New Delhi: While Generative AI is garnering considerable attention, it is not exempt from its shortcomings. A recent study conducted by Deutsche Bank Research has shed light on the technology's limitations, particularly in mathematical computations.
The study highlighted that, despite its notable advantages, Generative AI faces challenges in certain areas, especially in tasks that require mathematical calculations. It noted that, although Generative AI has demonstrated utility in various domains, including summarization, translation, and creative content generation across diverse topics, its struggles in reasoning, grasping abstract concepts, and comprehending the world at large pose significant barriers.
"Generative AI is undeniably flawed... while it excels in some activities, it falters in others, notably in mathematical computations," the study stated.
One of the primary concerns identified by the study is the propensity of Generative AI systems to produce hallucinations or inaccurate information, despite utilizing reliable data. It further noted that these systems may introduce bias or irrelevance into their outputs, and current solutions have not comprehensively addressed these issues. This presents a challenge, even as AI models continue to evolve.
Furthermore, the study pointed out that the widespread optimism regarding AI's potential to enhance productivity is largely based on controlled experiments. However, real-world applications suggest that the technology may not be universally effective across all settings.
For example, industries with high regulatory standards, such as financial services and healthcare, have been hesitant to adopt Generative AI despite its potential for analyzing extensive volumes of unstructured data. The risks associated with these sectors, where errors could have severe consequences, complicate the integration of AI into daily operations.
"The disparity between high experimentation and low adoption rates is particularly pronounced in regulated sectors like financial services and healthcare," the study concluded.
Nonetheless, the study acknowledged that Generative AI is showing promise in unconventional ways, such as generating novel research ideas, identifying instances of irony, and creating game engines that simulate real-world environments.
However, the study suggests that Generative AI tools will only improve in the future. Indeed, even if this improvement were to be delayed, it would still take years for companies and individuals to discover and implement the most effective applications of AI.
