LLMs for the Rest of Us

Testing the LLM application can be done in a variety of ways. “People who are familiar with the knowledge in a chatbot, for example, can write questions that cover different aspects of the knowledgebase,” Carlton noted, “to see if the answers are what is expected.” This process can also detect gaps in the content. If the answers differ from those verified by the subject matter experts, then the application can be revised. “Even for simple or internal tools, it’s important to remember that you can’t test for everything. One best practice is to put in a way for users to easily provide feedback, even if it’s just a thumbs up or thumbs down,” Carlton advised. “That way, inaccuracies can be spotted quickly.”

LLMs are rapidly becoming pervasive across a wide range of use cases, both for individuals and SMBs. The underlying technology is complex, but using them is just a matter of writing your first prompt. With some practice and the right model, you can end up with a powerful tool that enhances your productivity in numerous ways. Indeed, LLMs are a technology whose time had come.