Google Gemini falls short in analysis despite high expectations. #AI

Generative AI, which is known for its ability to analyze written content, may not be as effective as previously thought, according to two recent studies. One study found that Gen AI struggles with understanding long-form books, while another showed difficulties in answering questions about videos. This raises concerns for companies looking to integrate Gen AI into their workforce.

The rise of Gen AI has impacted freelance work for copywriters and other gig workers, leading to a decrease in opportunities. Despite the initial hype surrounding AI, these studies reveal limitations in the ability of AI language models to comprehend and build upon extended narratives and analyze large amounts of data.

For example, Google’s Gemini generative AI models, which are designed to process vast amounts of information, have shown shortcomings in understanding long-form written content. The studies also highlighted challenges in visual language models’ performance, particularly in tasks involving long visual contexts.

Researchers found that current state-of-the-art VLMs struggle with ignoring irrelevant information in long visual contexts, impacting tasks like mathematical reasoning and visual question answering. Gemini Flash, one of the AI models tested, had difficulty transcribing handwritten digits from images accurately.

Overall, these studies shed light on the limitations of Gen AI in understanding and processing complex written and visual content, urging caution for companies relying on these models for tasks involving extensive data analysis and comprehension.

Source link

Source link: https://www.techspot.com/news/103610-google-gemini-not-good-analysis-hype-other-ai.html