By Rui Vasconcelos, Product Manager for AI/ML at Canonical – the publisher of Ubuntu
The last few years have seen the promise of general AI acclaimed across multiple industries and this vision has been particularly strong in the finance sector. We’ve currently hit the trough of the hype curve and it will take some time for engineering solutions to deliver on the touted promise. The potential is so great, that hope for general AI will require a longer term and collaborative investment, rather than a quick ROI for a single financial company. As a result, we need to see a continued collective effort from organisations in the direction of making general AI a reality – whether that is in the near future or further in time.
Artificial intelligence within financial organisations has developed from an almost unfathomable vision into tangible deployments, with applications ranging from back-end decision making to front-end customer-facing services. Financial services companies are now placing a much greater focus on AI/ML and are rearchitecting their IT and business operations to take advantage of what this new technology can offer. However, these implementations are what is known as ‘narrow’ AI, which is focused on a single or limited task and operates within a pre-programmed state. Almost all of the AI that surrounds us today is narrow AI. Everyday examples within the financial industry range from Robo-advisors to tailored credit and insurance tools. In distinction, general AI is a progression of this and is often described as an AI solution that can solve a wide range of financial services issues – from natural language understanding to anticipating risk and detecting fraud – with the additional advantage of self-learning to solve any problem without human intervention.
Narrow AI is goal-oriented and solves a particular problem, which is not necessarily bad. We have seen AlphaGo perform a singular task (playing the complex game of Go) and beat the top human expert at it. Organisations focused on being highly competitive in specific use-cases, should concentrate on narrow AI, however it is a short-term win. Those looking at wider-range problems and planning to gain long-term competitive edge need to consider investing in work that will make general AI more accessible, benefitting both the company and society in the long run. Getting there will harness tools and insights that will be very valuable to other financial services applications, even if we do not reach general AI in our lifetime. Where a ‘narrow’ AI would take into consideration historical stock prices to make time-series predictions, general AI would look into all types of accessible data that might influence the mood of investors on that day.
It’s unsurprising that AI development is a resource heavy and challenging process, and general AI development will be even more so. However, we possess an unparalleled capacity today to move it forward, both in terms of computation and human collaboration. The open source community may be able to help tackle some of the hurdles to general AI development by encouraging collaboration as well as pooling knowledge and resources. For instance, open source software allows IT teams in finance companies to benefit from frameworks, data sets, workflows, and software models in the public domain at reduced costs. In addition, the open source community sees projects as a shared responsibility, so provides an extra layer of security by continually monitoring source code for potential flaws and vulnerabilities.
A further advantage of the open source community is that it assists financial businesses to overcome the AI skills gap – one of the most frequently discussed obstacles to AI adoption. In fact, recent research shows that a third of IT teams cite a lack of skilled people and difficulty hiring for required roles as the third most-common challenge. The first hurdle is a lack of institutional support from within the business. In another study, technology’s lack of transparency was also cited as a major hurdle. With collaboration promoted at its very core, an open source approach to AI allows smaller IT teams to benefit from the wider expertise of the much broader community.
Open source will be fundamental to democratising the development of general AI. Financial services organisations who are invested in refining and improving AI for the benefit of their own operations and society will look to open source for future development. However, realising general AI will require long-term investment. Without it, the likelihood of reaching general AI in our lifetime is low. So, it’s up to financial services businesses to start concentrating resources into general AI now to make this future a reality in a short timeframe.
