Week 2: Commercial Large Language Models

Hello Friend

Today we are looking at week to of our class - AI for Creative Leaders. Written from the perspective of one of our students - Julian.

If you’re looking to learn and follow our process, this is a great opportunity to do so.

It is packed and there is a lot to go through. I’ll share the journey week by week.

Hope you join us.

Building your creative team

My previous experience with ChatGPT had been superficial. I did not have direct use for it. I played a little with it, like many, I asked it questions and gave it small tasks, but never truly explored its depths. Today's lesson changed that view.

What struck me most was the capabilities of commercial LLMs like ChatGPT, Claude, or Gemini. I was amazed at how we can create specialised assistants with them. Each custom GPT can serve as a dedicated team member - a researcher, a script consultant, a prompt engineer, or any role you can think of to help with your creative process.

The power of persona

The key to effective AI assistance lies in building detailed personas: comprehensive character profiles that guide your assistant's thinking and responses. Do you want a screenwriting assistant with deep knowledge of film noir? Or a research specialist focused on the historical accuracy of your script? By building a knowledge base - a tailored collection of relevant documents and instructions - you train an AI that actually knows its field.

In the classroom

Some of us have been playing with ChatGPT in preparation for building storyboards and working on the scripts for our 'productions'. Already, Stephanie Gornick 's approach of organising her GPTs into "departments" - each with specific knowledge and purpose - demonstrates how she enhances her creative process. By limiting each AI assistant's scope, she maintained better control over her creative direction.

Not all experiments went smoothly. Dirk Standaert 's GPT misinterpreted its knowledge base as user-uploaded content, despite precise instructions. This highlighted the limitations of the technology.

Our course leader, Nejc Susec , reminds us, "Garbage in, garbage out." The COSTAR method (Context, Objective, Style, Tone, Audience, Response) provides a structured approach to creating effective prompts. Beyond the technical aspects, it's about learning to communicate clearly with your AI assistants.

LLMs in film development

The added value of using Large Language Models in film development is obvious. Custom AI assistants help with tasks that traditionally consume substantial time and resources:

• Analyzing and breaking down scripts

• Developing character backstories and relationships

• Generating dialogue variations while maintaining character voice

• Research assistance for historical accuracy and fact-checking.

• Story development through plot analysis and structure suggestions

But be aware, when working with it, the AI can confidently present something that’s completely wrong. I don’t take its output at face value and check everything it generates. Simple as that.

Looking Ahead

Today's key insight is clear: build your own team of AI assistants - from research specialists to creative consultants - but keep your artistic intuition in the driver's seat. These assistants won't replace human creativity; they'll amplify it, if we find the right balance between automation and artistic control.

Want to follow along our class?

Create your first assistant.

Look at the creative process your wrote down and think about an assistant that could help you with your work.

Create a Persona for that assistant:

• Who role will your assistant have?

• What would you like it to help you with?

• Guidelines and examples it should follow.

Would love to see the assistants you create!

Reply to this email, I’ll be happy to review and help you finetune your personas. 🤗

Looking Ahead: Local Large Language Models

Next week, we’ll step into the world of Local Large Language Models (LLMs). What are the available models, how can we run them, what do the parameters mean and how can we adjust them.

If you’re interested in the next cohort, the applications are open.