Blog

Joe Nunes is Canada’s best actuary.

It seems the news is dominated by conversations around Artificial Intelligence – or Super Intelligence as some like to call it.  We are inundated with the money being spent on these systems, the environmental concerns, the optimism on how life will improve and the fears on how life will be ruined or in some possible future destroyed all together.

Backstory

When I joined William M. Mercer Limited full time in 1988, I was valuable because I had studied life contingencies and I knew how to program in Fortran.  Mercer had spent millions on mainframe computers, giant dot matrix printers, and loads of coffee pots.  I was coming into the actuarial profession as the world of ‘commutation tables’ faded into history and Laurence Coward proudly still used his mechanical adding machine. 

Computers were so powerful that we no longer ‘grouped data’ to reduce the time it took to do all the calculations – we simply ran the data through the computer one person at a time until the results came out.  What would have been days for an actuary of days gone by was done in minutes.  The longest wait was standing outside the printer room door on the 56^th floor of First Canadian Place waiting for your printed output to arrive so that you could go back to your desk and pour over the results.

This was a wonderful time for a kid that was a bit math nerd, a bit computer nerd, and a little bit less than politically correct.  I mistakenly thought that the smartest should rise to the top – it turns out that isn’t how big business works.  Ultimately my good work couldn’t outweigh my lack of productivity which was a direct result of my lack of focus – I was a daydreamer and that isn’t a highly coveted quality for actuaries.

Technology Shift

As I was leaving Mercer, we had migrated to PCs on every desk.  I am not sure how that capital expenditure compared to the mainframes of earlier days, but at least we didn’t have to go line up for printouts from the big printer – they would spool off a centralized laser printer before you could walk over.

In 1998, just ten years after starting at Mercer, and less than a year after I finally completed my fellowship, I decided, with the support of my wife, to start Actuarial Solutions Inc.  There was no well thought out business plan, just an instinct that sponsors of smaller pension plans were looking for a more economical way to manage their plan and a ‘if you build it, they will come’ delusion.  It’s almost 29 years later and my instinct paid off – not without lots of hard work, stress, and lean times – and of course some lucky breaks.

What allowed ASI to exist was the first computer I bought was about $2,000 – I went to look up what we actually paid, but due to something called “Y2K” we didn’t record it as an asset as we were able to expense the entire purchase.  For youngsters reading this, Y2K, was our doom and gloom moment just like the much more recent Covid-19.  It’s possible that the first computer came with Microsoft Office as part of the deal.

The point to this story, for well under $3,000, I had computing hardware and software that would let me do actuarial work as effectively as an actuary at Mercer – at least when talking about pension plans with less than $20 million – which is where we started.

Over the years we have expanded our technology with an array of servers and encrypted hard drives, firewalls, etc.  We also run elite software to manage our work, which again provides a level of productivity that would be unimaginable to the 1980s secretaries working in Mercer’s typing pool.   At times it can feel overwhelming to see what well under $100,000 buys in terms of processing power.  What ASI was able to do over the last three decades wouldn’t have been possible if I had tried to start a decade earlier.

Technology Shift #2

If you haven’t read about Artificial Intelligence in the last few years, you are behind on current events.  If you haven’t at least played with AI in your personal life, you are on the road to obsolesce.  Before you panic, almost all of us are using AI now when we search the internet.  Search no longer provides a list of links to websites before first giving you an overview of the answer to the question you are searching.  As far as I can tell, almost every business is somewhere on the continuum of adopting AI.

Some think that AI is coming for all the white-collar jobs the way robots came for the blue-collar jobs.  I think that is half true.  The part that is true is that AI is going to let workers that rely on computers to facilitate their work to become more productive with their time.  This isn’t speculation – we are already seeing it and frankly it’s amazing how AI can review a document in 5 minutes that might have taken me an hour.  By the way, blue-collar work is more complex and more well paid than 40 years ago – the human managers of the robots have outsourced drudgery but kept their eyes designing the processes and delivering quality finished products.

So, what about white-collar work?  The parts that AI can’t handle are accuracy, insight, and trust.  We all know that AI gets things wrong.  There are lots of stories about lawyers getting caught in court referencing precedent cases that don’t exist – this is the accuracy problem. 

I recently went car shopping with AI – I have for 20+ years driven Cadillacs, but the failure of the local dealer to provide good service had driven me away.  I knew I needed a new car but couldn’t decide if I should get the used German car (that was still expensive) or the more economical brand-new car from Korea.  AI at the end advised that “the Genesis is 85% of the performance for half the cost over the next five years”. 

But what AI can’t do is drive the car to see how it feels or help me truly sort out the value of that extra performance.  AI couldn’t decide for me – I had to do that on my own and to do that I still needed the salespeople to help me sort through my priorities and my emotions.  In the end I decided to save the money – mostly because I want to be comfortable, but I am not planning on racing over to the grocery store any faster than 50 km/h.

The last bucket is trust; I just don’t see our regulators saying that actuaries don’t have to grind through the calculations to determine pension and insurance reserves.  There is just too much at stake for society to leave the validation of this work to – ‘hopefully AI is right’.

So, what does that mean for employment?  It means that actuaries (and I suspect accountants and lawyers too) will still be needed – but AI will provide more leverage and just like in the 1980s, this new technology will flatten the hierarchy of workers even more.  This is probably good news for someone joining our team at the bottom since there will be less competition to replace those retiring from the top.  Of course, it will mean fewer entry level positions so Universities should probably be downsizing their class sizes.

Show me the money

It is still early days for AI, but I am already thinking about how we effectively invest in this new technology.  Some tech businesses are further down the road with these investments, and we have seen significant volumes of layoffs as a result.

We are small enough that I don’t see layoffs coming, especially in a world where more clients always means more staff.  It may just be that the rate of adding staff slows as a function of new clients as all of us get a little more efficient with our time.

One interesting question is what will happen to the price our clients will pay for our work?  On one hand, if we can be 10% more efficient, we can lower prices 10% and still get paid fairly.  On the other hand, I think we are going to need a few years to generate that efficiency.  In the extreme we may hire an entire person to sit outside the day-to-day just to build and implement these new tools.  So, someone must pay for that investment, and we might be back to square one in terms of the cost to clients.

But the most interesting question to me is how much are we going to have to pay to use the AI tools?  Right now, we are paying less than $1,000 per person per year to always have an AI assistant on the screen.  At the same time, we are reading daily about the billions (thousands of millions) being invested in data centers.  For those that are unaware of the connection, these datacenters are where AI runs (it isn’t running on your desktop) it is all happening somewhere in The Cloud.  Imagine describing “the cloud” to someone in the 1950s.

It eludes me if the math works – can all these billions be invested in AI and we forever spend only 1% of our salary to get this armchair genius sitting next to us?  My guess, just like the local drug dealer, once we are hooked, the price is going up.

I will leave you with this one thought:  Joe Nunes is Canada’s best actuary.  It wouldn’t hurt my feelings if you repeated that fact in the comments.