Applying AI to Finance

Also: Pitfalls with AI, what to do, and Token picker

Applying AI to Finance

You can’t go online anymore and don’t read about the latest advancements in AI. ChatGPT set the record for the fastest-growing user base in history, with over 100 million active monthly just two months after launch. And if you’ve used it, you probably get why. But how are banks and fintechs leveraging this technology? Here are some relevant links we’ve come across during the last weeks:

  • Parthean AI: A chatbot where you can ask questions about your money, build personalized financial plans and get advice about your money.

  • GPT Portfolio: People are reporting that ChatGPT is outperforming the most popular investment funds; therefore Autopilot created “The GPT Portfolio” and invested 22 million dollars so far based on advice from ChatGPT. So far the strategy is +3,4%, while S&P 500 is +4.60% in the same timeframe. 😬

  • BloombergGPT: Bloomberg has released Bloomberg GPT a 50-billion-parameter large language model for finance trained on press releases, news articles, and filings. This is useful for streamlined financial reporting, advanced risk assessment, and customized financial advisory.

  • The Impact of AI on Developer Productivity: Evidence from GitHub Copilot shows that developers with access to the AI pair programmer completed the task 55.8% faster than the control group.

  • Ramp: Uses AI to help you understand whether you pay too much for software and auto-fill your expenses.

AI is a numbers game

The possibilities of AI seem endless, and as the technology advances, we can expect to see more innovative applications. But AI is a numbers game.

“The quality of machine learning and NLP models comes down to the data you put into them,”
– Gideon Mann, Head of Bloomberg’s ML Product and Research team.

The more data you have, the better your predictions will be. Banks have a significant advantage over traditional software vendors since they access large amounts of transactional data. The challenge is however to make sense of all the data and extract information that can be used to make accurate predictions. As I’ve written before: Context matters when it comes to AI in banking.

Pitfalls with AI

While AI has the potential to revolutionize the finance industry, it's not without its pitfalls, mainly if you use AI to arrive at decisions. In financial institutions, you need to explain why your model arrived at a particular decision, both to regulators and customers. Additionally, if the data an AI is trained on contains a bias, you can be sure the model will learn and perpetuate it.

KPMG has written a report on this topic named «Generative AI models — the risks and potential rewards in business,.» The report discusses generative AI models, their potential opportunities, current considerations, and risks. It emphasizes the need for responsible AI usage and how to build trustworthy and safe AI solutions through frameworks, controls, processes, and tools. The Norwegian Digitalisation Agency has also just released a guide for responsible development and use of artificial intelligence (in the public sector) that could be relevant for people working in the financial sector.

What to do?

Fintech companies seeking to leverage AI should begin by gathering their existing structured and unstructured (non-sensitive) data, as this will enable them to capitalize on future opportunities and train their own models. Without a top-tier learning data set, building a deep learning capability is pointless. And it is probably wise getting your ducks in a row cause AI development isn't going away as Standford 2023 AI Index Report shows.

Even though it seems like many companies have a head start, even Google doesn't think they or OpenAi are positioned to win the arms race since Open Source has quietly been eating their lunch. Why? Because data quality scales better than data size.

Token picker

Say what you want about Atlassian and its products, but their new token-picker in their design system is a great idea: