AI Systems Demonstrate Core Cognitive and Linguistic Limitations

Key Intelligence

• ■Recent research indicates that AI chatbots struggle with cognitive tasks requiring attention and inhibition, failing classic tests like the Stroop test.
• ■Large Language Models (LLMs) frequently provide confidently incorrect answers, highlighting issues with reasoning and truthfulness.
• ■AI performance exhibits linguistic biases, particularly with Romance languages, suggesting potential shortcomings in training data or model architecture.
• ■Experts emphasize the critical need for greater explainability in LLMs to understand their decision-making processes and address inherent flaws.