Google’s generative AI models, Gemini 1.5 Pro and 1.5 Flash, are claimed to be able to handle and analyze large volumes of data with long context capabilities. However, recent research has shown that these models may not live up to expectations. Studies have found that the models struggle to understand content and answer questions correctly when processing datasets as long as “War and Peace.” Despite Google showcasing the models’ capabilities, they have been found to have limitations in evaluating truthfulness in fiction books and video reasoning tasks.
Researchers have raised concerns about the limitations of generative AI, with CEOs expressing skepticism about its productivity advantages due to mistakes and data security concerns. Workers from top AI firms have also criticized the financial incentives that inhibit supervision of AI systems, warning about the potential risks of unmanaged AI spreading disinformation and worsening inequities.
Despite these concerns, advances in AI are expected to transform industries and technology, with AI-enhanced software sales projected to reach $280 billion by 2032. Investors are increasingly focusing on the AI software sector for its potential to streamline processes, boost productivity, and lower costs. However, there are calls for AI businesses to be more transparent about the capabilities and limitations of their systems and to allow employees to raise risk-related issues without hindrance.
Source link
Source link: https://www.techtimes.com/amp/articles/306184/20240630/googles-gemini-ai-struggles-large-data-sets-studies-reveal.htm