AI Good at some tasks, but bad with Maths says Deutsche Bank research

Generative AI, while gaining widespread attention, is not without its flaws. A recent report by Deutsche Bank Research highlighted that despite its strengths, the technology still struggles with certain tasks, particularly when it comes to mathematical calculations.

The report noted that though generative AI has proven useful in many areas, such as summarizing, translating, and even generating creative content on a wide range of topics, its limitations in reasoning, learning abstract concepts, and developing an understanding of the world remain significant obstacles.

“Generative AI is certainly flawed… while Generative AI is surprisingly good at some activities, it is surprisingly bad at others, such as making mathematical calculations” said the report.

As per the report one of the key issues is the tendency of generative AI systems to produce hallucinations, or inaccurate information, despite using reliable data. It noted that these systems can also introduce bias or irrelevance into their outputs, and existing solutions have not fully addressed these problems. This remains a challenge, even as AI models continue to evolve.

The report also pointed out that much of the optimism surrounding AI’s potential to boost productivity comes from controlled experiments. However, real-world applications show that the technology may not be as effective in every setting.

For instance, highly regulated industries like financial services and healthcare, where the stakes are especially high, have been slow to adopt generative AI despite the potential benefits of analyzing vast amounts of unstructured data.

The risks in these sectors–where errors could lead to serious consequences–make it harder for them to integrate AI into everyday use.

“The gap between high experimentation and low uptake is particularly striking in regulated industries such as financial services and healthcare” the report added.

In some cases, generative AI is showing potential in unexpected ways, such as generating novel research ideas, identifying irony, and even creating game engines that simulate real-world environments.

However, as per report the generative AI tools will only get better from here. And indeed even if they never did, it would take years for companies and individuals to find and implement the best use cases of AI.