Episode 253: AI with Katerina Skroumpelou

Web Rush

Host(s):

• John Papa @John_Papa
• Ward Bell @WardBell
• Dan Wahlin @DanWahlin
• Craig Shoemaker @craigshoemaker

Guest:

• Katerina Skroumpelou @psybercity

Recording date: Sep 14, 2023

Brought to you by

• AG Grid

  

• Narwhal

  Visit nx.dev to get the preeminent open-source toolkit for monorepo development, today.

  

Resources:

• Soul Train the Musical
• ChatGPT
• Nx Docs
• Link to Nx AI Assistant (Disclaimer: I did NOT build the UI (I am useless in UI stuff).)
• NxConf event
• Grounding LLMs
• Hallucination (artificial intelligence)
• AI & Vectors
• PostgreSQL
• What is RAG?
• Supabase
• Based on/inspired by: (we'll have our own blog post explaining our AI docs once they're out)
• Mastering temperature and Top P in Open AI
• SkyNet
• Text Blaze
• "The numeric representation of knowledge and logic—Two artificial intelligence applications in medical education" W. D. Hagamen 1986
• Ward's citation: "Medical applications of data-driven APL programs" 1975 (Note that the abstract had to explain what "data-driven" means even to a highly technical audience. That's where we were in those days.)

Timejumps

• 00:28 Welcome
• 03:54 Introducing Katerina
• 04:45 How did you get into AI?
• 05:45 Why AI assistant is important for documentation?
• 09:46 Sponsor: Ag Grid
• 10:51 How do you feel it's been going?
• 16:32 What approach did you end up using to train AI?
• 18:36 What are embeddings?
• 20:32 Sponsor: Narwhal
• 21:06 Acknowledging our AI knowledge
• 24:25 How AI answers a question
• 28:31 How did you decide how often to call the API?
• 32:41 What is Rag?
• 36:49 Final thoughts

Podcast editing on this episode done by Chris Enns of Lemon Productions.