Technology experts discussed the role of artificial intelligence in the banking industry at The Asian Banker Future of Finance Summit in Thailand. Panellists addressed the relevance of this new technology, its limitations, and pondered the question of whether AI will eventually “eat your job”. 

The summit audience heard from Chrisada Sookdhis, head of data science at AIA Thailand, industry thought leader Kumardev Chatterjee, and Peter D Finn, co-founder and CEO of Synectify. The session was moderated by Emmanuel Daniel, founder of The Asian Banker, and international resource director Gordian Gaeta.

ChatGPT, short for generative pre-trained transformer, is a type of language model based on the transformer architecture that shows promise in improving productivity and automating certain tasks, but observers agree that it is far from replacing human intelligence. The talk covered the potential for decentralisation and personalisation in AI systems, as well as concerns about data privacy, biases, and security.

Sookdhis described the evolution of language technologies, including transformer architecture-based large language models for tasks like spam detection, summarisation, sentiment analysis, and auto-moderation. He explained AI’s use in banking and insurance at AIA, from traditional machine learning to cognitive and generative AI.

Daniel commented that AI forces the institution to take a holistic view of all its data sets. Traditionally, banks value the data that sits inside the bank as being more valuable than the data that sits outside of it.

Finn shared his experiences with AI and ChatGPT in the startup space. He elaborated on the advantages of having fewer constraints and more freedom to innovate. He explored AI’s potential in finance by training large language models (LLM) for personalised financial interests, businesses, and histories. One of the challenges faced by financial institutions is the lack of knowledge about training data and biases, as well as the training data used.

He said: “You can’t crack open an LLM and understand what it is deterministically going to come up with, in terms of potential biases because it’s all just neurons and synapses and statistical weights in the neural network.” 

Sookdhis said the use of AI, particularly, generative AI, and ChatGPT can improve productivity at an individual level by 20% or 30% in terms of saving time. There may be job displacement but industry pundits stress that AI is a tool and not a complete replacement for human judgment. 

ChatGPT and AI are promising technologies for the future, but security concerns continue to arise. As advancements in technology unlock new vulnerabilities, it’s essential to consider potential abuse. 

Sookdhis explained: “If someone asks [AI] for instructions to do some illegal activities—how do I jack a car, for example—it would refuse to do so. People have jailbreak, these sort of guardrails, but I do think that smart people are working on these aspects. We will be able to arrive at a place where they are safe enough, if we choose to use these models that have this built-in.”

Chatterjee elaborated on the diverse nature of AI and how it encompasses a range of technologies rather than being a single monolithic capability. He talked about the potential impact of AI in areas like compliance, know-your-customer (KYC), and payment clearing. 

He said: “In the next five or 10 years, you will see that many of these things can be highly automated, including customer service, which will be done by avatars, who will be more personal than any customer service agent can ever be.”

Gaeta concluded the session saying: “We have every reason to look into the future with ChatGPT and AI. ChatGPT can be massively abused. AI can be massively abused. Training and data feeds can be massively abused, but that’s not a reason to stop progress.”