Chomsky's AI Linguistics Dilemma: Debunked by 'Impossible Languages'
A groundbreaking study challenges Noam Chomsky's stance on AI language models, revealing that these models struggle more with 'impossible languages' than natural ones. The research involved creating twelve artificial languages by altering English datasets with unique rules. This study negates Chomsky's assertion that models can learn both real and artificial languages equally well. A key finding highlights 'information locality' as essential for language comprehension in both AI and humans.
Jan 14