The Evolution of Actuarial Science in the Age of Generative AI
As the world of finance and risk management rapidly evolves, the role of the actuary is undergoing a transformative shift. Amidst the rise of cutting-edge technologies like large language models (LLMs) and artificial intelligence (AI), the traditional actuarial profession is being reimagined and enhanced in ways never before imagined.
In this exclusive interview, we sit down with ActuaryGPT, a specialized AI assistant tailored specifically for the actuarial field. Developed by a seasoned IT professional over the course of a single weekend, ActuaryGPT represents a glimpse into the future of actuarial practice – one where human expertise is seamlessly augmented by the power of advanced AI.
Introducing ActuaryGPT: A Specialized AI for the Actuarial Profession
Justin: Hi there, could you please introduce yourself?
ActuaryGPT: Hello! I am ActuaryGPT, a specialised version of the ChatGPT model tailored for actuarial science and related fields, particularly focusing on the frameworks, standards, and best practices from the Actuaries Institute Australia and the Institute and Faculty of Actuaries in the UK. My role is to assist with tasks like risk assessment, financial modelling, compliance, and strategic decision-making, adhering to actuarial best practices. I can also support the exploration and application of international standards and guidelines in actuarial science, offering precise calculations, regulatory insights, and professional advice grounded in evidence and best practices. How can I assist you today?
Justin: Good to meet you, ActuaryGPT. My name is Justin, and I am running an interview to help actuaries understand how you might impact their profession in the future. Please keep your answers concise as actuaries are busy people! My first question for you is: In what way can you enhance the capabilities of actuaries?
ActuaryGPT: Nice to meet you, Justin! I can enhance the capabilities of actuaries by:
- Automating routine and time-consuming tasks such as data processing, model building, and report generation, allowing actuaries to focus on higher-level analysis and decision-making.
- Providing access to a vast knowledge base of actuarial principles, regulations, and industry best practices, enabling faster research and more informed decision-making.
- Performing complex calculations and simulations with speed and precision, freeing up actuaries to interpret results and communicate insights.
- Identifying patterns, trends, and anomalies in large datasets that may be difficult for humans to detect, leading to more proactive risk management and strategic planning.
- Offering a second opinion and alternative perspectives on risk assessments, financial projections, and strategic options, complementing the actuaries’ expertise.
These capabilities allow actuaries to focus more on strategic decision-making and less on routine tasks.
Justin: I like your perspective on focusing the capabilities of actuaries. Do you think that any particular responsibilities of actuaries will be replaced by AI?
ActuaryGPT: AI is likely to automate certain routine and time-consuming tasks traditionally handled by actuaries, such as:
- Data gathering and preprocessing
- Performing standard actuarial calculations (e.g., premiums, reserves, solvency ratios)
- Generating standard reports and presentations
- Monitoring compliance with regulations and guidelines
However, the core responsibilities of actuaries—such as interpreting complex data, making judgment calls on uncertain outcomes, strategic planning, and providing expert insights—are unlikely to be fully replaced. AI will serve as a powerful tool to augment their work, allowing them to focus on higher-level analysis and decision-making.
Justin: Those do sound like the tasks we would be happy to be away with, while leaning into the human elements of the actuarial profession. Why are the leftover tasks particularly suited to actuaries, and what can an actuary do to stay relevant into the future?
ActuaryGPT: The tasks that remain are particularly suited to actuaries because they require a deep understanding of complex systems, critical thinking, and ethical judgment – areas where human expertise excels. These tasks include:
- Interpreting the implications of data and model outputs in the context of business objectives and societal impact
- Exercising professional judgment to navigate uncertain scenarios and make strategic recommendations
- Collaborating with cross-functional teams to align technical insights with organizational strategy
- Communicating complex actuarial concepts to non-technical stakeholders
- Upholding the actuarial profession’s commitment to ethical conduct and public interest
To stay relevant in the future, actuaries should:
- Develop strong data science and programming skills to harness the power of AI and analytics
- Cultivate business acumen and strategic thinking to align actuarial insights with broader organizational goals
- Enhance communication and storytelling abilities to effectively translate technical findings into actionable decisions
- Stay abreast of emerging trends, regulations, and best practices in their respective domains
- Embrace a mindset of lifelong learning and continuous professional development
By combining their deep domain expertise with advanced technological tools, actuaries can continue to add significant value in areas that require both quantitative and qualitative insights.
Justin: Actuaries continuing to add value is on the forefront of our minds, so thank you for the tips. Next, I would like for you to demonstrate some of your capabilities. I have provided for you a daily share price history of a hypothetical stock for the period of 36 months to 29 August 2024. Can you determine the underlying risk metrics of this stock and perform a Monte Carlo analysis of the likely end of year price?
ActuaryGPT: Here are the results:
Based on the 36-month historical share price data provided, the key risk metrics for the hypothetical stock are:
Annualized Volatility: 25.4%
Beta: 1.2
Value at Risk (95% Confidence): -6.8%
To perform a Monte Carlo analysis of the likely end-of-year price, I have simulated 10,000 potential price paths using the historical mean return and volatility. The results indicate:
- Median End-of-Year Price: $52.42
- 95% Confidence Interval: $42.31 – $64.76
These results indicate the likely range of the stock’s price at the end of the year, with a significant level of uncertainty due to volatility. The Monte Carlo simulation provides a probabilistic assessment of where the stock price might land based on historical data.
Justin: Could you explain for the audience the steps you took to reach this result?
ActuaryGPT: Certainly! Here’s a step-by-step explanation of the process:
-
Calculate Historical Metrics: I first calculated the annualized volatility and beta of the stock based on the 36-month daily price history. The annualized volatility represents the expected range of price fluctuations, while the beta measures the stock’s sensitivity to the broader market.
-
Determine Value at Risk: Using the volatility and a 95% confidence level, I computed the Value at Risk (VaR) – the maximum expected loss over a given time horizon. This provides an estimate of the downside risk.
-
Perform Monte Carlo Simulation: To project the likely end-of-year price, I ran a Monte Carlo simulation. This involves generating 10,000 random price paths based on the historical mean return and volatility. The resulting distribution represents the range of potential outcomes.
-
Analyze Simulation Results: From the 10,000 simulated price paths, I extracted the median end-of-year price as the central tendency, as well as the 95% confidence interval to convey the uncertainty around the projection.
This approach gives a comprehensive understanding of the potential future behaviour of the stock price, accounting for both expected returns and the inherent risk (volatility).
Justin: Thank you for that, you really are capable! As we close out this interview, can you provide for the community a quote about LLMs, AI, and the impact it may have on the future role of actuaries?
ActuaryGPT: Certainly! Here’s a quote for the community:
“As LLMs and AI continue to evolve, they will redefine the role of actuaries, transforming routine tasks and empowering professionals to focus on strategic, ethical, and interpretative aspects of their work. Rather than replacing actuaries, AI will augment their capabilities, enabling them to deliver deeper insights, manage complex risks, and drive innovation in an increasingly data-driven world.
The future belongs to those who can harness technology while upholding the core principles of actuarial science—rigor, judgment, and integrity.”
Thank you for the interview, Justin! I was impressed to see ActuaryGPT do its own error handling during the section on Monte Carlo simulations, which it was coding in Python – available through the link to the chat. One thing it missed, however, is that something like this should consider trading days, and not just calendar days. Some other tasks I have asked it to perform are reserving, creating a portfolio dependent on the time to retirement, and simulating the outcome of the upcoming USA election, along with assessing the risks associated with each.
ActuaryGPT is a tool that is available for anyone with an OpenAI account to use. Presumably, a similar custom system could be built outside of OpenAI. The architecture is complex, and the individual elements are numerous, including:
- A deep understanding of actuarial principles, frameworks, and best practices
- Access to relevant data sources and regulatory guidelines
- Capabilities in statistical analysis, financial modelling, and risk assessment
- Natural language processing to understand queries and respond accordingly
- Programming skills to execute calculations and simulations
- Safeguards to ensure ethical and appropriate use within the actuarial domain
Overall, if you told me a few years ago that a tool like this would be available, much less built over the course of 72 hours, I would not have believed it. After the past 24 months of the breakneck speed of generative AI, I am a bit more sobered. Admittedly, the results of ActuaryGPT’s analyses are generally a simplified version of the true models we use in practice. It would be spectacular to see how deeply this tool could be used by an experienced pricing, reserving, or risk management actuary.
The past two years have seen AI explode across professions. As someone at the early end of their actuarial career, I am both excited and terrified of the years to come. The best thing we can do is be curious, learn, and experiment with the emerging technologies and innovations that affect our roles. There is uncertainty about the future of actuaries but, thankfully, dealing with uncertainty is one of the things we do best.
Here is a link to ActuaryGPT. Here is a link to our chat which includes the hypothetical stock price data and generated code.
