| | |

AI Consulting Ideas: Y Combinator Podcast

Major takeaways from video:

  • 50% of YC focused on AI, but not because they focus on AI, but because they focus on smart good founders, and many smart people are focusing on AI
  • Tarpits are things that seem really exciting for a lot of founders, but actually isn’t that great…might be hard to solve the real problem
  • Example tarpits for AI: copilot, chat interfaces
  • Use UX and software, and instead of chat interface, add in LLMs in background
  • Boring is often good
  • Example of boring: AI that is able to search government contracts and apply for relevant ones
  • If someone doesn’t want to buy your AI product, try to compete with the market itself
  • For example: if you develop a product for a industry and they won’t buy it, see if you can build your own company in that industry and see if you can beat them
    • I think this is KEY for consulting for big ideas like I want to do (solve their biggest issue) I need to think of ideas that would run them out of business if I made a company enhanced with AI
  • Specific is important: don’t do a catchall, include a lot of business logic
  • Prompting and GPT wrappers is the future: SAAS is basically a MySQL wrapper
  • AI security is the future
|

Developing My Own LLM Challenge

I want to learn how to build my own large language model leveraging ChatGPT and my own proprietary data. There seems to be a couple of things that I need to learn before I do that:

  1. How ChatGPT fine tuned models work
  2. What a vector database is
  3. What is Langchain

Some helpful videos on each:

ChatGPT

Vector Database

Langchain

Hands on Coding