” Next room. MULTI-AGENT SYSTEMS....