Can AI Have Common Sense?

Artificial intelligence is set to revolutionise all aspects of the world we live in today, finding its way into more and more industries. The prospect of AI replacing finance jobs and a wider application of artificial intelligence in finance is a very real possibility with the advances being made.

This was part of a presentation at Level39 by Dr Catherine Havasi, a world-leading expert in the field of AI who helped to create the OpenMind Common Sense Project, now known as ConceptNet, and one of CFTE’s Contributing Experts on the AI in Finance course. Here are some other key takeaways from the session:

Insights-driven businesses will disrupt the industry

According to a Forrester report, businesses that are driven by insights will “steal $1.2 trillion per annum from their less-informed peers by 2020”. This is a huge number and shows that it is only a matter of time before AI in the finance industry becomes widespread.

Humans still have one distinct advantage over AI

Unlike human beings, current AI systems are unable to generalise in the same way humans can. Using an easy comparison, Google’s AlphaZero system was only able to beat a human opponent at chess after having played 68 million games against itself and analysed its moves. In short, the impressive power of the deep learning model is built on incomprehensible amounts of data and brute force, which humans do not require thanks to “common sense”.

Deep learning AI is not practical

Dr Havasi highlighted that there is a fundamental flaw in the way AI currently learns. It picks up on nuances and patterns by sheer data crunching which requires much more data, time and calculations than is practical. This is best illustrated by Andrej Karpathy’s (Senior Director of AI at Tesla) quote “I don’t have to actually experience crashing my car into a wall a few hundred times before I slowly start avoiding to do so.”

Humans adapt quicker

Thanks to a common understanding, humans can learn and adapt quickly from a few examples. Generalising from known occurrences, humans build analogies to understand complex concepts. However, the problem with robotics is that the learning process is very stringent and set – there is no room for expansion currently. This effect is compounded in the finance industry where jargon and nuance can make a world of difference.

Is it possible for AI to learn or pick up common sense?

In a word, yes. Currently, AI is very narrow as it is trained to react to every possible situation at the cost of practicality. However, the future suggests that it is possible to have human-level AI. Once the concept of transfer learning and domain adaptation, which refers to what has been learnt in a session, is utilised to improve generalisation in another setting, AI will learn at an exponentially quicker pace.

Although there is definitely merit to using artificial intelligence in finance, it is unquestionable that the revolution will certainly pick up speed once AI reaches a level where it can incorporate common sense into its learning. It is very likely that the widespread adoption AI in the finance industry will cause a huge disruption, the scale of which we have never seen before!

As always, for the latest developments and events in AI & FinTech, feel free to subscribe to our weekly newsletters after the jump below!