LLM + LangChain + Function Calling = 🤯

OK, so for a while, I was having a lot of difficulty seeing how these LLMs are anything more than fun little gimmicks. Then I began to see how you can treat them as sort of pre-trained experts, but with stale knowledge (kind of like a senior dev who hasn't taken much time to look at what's new out there in a while 😳). Of course, Dave has done a great job of explaining how easy it is to write small utility apps that aren't really too gimmicky.

But then discovering function calling made it all click. It makes the LLM a part of a pipeline, rather than a a solitary service that you have to train as an expert and make it do all the work. Now this stuff makes sense and, when combined with other ML and standard programming models, makes for a powerful toolkit for application programming, particularly if a set of standard starter-kits are compiled together.

In any way, the rusty old gears in my head are turning now. Excellent stuff.

1 comment