But let’s face it, understanding AI basics can still be a bit daunting. So, let’s simplify a bit: think of generative AI, especially Large Language Models (LLMs), as a supercharged version of your iPhone’s predictive text feature. It predicts text based on vast datasets, much like how your phone suggests the next word in a text message. This analogy helps to illustrate that we’re already using basic forms of AI in our daily lives—just on a smaller scale.
Best practices: getting up to speed with AI
To help you get the most out of GenAI, here’s a rundown of some core industry best practices:
1. Control the source of data: To avoid AI hallucinations (when the AI goes off the rails with incorrect or nonsensical answers), be specific about where the AI pulls its information from. Precise prompting can ensure the AI taps into reliable and relevant datasets. For instance, if you’re working on a financial report, you might instruct the AI to draw from recent market analyses or verified financial statements rather than general web searches.
2. Temperature control: Guide the AI’s responses by adjusting the “temperature” of your prompts. This allows you to manage how creative or specific the AI’s responses are, keeping them within your desired context. A lower temperature setting makes the AI’s responses more focused and deterministic, while a higher setting allows for more creativity and varied outputs.
3. Ensure your AI platform uses Retrieval-Augmented Generation (RAG): Information retrieval is enhanced by using RAG, ensuring the AI accesses the most accurate and relevant data, so this is an important consideration when choosing an AI solution. RAG allows the AI to pull from specific datasets that are continually updated, ensuring the information is both current and precise. At Moody’s, this technique has been crucial in reducing hallucinations and providing traceable, reliable outputs.
Real-world impact and future prospects
Moody’s journey into AI began with products such as QuiqSpread, which used machine learning to extract data from financial reports. Since then, advanced GenAI tools have been introduced that integrate structured and unstructured data and improve risk assessments and financial analysis.
One of the most important innovations is the Moody’s Research Assistant, which was introduced in collaboration with Microsoft’s secure Azure environment. This tool uses RAG to ensure that answers are based on solid data and minimize the risk of hallucinations. By combining structured financial data with unstructured data from various sources, more comprehensive and accurate analyses could be produced.
The practical applications of these tools are vast. For instance, financial analysts can now generate detailed credit risk assessments in a fraction of the time it used to take. By automating data retrieval and initial analysis, these tools free up analysts to focus on more complex and nuanced aspects of their work, ultimately leading to better decision-making.
Looking ahead, mastering basic prompting skills is essential as we gear up for more sophisticated AI applications. This includes creating comprehensive reports, generating polished presentations, and developing custom analyses tailored to specific needs. Moody’s is also exploring AI integration across various platforms, including tools for portfolio monitoring and custom alerts, further enhancing AI’s utility in finance.
The evolution of AI in finance
Artificial intelligence has actually been part of the financial industry for quite some time. Early applications included optical character recognition (OCR) for digitizing paper documents and basic machine learning algorithms for financial forecasting. However, the pace of AI evolution has accelerated dramatically in the last decade, with GenAI representing the latest leap forward.
The fundamental difference between earlier AI applications and GenAI lies in the ability to generate human-like text based on context and probability. Traditional AI could process and analyze data, but GenAI can create new content, interpret context, and provide insights in a conversational manner. This opens up new possibilities for automating and enhancing various processes across finance as well as a slew of other industries, like marketing, content creation, and business, among others.
Of course, as with any transformative technology, the actual integration of GenAI into how we work comes with its challenges. Two major concerns often raised are data privacy and the potential for AI to generate incorrect or nonsensical information, known as hallucinations.
At Moody’s, these challenges were met head on. To ensure data privacy, a partnership was formed with Microsoft to create a secure environment for AI tools. By using Microsoft’s Azure infrastructure, it was ensured that all data used by AI models is protected and remains confidential.
Strict data management practices should be applied to minimize the risk of misreporting. By using RAG, it can be ensured that the AI only accesses verified and relevant data, reducing the likelihood of false results. In addition, data sets are continuously updated to ensure that the information used by the AI is current and accurate.
It is important for Moody’s that users understand that while GenAI is incredibly powerful, it cannot replace human expertise. Instead, it provides us with tools to augment human capabilities. For example, an analyst could use GenAI to create a first draft of a financial report. The AI can retrieve data, create charts and summarize the key points. The analyst then reviews and refines the report, adding insights and interpretations that only a human can provide. This collaboration between AI and human expertise results in more accurate reports that are produced in less time, without compromising on quality.
The future of GenAI in finance
You’ve heard it before, but it bears repeating that the potential applications of GenAI in finance are many and continually evolving. People have only just begun to scratch the surface of what is possible. Future developments may include more sophisticated AI-driven risk assessment tools, enhanced customer service applications, and even more integrated AI systems that can handle complex financial modeling and scenario analysis. From automating routine tasks to enabling more sophisticated analyses, GenAI is poised to become an indispensable ally in our professional toolkit.
Source: Moody´s, 26 June 2024.
