Scarcely a day goes by without a news story on AI or Artificial Intelligence. Sometimes the story is about a shiny new future, like self-driving cars. Often the story is more scary, for example how AI will disrupt the jobs market. There are very few areas of life that will remain untouched by AI in the future. Personal finance has already felt its touch in the form of robo-advisor investment, as well as by more guidance-oriented services such as from EvolveMyRetirement®. But there has been a backlash. Many experts say that humans provide these services better than AI.
Types of AI
The phrase ‘Artificial Intelligence’ suggests a human-like ability to apply common sense. But this is far from how current AI systems work. At present, there are two main types of AI system:
- Reactive systems, which analyse data to determine the best response, often with the ability to train them. Such systems typically involve a neural net.
- Evolutionary systems, which have the ability to adapt and improve their responses on the fly.
Both of these types of system can often out-perform humans in well-defined tasks. For example, reactive systems have beaten the best human Chess and Go players. And evolutionary systems have found solutions to engineering problems that would have taken years for humans.
Both types have started to be used for financial planning. AI systems can’t be thrown off by emotion. Based on its training, a neural net can arrive at the logical solution to a problem. Based on its starting assumptions, an evolutionary system will remorselessly home in on the best solution. These are strengths, but also potential weaknesses.
If you’re interested in learning more about the different types of AI, a good place to start is this article from Live Science.
The limitations of AI
There’s a popular acronym in computer science: GIGO – Garbage In Garbage Out. This applies just as much to current AI systems as to traditional computer systems. If your system assumes that the stockmarket will double in value every 5 years, then this is likely to be a garbage assumption. Any results based on it will be worthless. This is just a simple example. Real life financial AI systems depend upon numerous assumptions. Their results will be very sensitive to some of them.
Neural nets often learn their assumptions from training data. Unless the training data is representative of future data, the neural net will learn the wrong rules.
The dream of AI developers is to create a system that can apply common sense. It could then detect unrealistic assumptions and avoid GIGO. This dream still appears a very long way off. In the meantime, the best way to avoid GIGO is to involve human experts. It’s rather like a driver that gets directions from a sat-nav, but knows that the recommended road is closed. A sensible driver will ignore the sat-nav. Following the sat-nav may be the right thing to do 90% of the time. Getting the most out of the sat-nav means being able to spot the other 10% and ignore it when necessary.
Using AI sensibly
Unlike machines, humans are prone to emotions and cognitive biases. This applies to you and me, and it also applies to financial advisors. A few months ago I wrote an article on how confirmation bias can get in the way of good retirement planning.
On the other hand, AI systems don’t have emotional or cognitive biases. So it makes sense for both private individuals and financial advisors to make use of the best tools available, but not blindly follow the results. AI systems can provide objectively sensible suggestions. But sometimes what’s objectively sensible just doesn’t work for an individual. Human common sense needs to be applied both to the assumptions and to the results.
EvolveMyRetirement® is the only freely available retirement planning tool based on an evolutionary system. It uses a Genetic Algorithm to evolve an optimal strategy, as well as Monte Carlo Simulation to take account of uncertainty.