🀯 AI Struggles with Common Sense

– AI chatbots like GPT-3 are impressive but struggle with common sense questions.

– Max Bennett explores human brain evolution to bridge the gap.

– GPT-3's training process involves predicting the next word in a sequence.

– GPT-3 can predict novel sequences, mimicking human language prediction.

– However, human brains do more than predict; they simulate scenarios.

– GPT-3 struggles with common sense questions, lacking understanding.

– Human brains differ by simulating and rendering inner worlds.

– Language learning in humans involves tethering symbols to inner simulations.

– Humans can check answers through mental simulation, unlike GPT-3.

– Human brains combine language prediction and inner simulation, as shown in cognitive reflection tests.