FinTech LIVE New York: AI - A Modern Deus ex Machina?

"AI is not a new theory," says Kouri-Maglaras, "The concept of giving computers the ability to do tasks traditionally only humans could do goes back in the US to the 1950s."

There are two primary approaches to AI: rule-based intelligence, also known as expert systems, and machine learning. Rule-based intelligence involves programming specific rules into systems, making them static and deterministic. In contrast, machine learning emulates human decision-making by learning from examples of inputs and outputs.

Highlighting the pioneering work of Frank Rosenblatt, Kouri-Maglaras notes, "Frank Rosenblatt is sometimes called the father of deep learning for his pioneering work on artificial neural networks and the invention of the perception in 1958."

This early neural network could learn new skills through trial and error, simulating human thought processes. However, the limitations of the time led to the first "AI winter," characterised by stalled progress due to the challenges in handling unstructured data and insufficient computing power.

The resurgence of AI research in the 1980s marks a historical moment for technology.

Kouri-Maglaras explains: "Research on neural networks returned to the mainstream in the 1980s, and new researchers started to study Rosenblatt's work again."

The development of backpropagation algorithms, allowing adjustments to neural networks based on errors, was a significant advancement. "Backpropagation is a simple way of using calculus to train a multilayer neural network by propagating backwards the effect of an error."

The breakthrough came with advancements in hardware, particularly the development of GPUs by companies like NVIDIA.

"NVIDIA's GPUs were invented for graphical processing, widely used in video games and overall graphics rendering in the late 1990s," says Kouri-Maglaras. Unlike CPUs, which handle tasks serially, GPUs perform parallel processing, crucial for the massive computations required in modern AI models like GPT-4, which boasts an astounding 1 trillion parameters.

Entering the new millennium, the convergence of hardware innovation and the explosion of digital data catalysed AI's exponential growth. Kouri-Maglaras quotes Ernest Hemingway to illustrate this phenomenon:"Gradually then suddenly, which is exactly how exponential functions work."

She points out venture capital investments in AI began to surge around 2014, further accelerating development.

The impact of AI on computing power has been staggering. According to Moore's Law, the number of transistors on integrated circuits doubles approximately every 18 months, leading to a billion-fold improvement in computing efficiency over the past 30 years. "Our cars have not even become ten times more efficient in the same span," says Kouri-Maglaras.

This exponential growth in computing capabilities, coupled with the vast amounts of data generated by digital communications, has been pivotal in training increasingly sophisticated AI models.

In the insurance sector, AI is revolutionising operations and customer interactions. Prudential Financial, under Kouri-Maglaras's leadership, leverages AI for customer service, product recommendations, and financial wellness optimisation.

"We can take care of all requests like give me the status of my claim, get me my personal contract, let me pay my premium," she explains. AI-driven chatbots and virtual assistants enhance customer service by providing instant, accurate responses.

AI's potential extends to loss prevention and promoting health and financial wellness. Kouri-Maglaras highlights the use of AI-driven wellness apps to monitor physical activity, measure health metrics, and incentivise healthier behaviours.

"The idea is to use a total wellness app to monitor physical activity, measure steps and heart rate via smartwatches and smartphone apps, and to change behaviour through incentives and goals," she said. Similar principles apply to financial health, where advanced optimisation algorithms offer personalised, dynamic financial planning and recommendations.

However, the rapid pace of AI advancement necessitates responsible and ethical use. Kouri-Maglaras underscores the need for proper governance and guardrails, stating, "AI is just as good as our data, just as good as we are." She drew a parallel to ancient Greek drama, where characters like Euripides, Aeschylus, and Sophocles portrayed varying worldviews, reflecting the ethical considerations needed in AI today.

Kouri-Maglaras echoed Bill Gates' sentiment: "Sometimes we are overestimating the impact of AI in the short term and underestimating it in the long term." She emphasised that while current AI applications are transformative, the future holds even greater potential as computing power continues to grow exponentially.

"We are definitely scratching the surface."

**************

Make sure you check out the latest edition of FinTech Magazine and also sign up to our global conference series – FinTech LIVE 2024. 

**************

FinTech Magazine is a BizClik brand.