The Synergy Between Machine Learning AI and Generative AI

The financial technology (fintech) sector in the United States has always been at the forefront of technological innovation. The recent emergence of generative AI has reinvigorated traditional machine learning (ML) AI, leading to the development of some sophisticated solutions that have the potential to transform our industry. This article delves into how generative AI has enhanced ML AI and highlights the specific types of machine learning solutions that have emerged in US fintech as a result. It will take those solutions and mesh them against the US mortgage industry to identify what might be in-store.

Machine Learning AI

Machine learning AI is a subset of artificial intelligence that has been around for decades. It focuses on developing algorithms and statistical models that enable computers to learn from and make predictions or decisions based on data. Instead of being explicitly programmed to perform a task, machine learning algorithms use patterns and inferences derived from the data to improve their performance over time. This approach allows machines to adapt to new and changing data, making them capable of handling complex and dynamic problems across various applications.

The Emergence of Generative AI

Generative AI, a subset of artificial intelligence, involves algorithms that can generate new data or content based on the input they have been trained on. Unlike traditional ML models, which primarily focus on identifying patterns and making predictions, generative AI can create novel content, from text and images to entire synthetic datasets. The recent advent of models like GPT-4 and DALL-E has demonstrated the power of generative AI in producing high-quality, human-like content.

Enhancing Machine Learning AI with Generative AI

The integration of generative AI with traditional ML has several key benefits:

1. Data Augmentation: The value of machine learning models generally increases with the volume of data they are fed. Generative AI can create synthetic datasets that augment existing data. This is particularly valuable in fintech, where high-quality, labeled data can be scarce. Synthetic data generation helps improve model training by providing a more diverse and comprehensive dataset, leading to better generalization and robustness. The obvious downside is if the augmented data is created based on invalid or inappropriate inputs, it can do more damage to a ML model than good.

2. Model Explainability: Generative AI models can be used to generate counterfactual explanations, helping stakeholders understand how and why a particular decision was made by the ML model. This enhances transparency and trust, which are crucial in the highly regulated fintech sector. The irony is that generative AI itself is often considered opaque in how it reaches its decisions.

3. Personalized Customer Experiences: Generative AI can utilize machine learning results and make them tangible. When fed customer behavior patterns and preferences, generative AI to create personalized financial products, services, and communications. This is the most prevalent use case available today because machine learning of customer behavior is quite mature.

Emerging Machine Learning Solutions in US Fintech

The synergy between ML and generative AI has recently led to the development of innovative solutions that are reshaping the US fintech landscape:

1. Automated Underwriting: Leveraging generative AI for data augmentation and traditional ML for risk assessment, automated underwriting systems can provide faster and more accurate loan approvals. These systems analyze vast amounts of data, including alternative credit data, to assess creditworthiness, reducing biases and improving accessibility to credit. At Brimma, we are hesitant to say that you should entrust AI to perform full underwrites. There is an enormous amount of learning and tuning required to perfect such a model. The good news is that generative AI can provide a bridge for the necessary feedback loop that enables such ongoing learnings to occur.

2. Customer Service Chatbots: Advanced chatbots powered by generative AI provide more natural and context-aware interactions. These chatbots can handle a wide range of customer inquiries, from account information to financial advice, reducing the need for human intervention and increasing operational efficiency. But that's not new news. What is new is driving the generative AI based on machine learning patterns. Imagine, for example, that you have a machine learning model that knows all of the products an agent pitches to customers AND knows which products were actually purchased. With a machine learning model at its disposal, generative AI can predict the products most likely to sell and can "cut to the chase."

3. Personalized Financial Planning: ML models analyze user data to provide tailored financial advice, while generative AI can simulate various financial scenarios, helping customers make informed decisions about their investments and savings. This is similar to the "sales" model on the previous solution except that in this case, the variables relate to the person(s), their financial situation, their financial goals, and the current and forecasted market dynamics. Again, at Brimma, we are not necessarily fans of turning financial counseling over to AI because the complexity and implications of nuanced choices can be significant and those types of conversations are often better left to licensed professionals.

Mortgage <> Fintech

As we know, the mortgage industry is hardly fintech. Not to be mean, but mortgage technology as a whole lags fintech by 5-10 years. That does not mean there will not be some cool ML+GenAI solutions in mortgage. It's just unlikely they will get widespread adoption.

So what should the mortgage industry expect? Let's baseline by saying that we have had two AI waves already:

• Wave #1 was the original machine learning AI wave that started >10 years ago. In the scheme of wave sizes, this one might be just big enough to body surf on.
• Wave #2 is the generative AI wave we are currently experiencing where chat-centric functions are being automated. This wave is a solid surfing wave. Maybe not Hawaiian-style surfing, be, still, pretty sizeable.

We adhere to the thesis that the third wave of AI (ML+genAI) will be about serving-up recommendations and coaching. In this way, this combo-AI will start to "be everywhere", including embedded in the tools we use all day, every day. Our prediction: This will be seen as a smaller wave than GenAI by itself. because this will be less obvious that it is enhancing productivity. But, when the underlying machine learning models achieve perfection, it will enable a forth wave where that same AI that was coaching and recommending suddenly is able to "do" all of those tasks...perfectly!

What that means is that you are going to see solutions where the actions inside the solution are used to train machine learning models. But unlike today's rigid models, the models will be adaptive, making them flexible enough to know that a "one size fits all" machine learning model is not appropriate in most decision-making. At Brimma, we're adding this kind of AI into our Vallia family of products. If you want to learn more, check us out at www.brimmatech.com