Immediate Impacts of GPT-4

Developer Productivity

Customized how-to's for every new topic
Faster throughput on simple debugging From Error Message —> Google —> Stack Overflow/Github Issues —> Code Change to Error Message —> Fix with AI
Huge throughput increases in languages and platforms developers are less fluent in
Autocomplete for whole functions
Solve the blank page problem
Overall improved ergonomics for retrieving information leads to fewer interruptions and more flow state

Every product can add Google-level semantic search

With not but an API call and the use of pgvector developers can add powerful semantic search and topic clustering features to any product.

World class zero-shot classification

Before GPT-4 classification was a separate problem to be solved but classification is an emergent property of 3.5 and 4 so instead of spinning up a separate model developers can make a simple API call.

Chat interfaces that work

Not every UI should be replaced with Chat but many interactions with apps would be improved with a UI that is essentially “just tell me in plain English what you want to do”.

In the future, I expect to see Generative UI — so the model generates a UI instead of English. I expect this to replace a huge chunk of UI code. Why build a UI when you can just build an API, teach your model about it and have it generate UI on the fly for users?