For all the media attention generative AI is attracting right now, few commentators are examining the game-changing advances it is enabling in the financial services sector.
But those looking for a deep dive into this area need look no further than the Generative AI Summit, taking place at Hilton Syon Park on 16 and 17 May 2023. Among the experts gathering to share their knowledge and experience is Debasmita Das, manager – AI research & product development at Mastercard.
Das will be speaking at the summit in a session on how generative AI can increase innovation, efficiency and agility in financial services.
Through her work at Mastercard, Das has witnessed first hand the ways in which generative AI is shaping the future of financial services and the specific areas it is impacting most. And she believes one of the most important applications of generative AI in the banking industry is tracking down fraudulent transactions.
“AI labs in financial service institutions have been exploring this area for quite some time using the Generative Adversarial Network,” Das explains.
The Generative Adversarial Network is a type of machine learning model that involves a generator and a discriminator. The generator takes random noise as input and creates fake data samples, while the discriminator tries to distinguish between the generated data and real data from the training set. In this way, the two networks learn together in a game-like scenario where the generator tries different ways to fool the discriminator, while the discriminator gets better at spotting the fake data among the real data.
According to Das, this method of distinguishing between legitimate and illicit activities has reaped “promising results”.
Another aspect of financial services where Das predicts generative AI will have a marked impact is investment and wealth-planning consultancy. “Financial advisors will be able to make situation-specific financial guidance by using generative AI to model diverse customer exigencies considering all types of economic scenarios,” she explains.
Das believes generative AI will also see extensive future use in compliance, algorithmic trading, the creation of personalized offers and automatic chatbots.
But for an example of how the technology has already been used to increase innovation, efficiency and agility in the financial service industry, Das points to the regulated use of synthetic data – artificially generated data that mimics the characteristics of real-world data. She believes this has the potential to solve many of the problems the banking sector is now experiencing, particularly with regards to data protection.
“Customer data that cannot be shared owing to privacy concerns and data protection rules can be replaced with shareable data created using synthetic data,” she explains.
“We can get rid of the conventional compliance obstacles and silos that come with working with sensitive data by using financial synthetic data.”
There are other applications for synthetic data, such as testing out uncertainties like market collapses or software failures. “We don’t always have the data from these circumstances,” says Das. “These gaps can be filled with synthetic data generations, which can also assist organizations in creating plans of action for situations of this nature.”
It’s certainly a brave new world, and one full of opportunities. And although generative AI has the ability to automate many tasks, Das plays down the chances of the technology replacing people in the workforce, insisting it should be seen more as an “associate”.
She explains: “Current advances in AI closely resemble human intelligence but human minds pick up knowledge through reasoning, learning, experience and sense of understanding. AI-based algorithms are quicker, more precise, can handle the ever-increasing volume of data, but they still lack intuition, emotion, or cultural sensitivity.”
What’s more, it’s imperative that the humans interacting with AI are aware of the risks, Das adds.
“The financial service company should be aware that they will be fully responsible and accountable for their use of generative AI and the content generated by the algorithms, which should not lead to any legal violations,” she warns. “Companies should be careful of how their employees are using prompts while interacting with open-source generative AI algorithms so that they do not end up sharing personal or sensitive company-related information.”
Das also highlights some of the other challenges associated with the technology. “Generative AI algorithms are trained on huge pools of data – the sources of which are in many cases unverifiable and also may be out of date. The training data might also include bias and various discriminations – which may lead to generation of output that would amplify the bias.”
So while the power of generative AI is formidable, that power comes with great responsibility.
Find out more about maximizing the value and navigating the risks of the technology at the Generative AI Summit.