Ask HN: How long until an AI tool can process curated financial data?

2 points by Biologist123 2y ago ↗ HN

I work for an organisation that collects loan data at a small pilot scale. As it scales we could use a regular fintech loan tracking tool, but I wonder how long it will be until we can submit our curated data to an AI tool for data manipulation and output. For example, if we submit 10,000 transactions, I would like the AI to generate a list of transactions made by only one borrower, available via a natural language Inquiry. Is this just around the corner? I have tried with ChatGPT and it manages it to about 90% accuracy with 100 transactions.

4 comments

[ 4.6 ms ] story [ 109 ms ] thread

zephyrus76 2y ago ↗

How uniform is the format of your dataset? Current AI tools should be pretty good at direct transformations.

If you want to generate queries, though, my guess is another 2-3 years before it's solid enough for your domains with high correctness requirements, like fintech.

Biologist123 2y ago ↗

Thanks for the opinion. It seems to me that is the point when the entire software landscape shifts.

wakahiu 2y ago ↗

Care to explain the context and why you are keen on natural language inquiries? It seems to me like the example you gave is very simplistic and can be solved by a simple dropdown filter on Excel and or SELECT * WHERE clause in a database. Note, I'm not implying that you are wrong to use natural language, simply that there seems to be more to the question.

Biologist123 2y ago ↗

Sure. For example, I might not want a borrower to see any data other than their own, so using a chatbot to access their data and output it in any way they want, eg “chatbot, did I make any payments over 50 dollars to Amazon last year. If so, when and can you list them out”. Or “was my groceries bill higher last month than this? Why?”.

I also ask because it seems that once AI can process data in response to queries like this, the software landscape shift hugely.