Comment by byzantinegene
6 hours ago
in theory yes, but all the money is going into AI or AI-adjacent startups that no one would actually build a product that solves problems if it doesn't incorporate AI in it.
6 hours ago
in theory yes, but all the money is going into AI or AI-adjacent startups that no one would actually build a product that solves problems if it doesn't incorporate AI in it.
Using AI is fine. The key is to use it to build processes/Systems that solve problems deterministically. Instead of "asking them" to solve the problems non-deterministically themselves. It's way cheaper and robust.
As an example, we had to be able to parse most uses b2b bank statements. That means understanding the structure of around 30 different formats, some with very subtle differences within them.
The naive and expensive approach was to train an LLM to do it (after OCR).
Instead we used AI to generate a generalized python parser that is "configurable" for different structures. (And also extracts data from PDFs without OCR, unless they are pure images).
It covers the 99% of the cases. And for that 1% we pass it through AI for immediate solution, and to generate the additional deterministic config to cover it.
But I digress (lol). The point is, there are so many interesting problems that can now be solved. We should have teams of 3 to 5 people doing crazy stuff.