AI is eating Finance: JPMorgan is getting into LLMs 😳; HSBC Launches Zing App to take on Revolut, Wise, & other FinTech rivals 📲; If you’re in FinTech, you’re in Media: Robinhood acquires Chartr 📊
FinTech is Eating the World, 4 January
Hey Everyone,
Happy Thursday & Happy 2024! 🥂 And welcome back to your favorite daily FinTech newsletter 😎 Today’s issue is especially interesting as we’re looking at JPMorgan getting into LLMs (what it’s all about & why it makes sense + more reads on how AI is eating finance), HSBC that just launched Zing App to take on Revoluts and Wises of the world (can they + some deeper dives into Revolut & Wise), Robinhood that recently acquired newsletter publisher Chartr (reminding us that if you’re in FinTech, you’re in Media + a closer look at Robinhood). So let’s jump straight into the cool stuff 🌶
AI is eating Finance: JPMorgan is getting into LLMs 😳
The news 🗞️ As expected, JPMorgan JPM 0.00%↑ is getting into Large Language Models aka LLMs. The banking giant just announced DocLLM, a layout-aware generative Al language model.
Let’s take a look at what it’s all about and why it matters.
More on this 👉 DocLLM is a lightweight extension to traditional language models for visually rich document understanding.
It incorporates both text semantics and spatial layout information without relying on complex image encoders, which allows this large language model to better understand different documents (invoices, receipts, reports, contracts, etc.).
And with that comes HUGE implications of DocLLM for the finance industry:
More accurate extraction and analysis of visually rich financial documents like bank statements, insurance claims, invoices, etc. Understanding spatial layout is key for these documents.
Automated processing of handwritten, scanned, or low-quality documents which is still common in finance. The infilling approach makes the model more robust.
Structured analysis of tables and forms in earnings reports, financial filings, etc.
Answering specific financial questions from SEC filings, loan documents, etc. The visual question-answering capabilities would be super useful.
Classification of financial documents like checks, account applications, and wire transfers based on spatial patterns.
And so much more!
We're literally just scratching the surface here as cost savings and efficiency gains alone could be massive and worth hundreds of millions of dollars 🤯
✈️ THE TAKEAWAY
Looking ahead 👀 Zooming out, we must also note that this is another niche, specialized model similar to BloombergGPT for the financial field, Google MED-PaLM in the medical field, or the Codex model in the programming field, which is the basis of Copilot. While it's possible that OpenAI's GPT-5 may simply outperform them all, it's clear that 2024 should be a good year for specialized AI models. JPMorgan gets it, and that's why the bank is doubling down on artificial intelligence. AI continues to eat the world.
ICYMI: BlackRock bets Big on AI 🤖
Generative AI will make Finance Autonomous 🤖💸 [taking a big picture view on the changes taking place now & what’s next + more deeper dives into AI & Finance]
Game-changer: the first Open-Source Financial LLM is finally here 🤯
JPMorgan is developing a ChatGPT-like AI service for investors 😳 [+6 more reads]