How can AI be useful right now

Day 81 / 366

I realize that lately, I have been skeptical and negative about AI. But I need to keep an open mind so that I don’t miss out on any good use cases that AI might offer. I still am a ChatGPT Plus subscriber and a daily user as well. So I thought I should list down some ways in which I think AI, as it is today, can be useful.

A coding assistant

While everyone is scared about AI replacing software developers, I don’t believe that we are there yet. But I still think LLMs, especially GPT-4 right now can act as a junior developer assistant for you to help save time in some basic coding tasks.

I use ChatGPT regularly to refactor my code or to write small utility functions. For instance, writing a function that validates password strength using regex would have taken me maybe 15 minutes of googling and debugging. But with ChatGPT I can do that in 2 minutes.

One thing to remember is to only use AI for tasks that you are an expert in as well so that you can spot when it hallucinates or gets things wrong.

Stock image generation

Dall-E, Stable Diffusion, and MidJourney are all amazing in generating basic images for use in thumbnails, posters, or assets in video games. Earlier you would have to rely on a stock image service for this, and then too you might not get exactly the image that you want. With Generative AI, you can provide a detailed prompt to fit the use case that you want.

Semantic Search

LLMs combined with RAG is an excellent way to do semantic search. The LLM can act as an agent to understand the user’s query and then the RAG ensures that the output shown to the user is accurate. This has already started being used to replace FAQs and basic customer care queries.