I have started working on a comprehensive OpenAI course that will be released after the Christmas break
So far I have covered:
Prompt Engineering for a wide range of use cases
Writing code in Python, Node, Curl, Typescript, PHP with examples
Tips, tricks and limitations for prompt engineering
Tweaking completion logic through parameters
Adjusting probablities for tokens
Coding for edits
DALL-E code examples
CodeX use and tricks/tips
Creating Fine tuning datasets (various methods)
Lots of fine tuning examples by use case
Processing Fine Tuning sets
Embedding in simple language with worked examples
Semantic Search of documents and text
Text Similarity Checking
Clustering to discover hidden commonality
Training a Chatbot (GPT3 based - not the new ChatGPT)
Using GPT for Blogging and SEO
Cost saving tips
Safety practices for deployment
I am leaving the new ChatGPT out for now
I wanted to know if you think I have left anything out or if there was anything in particular you would like covered.
Current status : About 280 slides completed with 10 or so to go. Will start recording audio and video soon. Hope to release at the start of January.
Thanks for the reply. I have completed the course content and will start recording it today.
Here is the final outline if you are interested.
It seems like embedding is something that is hard to come to grip with - so I have spent quite a bit of time explaining how it works with a walkthrough of a slightly modified code example from the cookbook
I’ll make a post when its done. If anyone wants an early preview before its finished, use the contact page on the website (for now)
Hey man, keen as, will keep an eye out for the newsletter!!
The hardest thing for ai generated content is management I think, its just another “person” you need to manage and requires more steps to get the content online.
Having said that, the hires I have had from upwork in the past required lots of “management” also…
My own personal interest is fact checking, but I’m not sure if GPT is the best use for this…
Hi Raymond. You might want to adjust one of the things you said about embeddings; something to the effect that ADA-002 can do an embedding for up to 6000 words. This maybe true but it would preclude using the vector for any kind of search because the max length for a query and completions is about 1500 words.
My best wishes for your course. Make attentions when you use the “ChatGPT” term, its a trademark not so much “Open” (I would call it CloseAI!) since basically they own all rights for courses, books, Footwear as well; clothing, i.e., Apparel, i.e., shirts, tank tops, belts, sweaters, hats, head caps, beanies, shorts, hoodies, sweatshirts, pants, sweatpants, head caps, underwear, socks, scarves, coats, jackets, hoodies, robes, pajamas, rainwear, swimwear, etc.