Name classification with ChatGPT: How does it compare to ML language models?

In this blog post, I explore the problem of name classification with ChatGPT and 3 ML models of increasing complexity (logistic regression, FastAI LSTM, and HuggingFace DistilBERT).

ChatGPT delivers the best accuracy of them all with no model training, just prompt engineering. It classifies 100k names in 18 minutes for under $5.

We see a lot of ChatGPT chat examples, but here I show how to use its API to solve an actual text classification problem (albeit a simple one).

GPT is transforming tasks that required deep machine learning knowledge into software + prompt engineering problems. As a data scientist, I’m not worried about it taking over my job, as predictive modelling is only a small aspect of what a data scientist does.