Old AI models are losing their minds— Literally

Scientists gave AI chatbots a real human dementia test (the MoCA), and guess what? The older ones basically drooled on the exam. ChatGPT-4 barely passed with a 26/30, but Gemini 1.0 scored a tragic 16—worse than your sleep-deprived grandpa.

Turns out, AI struggles with visual/spatial tasks, executive function, and, uh… remembering things. So if an AI tells you it’s “99% sure” you have the plague, maybe get a second opinion—from an actual doctor.