Last year, 81% of all 500 of the Fortune 500 CEOs interviewed, agreed that artificial intelligence (AI) is “extremely important” or “very important” to the future of their companies. A sharp increase from 54% the previous year.
In recent years AI has become a buzzword that every industry seems to have embraced in some form or another. From self-driving cars, GPS navigation, recommendation engines, such as Netflix, search engines like Google, all the way to advanced medical diagnostic practices. The science fiction of tomorrow is already the science fact of today.
Yet for many, we still do not have a clear picture or understanding of what AI is, how it is being deployed and how it may benefit us personally or in our industry. Some are even fearful of AI’s progress and its threat to the workforce.
At Sanlam we have been investigating and working with AI for more than three years. In this article I’ll try to share some more insight into what we have learned on our journey thus far and illustrate how industry and humanity will benefit from adopting AI.
One of the main reasons for the explosion of interest in AI is due to the world of big data. Every day we create 2.5 quintillion bytes of it. Every day we convert thousands of historical documents and files into digital information. Every day each of us adds to the big data library through emails, new online accounts, document, photographic or video uploads to the internet, or social media interactions. Modern estimates predict that by 2020 we will have created 50 trillion GB worth of data. It’s scary to note that 2020 is only months away!
But creating all this data is one thing. Being able to analyse, sort and interpret it is another. Enter the world AI and machine learning (ML).
What is artificial intelligence (AI)?
A wonderfully simple definition is to think of AI as the science behind creating thinking machines. Therefore, AI is the use of machines to perform tasks that would normally require humans and human intelligence.
Over the past two decades, two things happened in the technology industry that greatly increased the power of AI, namely: the decrease in computer manufacturing costs and the increase in computing processing power. These advances, in conjunction with big data sets, have catapulted AI into the forefront of technological innovation.
The power of Machine Learning (ML)
Machine learning (ML) is just one of many subsets of AI and is defined as the ability of machines to learn by themselves and improve their own performance. They do not rely on human driven rules-based programming, but instead use AI algorithms to identify patterns across data and thus predict future outcomes through this pattern identification and extrapolation. Harnessed together, AI and ML learn and adapt in order to improve the quality of the predictions as time goes on.
ML could be viewed as a way of getting computers to “know things when they see them” by producing for themselves the rules their programmers cannot specify.
What can AI & ML do well, today?
“For all its glamour, AI is no different from any other technology, in that it should have a well-defined purpose in pursuit of a specific goal.” As CEO Ben Lamm of HyperGiant (a consultancy on AI to Fortune 500 companies) said in a column for VentureBeat. “If you go in knowing you want AI, but not knowing why you want it, you will fail AI, and AI will fail you.”
This brings me to an important point – the goal of AI, if deployed successfully, is that it should become invisible. Just like electricity is invisible in our lives, yet was completely transformational in allowing us to progress humanity and industry.
There are literally dozens of examples where AI is being applied, with incredible outcomes, on a daily basis. Below are some examples of where AI is creating great buzz (being highly visible) as well as where AI has become effectively invisible.
Examples where AI is already (almost) invisible:
- Optical character recognition (OCR) is a built-in service available on any multi-function printer today. OCR is largely a solved problem which converts scanned text from books, magazines or letters into electronic text with 99% accuracy.
- Siri, Cortana, Alexa, and Google all have highly advanced natural language processing tools that extend beyond “text based” question answering. They can actually function across “knowledge based question answering”. For example, ask Siri, “Who was the Australian Prime Minister when Elvis died?” or “How old was Donald Trump in 2003?” and the answers may surprise you. AI has evolved to the point where several layers of logic can be integrated to deliver the answer.
- Optimisation techniques is a technique well utilised in mining operations, rotating crops, pricing insurance and parcel delivery. Groups like Fedex, UPS and others use AI optimisation to map and plan their delivery routes. For example, if we have a truck with 10 parcels and several options for each stop, that computes to more than 3.5 million possible delivery routes. This turns into a quintillion possible options at 20 parcels. In order to achieve the most efficient delivery times for all packages on a delivery route, optimisation models are used to dynamically adjust routing by factoring in distance, traffic congestion and delivery prioritisation.
And yet, Yann LeCun, previous Chief AI scientist at Facebook, suggested that we are so far away from machines that are as intelligent as humans that we have only seen 5% of what AI can do.
There are dozens of other examples of where AI is being applied, which is creating a lot of visible buzz, such as:
- AI beating humans at most problem solving games (chess, Go and even physically solving Rubik’s Cubes in half a second). These are not irrelevant successes as they help drive important mathematical models to apply to real-world problems.
- Tesla and other self-driving cars
- Amazon and Alibaba’s pick-and-place robots
- Boston Dynamics AI robots. Please do yourself a favour and watch what they are capable of on YouTube.
Are ‘the machines’ going to take over my job?
This is one of the most asked questions and the simple answer is: yes and no. AI will not replace most jobs. But they will replace existing skill-sets. As these skills are automated and replaced, many new opportunities will emerge. According to a recent article in Forbes, for every skill that is replaced by AI, nearly two new opportunities will be created. This is excellent news for us in the long run!
Humans, the flaw in the AI machine?
AI has no emotions and is morally and ethically neutral. It will be the responsibility of all of us to ensure that AI remains this way, much in the same way that we have checks, balances and controls in place for vehicles (such as servicing, inspections, and licensing). AI is really just another tool to be added to the tool bag. As an example, a motor vehicle can cause great harm and damage, but it has also exponentially improved the way we move around. In much the same way, AI is (and will continue to be) utilised for efficiency and improvement in technologies and industries, but we will need to develop robust controls to ensure its value-added benefit can be managed in a controlled manner.
For the naysayers, I personally believe that we have a moral obligation to investigate AI to help solve global challenges in industries like healthcare – to enhance diagnostic abilities, cure cancer and develop lifesaving drugs. (AI is currently being deployed in all of these areas.)
Rise of the machines
AI as a tool is very different to the notion of sentient robots (think Terminator), also known as the singularity (where AI exceeds human abilities in all respects). The reality is that science is so far away from even understanding our own human cognition that many believe it is highly unlikely for us to develop AI into a fully self-aware and sentient robotic organism in our lifetimes.
I believe (and hope) the fear around AI in this regard is greatly overstated.
How is AI relevant to investments?
The role of AI in investments is to try to cut through the noise, eliminate human error, emotions and behavioural biases and constantly learn and adapt to ever changing market conditions, thus trying to give managers and investors an edge.
Think of an AI investment team as never having to eat or sleep. They never get sick, never forget information, never argue, get embarrassed or upset. Their only purpose is to analyse, sort and interpret the data provided to them. And they are capable of doing this without bringing any human cognitive biases, which often result in value destroying behaviour, to the table.
Sanlam leveraging AI
At Sanlam we utilise an AI and an ML investment engine as part of some of our investment solutions. We view their addition to our solutions as a diversifier to human manager risks, and they are therefore useful in creating another allocation bucket or as an additional building block in the existing line-up of strategies or funds. The predictive AI investment engine analyses and interprets available data to:
- Identify and predict evolving market behaviour
- Minimize the capital loss and time in loss (drawdowns)
- Maximize portfolio outcomes (total returns).
The purpose of our AI proposition is to address investors’ primary concern, which is their risk of capital loss, by providing a solution that can pro-actively adapt as quickly as markets evolve in order to supply investors with improved investment outcomes.
By leveraging AI and ML, we believe investors will enjoy an improved investment journey and we will ultimately deliver investment outcomes that better match investor expectations, which align well to the trend of goals-based advice practices. We know that in the global context just about every investment bank and many wealth groups are trying to build AI or currently deploy AI in various contexts. As these models slowly roll out, the debate may very well change to how much to include of active, passive and AI each.
Other areas of early adoption of AI in the investment arena are:
- Optimisation in terms of automating and managing the rebalancing of portfolios and even facilitating the creation of Statements of Advice.
- Automation. This involves work that analysts would typically have done, such as reviewing corporate filings, reviewing company disclosures and financial reports.
- Chatbots. There has already been large growth (and not slowing) on financial service websites where customers converse with an automated AI chat system which attempts to act “human” and assist, guide or problem solve.
- Using Drones – in an ever increasing fight for an analytical edge, drones fly over carparks of shopping centres or stores like a Walmart and monitor capacity of car-parks to estimate sales growth
Where to next?
As more industries embrace AI, we will see improvements in “old school” technologies – smart homes and offices that are more energy efficient, self-drive cars that could revolutionise the public transport sector (our children may never need a driver’s licence) and our personal devices will blur the lines of health, banking and investments.
With the world’s markets continuing to evolve and advance at an ever increasing speed, it is exciting to know that many companies have or are working on AI strategies to improve their respective offerings.
We in the wealth industry have often been slow to react to change but it is nice to note that we have seen more than a dozen investment management AI offerings under construction!
At Sanlam, we like to say that it’s not man versus machine, but rather man with machine, which is far more efficient than man alone.
I understand the use of AI to assist in investment making decisions. However, I’m struggling with the fact that a trade is a zero sum game. For every buyer there is a seller and ultimately - depending on price performance thereafter a winner and a loser. So, 2 questions: 1, if everybody is using AI then won’t everyone have the same ‘stock idea output’ therefore, and I’m being dramatic here, the buy sell price reaches its ‘correct’ level and stays there I.e. winners at the moment exploit the irrational behaviour of investors that AI removes. 2, again AI can’t be the saviour of all our goals given the zero sum game aspect. There will still be the exact same loser/winner on either side of a trade right?
Hi Darren, This has the potential to be a very long answer; so I will provide a short response followed a longer answer (as a separate comment): I think one should view AI’s application as a product or solution much like existing investment products, which are: (1) Run by portfolio managers with different styles, approaches philosophies and portfolio construction views. (2) Formulated with different investment strategy objectives and investment styles – for example - long only/ long short, active/passive, fundamental, quant driven etc. (3) Not a perfect science or silver bullet (4) Do not necessarily take in or make sense of all available market information so will have different views of the same stocks. Therefore, AI will not create a Utopian investment arena where all AI solutions come to the perfect valuation (if there even is such a thing) and will certainly not create stock price parity. However, we believe AI will address or improve some investment considerations currently unsolved – such as diversifying against human manager risks and provide an improvement to analytical capacity, investment decisions and risk management processes in comparison to some human managers. Even marginal improvements can lead to excellent investment outcomes. When it comes to there being a winner/loser on either side of the trade that depends on your philosophy on a relative/absolute sense but simplistically - yes, I agree that there is and will continue to be a winner/loser on either side of the trade.