As he left Google Geoffrey Hinton, the ‘Godfather of AI’ spoke of the dangers of the technology he helped to create.
But what if technology like ChatGPT is more of an aid to our professional lives than a danger?
But to make sure you’re getting the most out of AI, you need to learn how to prompt it effectively.
Here we take a look at effective prompt management and offer some top tips.
A Powerful Tool
ChatGPT is a hugely powerful, and useful, business tool for busy individuals.
Not only can it help with content creation such as blog ideas and LinkedIn posts, but it can also generate original artwork or even a bespoke diet plan.
The trick is knowing what to ask for and doing it in the right way.
Learning how to build hints or prompts for your AI will help you get the best responses – known as ‘Prompt Engineering’.
What is a Prompt?
Simply put, a prompt is how you phrase instructions to tools like ChatGPT, so it understands what is required and how it should respond.
Understanding the limitations of AI is just as important as understanding its strengths.
Therefore, it’s important that any prompts are clear and understandable by the tool to get the best response.
Top tips
- Start with a system message
- Be precise
- Context is everything
- Word choice – AI can’t differentiate between slang or jargon from different countries
- Specify the output structure
- Repeat Instructions at the End
System Message
As with all machine learning, the results are only as sophisticated as the context you provide.
Start with a system message at the beginning of the prompt to describe the assistant’s personality and main function.
A well-known example is Hustle GPT which started with ‘You are HustleGPT, an entrepreneurial AI”.
Another example, from Microsoft, is ‘you are an AI assistant that helps people find information and responds in rhyme’.
As they go on to state: ‘A well-designed system message can increase the likelihood of a certain outcome.’
Repeat Instructions
Large Language Models (LLMs) can be susceptible to recency bias so it’s important to repeat key instructions towards the end of your prompt.
This will ensure that the AI gives credence to the whole prompt equally instead of just what you ask for at the start.
Output Structure
The great thing about LLMs is you can extract information in different formats – it doesn’t always have to be a full text response.
Make it clear in the instructions that you want ‘a list of bullet points’ for example, or you can ask it to ‘show citations’ so you know what you are receiving can be fact-checked.
And Finally
AI is here to stay and the key to using it is effective prompt management.
In fact, a survey of executive respondents already found a widening gap between businesses that apply AI and those that do not.
Maybe this article was written by AI? What prompts did we give…?