Not everything needs an LLM: A framework for evaluating when AI makes sense
1 min read
Summary
Sharanya Rao writes in VentureBeat that while machine learning (ML) has often been leveraged for repeatable, predictable tasks its use has expanded with the development of generative AI.
She argues project managers should evaluate certain key areas before determining whether ML is the best solution for a customer problem.
These include the inputs and outputs of the system; whether the same or different outputs are required for the same or different inputs; identifiable patterns in inputs and outputs; and the cost and precision of the desired system.
She provides a table demonstrating how different customer needs require different types of ML implementations, and sometime none at all.
The key takeaway is that while exciting, ML is not always the best solution and a more traditional approach should be taken if it is sufficient.