Going into production
You’ve made some great progress in increasing your knowledge of how to create a ChatGPT-powered conversational agent as you’ve progressed through the book. Hopefully, you’ve also been educated on some of the limitations, challenges, and pitfalls of creating systems powered by LLMs and, most importantly, getting them production-ready. These challenges are not to be underestimated, and it’s no surprise that despite the huge adoption of LLMs over the past year or so, examples of conversational AI agents entirely powered by LLMs are thin on the ground, and the ones that are have come under some intense scrutiny and unfortunately have underperformed in some cases. Let’s look at one in the next section.
Understanding the dangers of going into production
One recent example of how LLMs in production can quickly land you in hot water is the issues found with a chatbot deployed by a delivery firm of note. I’m not going to...
