Topic: Does prompting in an LLM matter?

🌟 Editor's Note

………………………………………………………………………

Content Source Link:

🚀 Social Post

Does prompting in an LLM matter?

The short answer is: yes.

Currently, I am in my AI period.
Reading many book on AI,
talking with AI engineers, and researchers.

I do this to filter out the 'hype' on LinkedIn vs the reality.

One of those conclusion is that prompting matters.

If you have AI developers in your company, then ask them next time how they use prompting.

They most likely leverage an API from an LLM like Claude, ChatGPT or Gemini., where they apply prompts.

Let's look at the size of popular models [parameters]:

- GPT: GPT-4 (MoE) [~1.8 trillion total, ~280B active]
- Claude: Claude 3 Opus [~200B+] (estimated)
- Gemini: Gemini 1 Ultra [~500B+] (estimated)

In short, this is HUGE!

The point of prompting is that you narrow down the focus.

Let's look at this simple example of a prompt:
"Tell me everything you know about apple"

The model does not know if you mean the fruit apple or the company?

In simplest terms with prompting you narrow down the focus to get to the right answers.

Of course, it is more complex then that and there are more nuances.

I summarised them in this below guide.

Marketer’s AI Experiments Library