This 150-Year-Old Puzzle Just Exposed AI's Biggest Lie
This 150-year-old puzzle just exposed that AI can't actually think. Apple researchers gave leading AI models the Tower of Hanoi puzzle - just move disks from one peg to another following simple rules. Even first-year computer science students solve this easily.
But here's what happened: With just 2-3 disks, AI looked smart. But add more disks? Total collapse. Google's Gemini scored only 24%, OpenAI's latest model got 2%. They made basic errors, contradicted themselves, and confidently gave wrong answers.
Turns out, AI isn't reasoning at all - it's just pattern matching from training data. When the pattern breaks, so does the 'intelligence'. We've been calling it reasoning, but it's really just very expensive autocomplete.
#shorts #ai #technology #apple #research #google #openai #chatgpt #gemini #puzzles #reasoning #machinelearning
This 150-year-old puzzle just exposed that AI can’t actually think. Apple researchers gave leading AI models the Tower of Hanoi puzzle – just move disks from one peg to another following simple rules. Even first-year computer science students solve this easily.
But here’s what happened: With just 2-3 disks, AI looked smart. But add more disks? Total collapse. Google’s Gemini scored only 24%, OpenAI’s latest model got 2%. They made basic errors, contradicted themselves, and confidently gave wrong answers.
Turns out, AI isn’t reasoning at all – it’s just pattern matching from training data. When the pattern breaks, so does the ‘intelligence’. We’ve been calling it reasoning, but it’s really just very expensive autocomplete.
#shorts #ai #technology #apple #research #google #openai #chatgpt #gemini #puzzles #reasoning #machinelearning
