The biggest challenge with AI
To figure out the why you need to understand the how.
For most of us, the biggest challenge with AI is not going to be which LLM model to use or what infra to deploy. It will be to understand why do I need AI.
Saurabh recently mentioned, “Understanding a business outcome in the context of AI's applicability is hard.” Which translates to - very few people understand how AI will help their business grow.
You need to treat AI as a toolset. Learn how the tools work. Ask someone. Or better, play with it. Once you have a mental model, you will find yourself applying it to all your problems. The more mental models you create, the larger your toolset becomes.
For example, once I understood, how embeddings and nearest neighbour search works, I could see it being used everywhere. Two trivial examples,
I could search through audio and video with queries like “Can you find me the episode where the author talks about deep work?”
With subtitles I could ask “Take me to the point in the story where the character realizes for the first time that they have special powers”.
If you are a developer, creating a new app might help in the really short term. But over time customers will want to augment their existing workflows with AI. And that won’t happen by simply integrating AI APIs in your applications. You will need to be intentional about using AI as a tool to solve the user’s problems.