• Home
  • |
  • Blog
  • |
  • AI for Actuaries: What Do You Need To Know?
artificial intelligence

The Dawn of AI

When we think of the term “AI”, what do we initially perceive? 

Does our mind instantly recall the chilling dystopian scenes of science fiction movies like iRobot, do we gasp at the thought of eventual human annihilation brought on by the Singularity in a world where superintelligence reigns supreme?

Though these are extreme outcomes of the arrival of AI and its integration into our daily lives, it is highly probable that we are, at least, apprehensive. That’s human nature, to fear the unknown.

But what if what we really fear is that which we cannot understand? As a species, we are in constant pursuit of the Theory of Everything, the idea that we will someday be able to explain everything in the Universe. However, many argue that the nature of our existence means that there will always be some phenomena that we will not be able to understand, so at some point will reach our limit of knowledge.

If the amount of new knowledge we can acquire eventually succumbs to exponential decline, we may be seeing this trend start to take on an empirical form. Studies show that trends in human IQ are potentially reversing, suggesting that we may even have already passed the summit of human potential. If we are indeed cognitively limited, AI is key to our continued advancement. We require this revolutionary extension to human potential to stay relevant in our continually evolving universe.

As actuaries, we have a unique appreciation of the complexity of our world. We strive to explain that which others cannot, and though the accuracy of our predictions may be correlated to our success and relevance, negating to recognise the fact that our models are ultimately wrong is a depressor on our advancement. In fact, in a world that is becoming increasing chaotic, our actuarial models may become increasing less relevant. We need a more advanced, augmented approach to making inferences from progressively noisy data. We need a revolution within our profession. Enter AI.

AI is a term that blankets a wide range of subsets, but in short it defines any form of human-like intelligence exhibited by a computer, robot or other machine. In the modern, data-abundant world, where we are seeing rapid development of extremely fast and accurate processing systems and constantly striving for an error free, Utopian existence, the emergence of AI was facilitated and, arguably, inevitable. Purposed with the simplification and lessening of human burden leading to better decision making, improved business efficiency and progression, AI has a plethora of applications ranging from smart client marketing to drug discovery.

A key element of AI is Machine Learning (ML), which processes data using algorithms to detect patterns and make predictions. Today, AI is currently limited to what it knows, referred to as “narrow intelligence” or “weak AI”. However, the development of more complex forms of AI such as Deep Learning, where the machine teaches itself to perform more accurate analyses via deep neural networks, and wider applications of AI, are being heavily invested in.

actuaries

AI for Actuaries

We can draw close ties between the work of AI and our professional endeavours as actuaries. The closer our models are to reality , the better our predictions. AI has the potential to make our models smarter as it can cope with greater amounts of unfiltered data than we have the ability to process, faster and more efficiently. As the stockpile of consumer data available is growing for actuaries, we need models that are constantly learning and adapting to new, and potentially noisier, information.

Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think we'll augment our intelligence.”

— Ginni Rometty

In the insurance industry, for example, huge amounts of accessible data and a greater availability of interconnected consumer devices means a much deeper understanding of clients and market insight is within reach through the implementation of AI. This has the potential to vastly reduce claim volume, improve pricing and attract more of the right clients, and correspondingly the right risks, to the firm.

Actuarial firms also expend significantly on administration and client servicing. Automation of basic communication, number crunching and report generating activities could provide actuaries with an opportunity to align themselves closer to the more human element of the profession, and the one less replaceable in the future. More emphasis and importance could be placed on high value activities whilst the automation of computational or administrative elements only requires actuaries’ involvement in end decision making and implementation of AI outputs.

The possible applications of AI in the actuarial field are so extensive that even some IT entrepreneurs without any prior actuarial expertise have launched start-ups in data-rich industries such as insurance.

Lemonade, a New York based start- up and provider of renter, homeowner and pet insurance policies has been so efficient in implementing AI that it holds the world record for claim processing in under 3 seconds. This is attributable to its extremely advanced chat bots which process reported claims, issue payments, help with new client onboarding and much more. Cytora, which helps improve commercial insurance underwriting, uses AI to churn through large volumes of property data inputs to allocate risk scores on an address-to-address basis.

In the pensions industry, AI can be implemented by actuaries to predict retirement scenarios and timing. The Finnish Centre for Pensions successfully uses ML to identify those who will retire on a disability pension based on predictor input data. AI also has applications as a user interface, dealing with member queries and providing clarity regarding members accrued benefits.

actuaries

Australian based Mercer offers clients its “Superbot” facility, a chatbot they can access via Facebook Messenger, which projects members' benefits to retirement and converts them to an achievable retirement income.

Relative to the pensions world is that of investment management, where AI is tasked with pulling investment data together and using it to detect patterns, looking for areas of improvement and measuring investment performance. Denmark based Grandhood have created a 100% digitised workplace pension for entrepreneurs and SMEs. They follow an algorithmic asset allocation which is based on low-cost ETFs and use risk profiling to match each individual to an optimal investment portfolio.

We can clearly see a trend in AI’s current applications in the actuarial field. Where repetitive, basic or overly computational activities hinder efficiency, AI can automate such processes. Where Big Data is a stumbling block in model building and inference drawing, or where it is hard to spot a pattern, ML can sort through the noise to find the signal more quickly and accurately than we can.

Future Actuaries using AI

Though these current applications for actuaries appear to be extensive, there is still much potential for the future. In insurance, we are seeing a shift from a “detect and repair” strategy, to one of “predict and prevent”. Even more advanced algorithms will further enhance the ability to detect who exactly is likely to claim and what events are likely to occur and when.

Applications are also possible in damage diagnostics, where technology is used to assess the extent of a claim as well as in automatic claim reporting. More advanced AI will also be able to adapt instantaneously to the individual client and their changing behaviour. In the pensions space, further development of chatbots through voice activation, greater language ability and better disability access will increase client experience and boost customer satisfaction.

Improvement in member dashboards could provide members with automatic scenario analysis to answer questions such as “What would my transfer value be if I decided to quit the corporate life and start a craft beer brewery?” or provide answers to distraught wives wondering what their pension sharing order would be if their husband spends one more day out on the golf course. More accurate modelling could also improve the fields of investment and risk management through better predictions and faster computational power offered by quantum chips.

AI has the ability to vastly advance our profession. By enabling new product discovery and upgraded engagement techniques, actuaries can increase and retain our client base and drive firm profitability. We can facilitate a more personalised delivery of services, better tailored to our clients and our business strategy. Automation of repetitive, replaceable activity can free up vital time to communicate optimised results, make better decisions and drive stronger growth.

However, would this really be an actuarial article without mention of risk? As actuaries we know too well how important a true understanding and analysis of the risks is, and the prospect of full implementation of AI into our lives should not be exempt from scrutiny. When it comes to AI, an advisable attitude to take is one which is open-minded to opportunity, yet is risk aware. Have we truly adopted a risk mindful attitude towards the implementation of these Fourth Industrial Revolution technologies? Have we seriously considered the risks they pose to our profession, and to the world? Have we drawn plans to reduce such risks and prepared ourselves for the unexpected?

It doesn’t seem so. Perhaps, following our depressing initial pondering of what an AI optimised world might look like, we’ve self- rationalised. We’ve put the Singularity down as a fictional eventuality, scored it out as a made- up concept for sci-fi entertainment purposes. Robot wars? Human extinction? We don’t think so. Yet the risks AI poses are very real and, for the most part, very preventable, if we as actuaries place enough importance on carefully considering them.

AI & Accountability for Actuaries

As we have discovered,  future actuaries may have a more important role post calculation of the model output. But who is to confirm that the enhanced model is actually more accurate? The risk of over reliance on the AI output could have numerous implications, for the profitability of the firm and for our professional reputation. We may become over trusting of our new AI capabilities, and may, in reluctance to admit our computational inferiority, not account for the fact that we might not fully understand the output. We may also have little to no comprehension of the biases that may exist in the functions AI performs, as smaller biases may be perpetuated if they go unnoticed.

Accountability for the results produces is also very unclear. To whom is responsibility attributed when things go badly wrong, if, for example, the chatbot exhibits racist tendencies as Microsoft’s Tay did in 2016? It was built using supposedly cleaned public data, so can we really blame Microsoft for a machine that taught itself such mannerisms?

The development of full artificial intelligence could spell the end of the human race….It would take off on its own, and re-design itself at an ever increasing rate. Humans, who are limited by slow biological evolution, couldn't compete, and would be superseded.”

— Stephen Hawking to the BBC

Another risk to the actuarial profession is that of over- automation. Perhaps AI will become so advanced and capable of making its own inferences that the relevance of our profession will eventually decline. Or it will become so complex that only data-savvy tech wizards will be capable of working alongside it.

As actuaries, it is part of our professional responsibility to serve the public interest, not just within the realm of our profession. Though there are risks to our professional roles, the risks to the human population are far more extensive. Without adequate controls, personal data is increasingly available for exploitation and privacy may become a thing of the past. A lack of transparency and general understanding of how AI works also means that many will simply be accepting the directions of something that they have no idea about. And with this may come the evolution of man to becoming more like computers; subordinate and succumbing to a future in which we are less creative, innovative and diverse.

Along with an awareness of the risks equally comes the means of managing them in order to reap the potential rewards of AI, without suffering in the long term. In a world where AI will have such a large role, similar to that of the integration of the internet, society as a whole should have some level of understanding of AI in order to progress alongside it. As actuaries, we have seen the numerous applications of AI in our profession.

It is therefore in our best interests to acquire a strong knowledge of such technologies, to understand their role in enhancing our work and to seek new opportunities to adapt to a world where AI is increasingly integrated. With our unique and transferable skillset, we are primely positioned in doing so. It is also our responsibility to ensure appropriate regulatory action. AI should be as transparent as possible, with businesses implementation of clear data procedures and clarity regarding usage of AI.

Concluding Thoughts

In conclusion, we must recognise the widening of AI as an inevitable result of societal development.

  • Though an understanding of the risks of AI is integral, we should approach the integration of this new technology with optimism, open- eyed to the potential for enlightenment it offers our profession, and the world. 
  • Rather than accepting that complete human replacement awaits, we should instead imagine a world free from narrow tasks, that we normally complete with significant human error.
  • As actuaries, we should adapt to focus more on the human, irreplaceable elements of our profession, and worry less about getting to the answer and more about interpreting it.
  • AI is created to serve a beneficial purpose for humanity, and it offers a unique means of heightening our relevance in the modern world. 

For this reason, it is vital that we focus on the real-life challenges we are probabilistically going to face, as well as the opportunities AI provides for progression, without acting in ignorance of the risks.

“This is a guest article written by Kerry Shaw. Kerry can be found on LinkedIn here.”

Related Posts

The Last Actuary

The Last Actuary

Should All New Actuaries Learn Data Science and Programming?

Should All New Actuaries Learn Data Science and Programming?

Actuary Vs Data Scientist

Actuary Vs Data Scientist

Should All New Actuaries Learn Programming and Data Science?

Should All New Actuaries Learn Programming and Data Science?
{"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}