GenAI Insights: Thinking, fast and slow

If concerns about AI taking over your job or role are on your mind, it's time to reconsider. We believe, it is more likely that individuals who adopt and embrace AI will discover enhanced career opportunities.

At Strawberry Sauce, we are dedicated to promoting a collaborative and creative ways of adopting AI, ensuring that nobody is left behind. Today we are going to look at ways of communicating and directing the behavior of GenAI.

What is prompt engineering?

To get the best results from GenAI, you need to know how to talk to it and ask the right questions - this skill is known as prompt engineering. Prompt engineering is a method of using AI and Large Language Models (LLMs) that involves creating specific and well-crafted inputs (prompts) that direct the AI’s outputs in a desired and effective way. It is basically how you communicate with LLMs like ChatGPT or BingChat.

Inspired by Daniel Kahneman’s “Thinking, Fast and Slow,” the ideas of System 1 and System 2 thinking can help you communicate with AI and get high quality results. This is how we understand it:

System 1 Thinking and Low-Quality Prompts

In prompt engineering, low-quality prompts are similar to System 1 thinking. These are prompts that are quickly written, unclear, or lack direction. They are like fast, instinctive responses that don’t need much thought or effort. Just as System 1 thinking in humans leads to quick, often rash decisions based on little information, low-quality prompts lead to quick, shallow responses from AI. These responses might not deal with the complexity of the query or provide the depth of information needed.

System 2 Thinking and High-Quality Prompts

On the other hand, high-quality prompts are the outcome of System 2 thinking. Creating such prompts involves careful thought and clarity. It is a conscious process that requires the user to think about what they are asking, how they are asking it, and what kind of response they want. This is similar to the slow, logical, and analytical thought process of System 2 thinking. When AI is given a high-quality prompt, it’s able to produce more detailed and accurate responses. The AI is effectively ‘encouraged’ to dig deeper, analyse more carefully, and provide more practical answers.

Let’s look at some examples. We will let you copy and paste them into your preferred GenAI and see the results.

Example 1

System 1 thinking: Tell me a joke

System 2 thinking: Tell me a joke that I can use at the start of a speech about innovation and strawberries. The audience is a mixed aged group.

Example 2

System 1 thinking: Plan a trip to Paris

System 2 thinking: Create a detailed travel plan for first-time visitors to Paris, focusing on budget-friendly attractions, French dining experiences, and cultural attractions. Include practical tips on navigating the city's public transport system and suggestions for off-the-beaten-path sights. The plan should be suitable for 2 friends in their 20s, and for 5 days.

So when you are next talking to your GenAI tool of choice, apply System 1 and 2 thinking to craft the prompts. This will enable the AI to generate more valuable, informative, and relevant responses, thereby maximising the potential of AI as a tool for information, analysis, and learning. Happy prompting!