Dearest Gentle Reader,

Last week there was a lot of chatter about the below:

Yes, this beautiful (spitting animal) was also the inspiration for Facebook to name their own Large Language Model - LLaMA

Last week they launched it's latest iteration - LLaMA 3.1. A LLM so powerful that re-ignited the debate over the pros and cons of open-source and closed models.

Why is this important, you might ask? Well last week was about food methapors. This week I am feeling more leafy, so lets use a garden analogy - Think of LLMs as gardens

• Open-source models, like LLaMA, are akin to community gardens. Everyone can see the layout, tend to the plants, and even introduce new species. It's a collaborative effort, with the potential for rapid growth and diverse flora.
• Closed-source models, like GPT-4 or Claude, are more like private estates. You can admire the beautifully manicured lawns and enjoy the fruits of their labor, but the gardening techniques remain a closely guarded secret.

Why does this distinction matter? Let's examine the pros and cons:

Open-source advantages:

• Transparency: You can examine every leaf and petal, understanding the model's inner workings.
• Collaboration: Everyone can contribute to nurturing and improving the garden.
• Customization: You can graft new branches or cultivate specific areas to suit your needs.

Open-source challenges:

1. Potential misuse: Not everyone may use the garden responsibly.
2. Less control: Once the seeds are scattered, their growth can be unpredictable.
3. Resource intensive: Maintaining a thriving garden requires significant time and resources.

Closed-source advantages:

1. Quality control: The owners can ensure the garden grows as intended.
2. Competitive edge: Unique gardeners remain exclusive.
3. Monetization: The fruits of labor (no pun intended) can be sold, funding further research and development.

Closed-source challenges:

1. Limited transparency: The soil composition and nurturing techniques remain hidden.
2. Dependency: Users rely on the estate owners for access and improvements.
3. Slower improvement: Fewer gardeners may mean slower innovation.

But why does this matter to you as a business leader?

You heard me before talking about my Build/Buy/Borrow Framework - and how deciding how to partner on AI is critical.

Open-source models offer the flexibility to tailor solutions to your specific needs without ongoing licensing costs, potentially accelerating innovation. However, they require in-house talent and resources to implement and maintain.

Closed-source models are potentially more expensive, but offer turnkey solutions with professional support, reducing the need for specialized in-house talent.

Deciding one over the other will influence not only your tech stac, but also your hiring strategy, budget allocation and partnership landscape.

This applies to you not only as a CIO or CDO, but if you work in marketing, sales, HR or Finance.

And there's no right answer, it's about selecting the right environment for your project to grow and flourish (no pun intended again).

See you next week.

(The) Lady Whistledown (of AI)

‍