What should we believe about the reasoning abilities of today’s large language models? As the headlines above illustrate, there’s a debate raging over whether these enormous pre-trained neural networks have achieved humanlike reasoning abilities, or whether their skills are in fact “a mirage.”
