Will AI tools like GPT-4 make actuaries obsolete?

Over the past few weeks, I've spoken with lots of actuarial analysts who are pretty anxious about AI. They've seen the increasingly-impressive results from large language models (LLMs) like GPT-4 and they're worried that there won't be any jobs left by the time they qualify as an actuary.

I think these fears are overblown and, in this blog, I'm going to share why I think actuaries aren't going anywhere. Things are going to change, but not in the catastrophic way people are saying online.

Groundhog Day.

Actuaries have always evolved in response to new tech. From the days of handwritten calculations, to slide rules, calculators, spreadsheets, cloud computing and now LLMs.

Of course, not every actuary has evolved… At the start of my career, you could split any group of actuaries into two groups - those who were comfortable using Excel/VBA and those who avoided spreadsheets. Right now, the split is those who can code (in R/Python) and those who can't. In the next few years, I think the most important split will become those who can leverage LLMs and those who can't.

The best carpenters adopted power tools. The best photographers adopted digital cameras. And the best actuaries need to adopt LLMs, or risk being left behind.

Don't Learn Tools. Learn Skills.

There are a lot of new AI tools popping up that promise to make your life as an actuary easier ("become more productive", "study more effectively", "boost your memory" etc. etc.) I try to avoid 99% of them and instead focus on using LLMs directly.

Why? Most of these tools are just a fancy front-end (they all use ChatGPT or GPT4 under the hood) and I think many of them will disappear within the next 6 months. Most importantly, they take away your capacity to learn two critical skills:

  • Prompt writing
  • Interpreting LLM output with a critical eye

I believe the most successful actuaries will ignore the hype around AI tools and focus exclusively on developing these skills.

Working alongside LLMs.

LLMs are still fairly limited at the moment and they regularly make stuff up (known as hallucinations). Only a few weeks ago, ChatGPT was terrible at calculations and it couldn't give up-to-date answers.

But new plugin functionality means it can connect to Wolfram Alpha and browse the web. We're only at the very beginning of what LLMs will be able to achieve.

As an actuary, I'm excited about the potential impact of LLMs on our work. Hopefully they can eliminate the boring, repetitive parts of the job and free up our time to focus on the interesting areas where human actuaries can add value.

Solving the Productivity Puzzle.

I've thought about whether the increased productivity from LLMs will reduce the total number of actuarial jobs. If each actuary becomes much more efficient, do we need fewer of them? I think the answer is no.

Firstly, even the most advanced LLMs can't replace the years of experience, deep industry knowledge, professionalism and communication skills required to be a good actuary.

Secondly, actuaries are in the business of quantifying risk. New risks will emerge, but I don't think the need to quantify risk will ever go away. In fact, I think LLMs will increase the number of actuarial employers by reducing barriers to entry in the insurance industry.

We're Not the Only Ones Having This Conversation.

Almost every white collar profession is concerned about the impact of LLMs. No doubt there are lawyers, accountants, architects and actuaries having an existential crisis right now. What's the solution? Keep calm and carry on.

There's no higher ground to escape tech improvements. Instead of worrying about the future, focus on becoming an adaptable, lifelong learner. Then whatever happens, you'll be able to cope.

I strongly believe that the core skills needed to be a successful actuary haven't changed because of LLMs. In fact, I think it's an exciting time to be an actuary!