How to use cheaper AI models

PLUS: Illusion Diffusion

Hi folks!👋🏻 This is The Prompt! Consider us your AI workout buddy. Ready to break a mental sweat?

Let's get it

NEW TECH

OpenPipe: Turn expensive prompts into cheaper custom AI models

There is a lot of interest in fine-tuning expensive AI models lately.

The big question is: How do we fine-tune a 7B parameter model like LLaMa to have the same accuracy as a very big model like GPT-4?

But, wait, what is fine-tuning?

Fine-tuning is like prompt engineering on steroids. Instead of writing your instructions in text, you actually embed them in the model itself. This is how you teach your model what is considered good and bad for the job you want it to do well.

And what’s OpenPipe?

OpenPipe is a new tool I discovered, that wants to help developers to fine-tune any big model — as simply as possible. They even let you evaluate different models using the same prompts and do experiments with prompt combinations.

I think this is a new, interesting path for creating personalized AI models.

You might want to consider exploring this “trend” and see what’s out there.

🚨 What else is going on

📕 Resources

🧰 Tools of the trade

  • JSONdataAI: Get JSON data about anything from your prompt

  • Briefly: AI content summarizer

  • Mindpal: Build AI agents for your tasks, with your data

  • Ultrafocus: AI-powered focus timer

  • StockInsights: Simplify your investment research with AI

✍🏼 Prompt of the Day

TOOL

Midjourney

PROMPT

square modern ios app logo design of a [type of app], ios app icon, simple ui, flat design, white background

RESULT