Banks Rewrite Old Code with AI
Hey, it’s Matt. This Week on AI Street:
✍️ Writing for Univ of Chicago Booth
🧟 AI decodes dead (programming) languages
📰 Regulators tout AI + more news & analysis
CHICAGO BOOTH
First Article in the Chicago Booth Review
Surprisingly, the most engagement I generally receive on social media is when I write a synopsis of an academic paper on AI + finance.
I didn’t realize there was so much interest until I started covering them.
This has led me to start writing for Chicago Booth Review, the research publication of the University’s business school.
In the article, I write about how the researchers applied transformer technology, the same that powers ChatGPT and the current AI boom, to fixed income markets. Their research suggests that much like how transformers can predict the next word in a sentence, they can do the same with bond downgrades.
This is a topic I’ve been writing more about here. How we’re in the early stages of companies applying transformers beyond just language. Stripe trained a model on the “language” of payments, and another company did so on grocery data. I imagine some firm, with a lot of computing power, is doing so for stock return data.
I’m writing a longer feature for Booth, which gives me room to quote experts. The piece is on “Interpretability,” which I just learned tries to verify how AI “thinks.”
(It’s still pretty wild to me that even the AI experts still don’t really know how AI works.)
But through some very preliminary reporting, it looks like experts are starting to make progress understanding AI interpretability.
If you’re in this field or have some suggestions of folks to talk to, please reach out: matt@ai-street.co
NEWS + ANALYSIS
AI Rewrites Dead (Programming) Language
COBOL is the NYC subway of the financial system. Both are still in use decades after their creators intended. You can’t take either offline for long…
Morgan Stanley is now aiming artificial intelligence at one of enterprise software’s biggest pain points, and one it said Big Tech hasn’t quite nailed yet: helping rewrite old, outdated code into modern coding languages.
In January, the company rolled out a tool known as DevGen.AI, built in-house on OpenAI’s GPT models. It can translate legacy code from languages like Cobol into plain English specs that developers can then use to rewrite it.
So far this year it’s reviewed nine million lines of code, saving developers 280,000 hours, said Mike Pizzi, Morgan Stanley’s global head of technology and operations.
WSJ: How Morgan Stanley Tackled One of Coding’s Toughest Problems
It’s funny that the infrastructure of the global financial system is in part built on a computer language so old that developers have to decipher it like the Dead Sea Scrolls!
You can see how this would be a big issue at banks.
Sandra Nudelman, CEO of RecodeX, a company that specializes in transforming code from one language into another, told me that while LLMs are helping, it’s still a tall order:
