Stuffsearth.com — According to Bernstein, Generative AI (GenAI) is emerging as a crucial tool in asset management, offering a “technical scaffold” that enables non-technical professionals to interact with models and derive deeper insights.
Speaking in the Bernstein “Quant Speaker Series,” the firm said a portfolio manager from GIC, highlighted that large language models (LLMs) are bridging the gap for those who previously found traditional machine learning challenging due to high technical barriers.
Bernstein notes that LLMs excel in processing unstructured data, such as text and natural language, where creativity and context are vital.
They explain that unlike traditional models suited for more formulaic tasks, GenAI thrives in handling variability and adaptability.
Furthermore, LLMs are said to be useful in summarizing documents, extracting key features, and even translating images into text, offering a valuable way to “separate signal from noise.”
However, challenges remain. Bernstein emphasized that GenAI models are still evolving and come with high fixed costs, making them more beneficial for larger firms with economies of scale.
They also believe smaller firms might struggle with broad-based adoption due to the steep investment required. Furthermore, organizational readiness is crucial, including having a solid data infrastructure and alignment of legal and compliance functions.
Bernstein also addressed the role of humans, noting that as AI lowers the cost of cognition, humans must adapt by either becoming specialists or understanding how to integrate GenAI tools into their work.
Despite its promise, Bernstein says GenAI still requires human oversight due to its limitations, including forward-looking bias and inconsistent outputs, which means that, for now, humans remain integral to the decision-making loop in asset management.
Overall, while GenAI holds significant potential for asset management, particularly in feature extraction and summarization, Bernstein feels its full adoption will depend on overcoming organizational and technical challenges.
Reference :
Reference link
