Recently, I've been using a free, AI-powered querying tool to explore GitHub data and gain insights. I can ask it any questions about GitHub, and AI will generate SQL based on that and tell me the answer. In most cases, it understand my intention and accurately answer my questions.
But it also makes mistakes. Sometimes, I ask about A, but it thinks I‘m asking about B; I ask only about A, but it thinks I‘m asking about A, B, and C.
This leads to incorrect SQL statements generated by AI. For example, when I asked about the most interesting metrics store projects, it thought I asked about the most popular, active, similar, fastest growing, and most contributors metrics store projects in GitHub. This resulted in invalid SQL generation.
The question link is here, if you're interested: https://ossinsight.io/explore/?id=3e634886-471c-4191-a65c-ae7fbcb5ffb1
It seems that AI can make things more complex than necessary sometimes. We should learn how to talk to AI to get it to do what we want.
Here is a course on prompt engineering. With content labeled with difficulty levels, it fits different levels of learners.
I'm learning this course. Hope it helps. :-)
Top comments (0)