Forum Replies Created
-
Posts
-
Looking for one that doesn’t totally get this question wrong! All the OpenAI GPT’s don’t produce anything meaningful.
prompt: How many m’s are in Canada?
Tried this out on garage-bAInd/Platypus2-13B and got something far different. It went the other way and told me all about the Provinces and Territories in Canada. Can LLM’s prompt you when they are unsure what is actually meant so they can solve the problem?
This would be very interesting. As an engineer, I have kept detailed notes and calculations for the last 10 years. Ideas and everything. My whole life is digital in that regard. I use OneNote for this. I bought a pen and use it to markup drawings all the time so a lot of notes are hand written, but there would be ways of solving that problem too.
Hi!
My name is Eric Wiens. This is an interesting place. I love reading about what Brian is up to. I am curious about things and right now AI is on that list. I wanted to learn about prompts so I subscribed. Now I am downloading garage-bAInd/Platypus2-13B and just ran my first local LLM query using Python. As an ex IT person many years ago this is fun. I just wish I had more time to play with things and so getting some pointers along the way would be helpful as it could me a lot of time, hence subscribing to multiplex to learn about prompts!
I have four young kids and I work as an electrical engineering consultant so I am also looking for how all this stuff could apply to my job. It’s been a rough ride as the LLM’s don’t do a very good job solving engineering problems. They get so much of it wrong. I did notice when GPT 4 first came out I got some decent answers but after they added more “safety” features to it, it got stupider. Another reason to try a LLM.
I work at a small firm and we won’t be able to get access to Co-pilot even though we are Microsoft subscribers so being able to access GPT with our own document data would be helpful too.
It can help in training new staff, but again it gets so many things wrong.
The next step will be learning how to get a LLM to use our Company data to help write reports and do designs. However, the design part will need a lot of work. I am impressed with image recognition results of GPT 4 so there might be a way to train it to read a drawing.
Anyway looking forward to discussing this topic’s an others with the people here.
Eric
