Questions related to Gen AI applications (week 1)

Course Q&A Generative AI for Everyone
,
1

Hello,

I have following questions on “writing” application video from week 1.

  1. What is the difference between “web version of large language model” vs “software version of the large language model”?Why web version is better over the machine version model?
  2. For some topic, I need to send several prompts to large language model. So does that scenario fall into writing category or chat category?

Thank you in advance.

Best regards,

Sukrut

2

I’m not that familiar with this course, and I’m not one of its mentors, so take this as my personal understanding.

If by the web version you mean ChatGPT or other commercially available web-based LLMs, then those models have been trained on enormous amounts of data by large teams of engineers. They’ve also been extensively fine-tuned, optimized, and customized to work well for a very broad range of users and tasks.

On the other hand, when you build your own machine learning model, you’re usually limited by the amount of training data, computing resources, engineering effort, and training time. Because of these limitations, a custom model generally won’t match the overall capabilities of commercial LLMs. However, for a very specific task or domain, your own model may actually perform better if it has been trained specifically for that purpose.

Regarding your second question, if you need to send several prompts to an LLM and receive responses, I would consider that to fall under the chat category rather than writing. My understanding is that the writing category is intended for tasks like drafting essays, reports, or other long-form written content. Since you’re interacting with the model through multiple prompts and responses, it seems more like chatting.