LLMs are trained unintelligently. They’re brute-forced into shape, like cracking a password with infinite tries. It’s ugly and energy-hungry. Yet probably worth it to help uncover the true essence of intelligence. Which is likely to be beautifully simple.