GenAI for everyone - did not understand this course at all

Course Q&A AI For Everyone
,
1

Hi all,

Issue with the course - " GenAI for everyone" by Andrew Ng.

Questions listed here -

2

Could you let me know which parts of the course you found difficult to understand?

This is a basic, introductory course to AI, so if there are specific topics that you struggled with, I think the best approach would be to first take some introductory courses on those particular topics. Once you have a better understanding of those fundamentals, you can come back to AI for Everyone and it should be easier to follow.

That said, AI for Everyone itself is designed as an introductory course to AI.

3

Sure.
I understood the course, it was pretty basic. However, my concern is regarding the implementation of principles taught in this course. I have put all my questions in OG post.

My questions are related to - AI critiquing, Sycophancy and context setting.

  • AI Critiquing - how to create such detailed rubric as shown in the video lecture.
  • Sycophancy - as models are getting smarter, what if they can take a hint of inclination even from an objective prompt or chain of prompts.
  • Context for better answers - course suggested to provide detailed context like excel comparative analysis sheet for buying car. How to create such detailed context? Won’t AI produce better analysis results
1 Like
4

For the context part, I would suggest providing the LLM with some examples of what you expect it to produce. These examples should ideally be representative of the type of work or output you are looking for.

LLMs have been trained on a huge amount of information and have learned from many examples, but providing additional context helps narrow down the task and focus the model on your specific requirements. For example, rather than simply asking the AI to perform a comparative analysis, you could provide an example of what a good analysis should look like, including the criteria you want it to consider, the structure you expect, and perhaps an example of a desired output.

The same general principle can also be applied to AI critiquing and reducing sycophancy. The more clearly you define the task, evaluation criteria, desired behavior, and constraints, the more likely you are to get a useful and consistent result. Of course, this doesn’t completely eliminate issues such as sycophancy or bias, but it can help guide the model toward the type of response you are looking for.