This is one of the questions I have to the AI Advantage team about how they will use our data as we build our clones - will they use our data to train other models? Information is gold in this economy. We will have ideas that will be worthy!

My company had the same concerns. We built private, segregated instances of LLMs. We did it to protect our private information. Of course, it is not an out of the box solution as you still need add guardrails to ensure the both the input and output are considered.

@Taresa O'Brien Thank you for your perspective! When I record the lectures, the professors are already being recorded and the full video replay is available on-demand; so they know -- I'm certain they prepare accordingly. I have tried to take notes manually, but found by the time I wrote down what I wanted, I was side-tracked, didn't hear an important concept the instructor mentioned, so I just thought about solutions to my problem -- it was to record the lecture, transpose it, and create notes. Now I don't miss anything important... and quite honestly, the notes generated by AI are frankly better than I can create. I mean they pick up everything and even add nice features I like mathmatical formulas are with the 'real' academic math symbols. The one thing that has help me the most is learning by doing. AI skills require hands-on to be successful. Learning proper prompt engineering to drive the results you want from the LLM is critical. I think you have many ways to go in your challenge - here are some thoughts and certain not encompassing - 1) Is there a codified 'learning types' framework out there? You could start by using AI to develop a plan or program for each learning style. Perhaps even as simple as kintentic, auditory, visual and combinations of those learning styles. 2) Make an 'open assignment' to the students and have them generate a prompt associated with a learning goal. You can take the prompt and verify - you might get some insights as to the thinking process of the students... it's hands on... and I think some of the students will love the ability to be creative.

@Sylvie Bracque that is the challenge. You need to have some level of domain expertise. At this point in time, AI is not always 100% correct. There are some ways to minimize the hallucinations though. 1) change the "temperature" of the response. Temperature controls how creative the response is. It is a value between 0.0 and 1.5. The default for ChatGPT is 1.0, which is on the creative side. Use the command (in the prompt) "llm temperature=0.5" (without quotes), for example. The lower the number, the more deterministic the response. The higher the number, the more creative it is. The second way is to ask clarifying questions. For example, at the end of the prompt add "Ask up to 3 clarifying questions" (without quotes). This will force the LLM to ask you questions to focus the response. I am pretty sure these topics will be covered during the Summit.

I'm from Austin as well! Howdy neighbor!! 🤠