Feb 28

Blending Artificial Intelligence with Human Intelligence

An article by Ian Kilbride

Artificial intelligence is penetrating all our lives. For example, I am using voice recognition software to craft this article without typing a single word. Amazingly, even when I make mistakes, I can instruct the voice recognition software to delete my error and start again. Remarkable! Having worked with and applied algorithms in the financial services sector for a number of years now, before commencing with this article, I also spent time Googling articles and asked AI Chat GPT to tell me more about artificial intelligence. These are powerful sources of information, which, used intelligently, can enhance knowledge and the quality of life significantly.

One assumes that in due course my brain will be plugged into a computer in the cloud, connecting directly to the Internet and software will allow me to think my way through an article and have it translated onto the screen. In some respects, this may be a scary prospect, yet it represents remarkable progress in our human intelligence and its interface with artificial intelligence.

It has taken humankind longer than expected to develop artificial intelligence, however, since the genius mathematician, Alan Turing, wrote in 1950, “I propose to consider the question, ‘can machines think’?”. Turing later amended his question to consider whether it is possible for machines to show intelligent behaviour. Turing is widely regarded as the father of modern computing, developing as he did the bombe machine, Ultra, which successfully broke the codes of the German enigma machines during World War Two. While Turing and Ultra would never claim to have secured victory in the second world war, together, they foreshortened it possibly by years and in the process saved countless lives.

Perversely, in 1952 this same heroic figure was convicted of acts of gross indecency for homosexual acts, stripped of his security clearance and subjected to chemical castration. Two years later at the age of 41, Turing was found dead from cyanide poisoning. While the ultimate test for artificial intelligence is that of emotion and value judgment, it is clear that a lack of human intelligence, combined with prejudice and an unjust legal system led to Turing’s death.

While formal AI research was launched at Dartmouth College (USA) in 1956, after heavy funding from the US defence department, the field underwent what is termed an AI winter, with a shortage of support and little substantive progress. It was only in the 1980s that the field experienced something of a renaissance – really taking off only in the 1990s and early 2000s. Perhaps the most famous example of this was IBM’s Deep Blue computer victory over chess Grand Master Garry Kasparov in 1997. It was also only in 1995 that two Stanford students, Sergey Brin and Larry Page, received government funding while researching web page ranking and tracking. Research from these grants led to the establishment of Google and the rest, as they say, is history.

But my major personal and professional interest lies in the application of artificial intelligence to the financial services sector where developments are moving at lightning speed and revolutionising aspects of our industry. This is most obviously the case in the banking sector in which client interphase is shifting to online internet banking and its related modalities. The upside for clients is the ability to access and control their personal banking remotely at any convenient time and not have to queue for services during traditional banking hours. The safety protocols built into online banking are also improving and thus obviating the need for the use of cheques and other forgeable paper-based transactions. For the banks, it is estimated that cost savings derived of AI will amount to $447 billion in 2023 alone. 

But for me, the more fascinating aspect of AI in financial services lies in the asset management space and in particular the utilisation of algorithms as tools for fund management. In essence an algorithm is a codified set of computerised instructions designed to achieve a specific task or set of tasks. Someone once said it was like writing the recipe for a cake or an instruction manual for showing a computer what you want it to do – or simply a codified map of how to do something to achieve a particular outcome. Of course, algorithms are far more complicated than this, but once you are clear about what you want computing power to achieve (the desired outcome or result), it is then a question of writing an algorithm that can compute the relevant possibilities within the given parameters set by the programmer.

To date, my team’s development of financial and asset management algorithms has proven to be highly satisfactory, and the results are very encouraging indeed.

Will AI and algorithms replace human engagement, client care and service in financial services? Never – nor should they. But what is clear is that AI is a powerful enabler of data management which allows us to enhance the performance, service offering and efficiencies we can deliver to our clients.