English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Advances in computational methods and big data availability have recentlytranslated into breakthroughs in AI applications. With successes in bottom-upchallenges partially overshadowing shortcomings, the 'human-like' performanceof Large Language Models has raised the question of how linguistic performanceis achieved by algorithms. Given systematic shortcomings in generalizationacross many AI systems, in this work we ask whether linguistic performance isindeed guided by language knowledge in Large Language Models. To this end, weprompt GPT-3 with a grammaticality judgement task and comprehension questionson less frequent constructions that are thus unlikely to form part of LargeLanguage Models' training data. These included grammatical 'illusions',semantic anomalies, complex nested hierarchies and self-embeddings. GPT-3failed for every prompt but one, often offering answers that show a criticallack of understanding even of high-frequency words used in these less frequentgrammatical constructions. The present work sheds light on the boundaries ofthe alleged AI human-like linguistic competence and argues that, far fromhuman-like, the next-word prediction abilities of LLMs may face issues ofrobustness, when pushed beyond training data.

Testing AI performance on less frequent aspects of language reveals insensitivity to underlying meaning