Believe it or not, generative AI chatbots such as ChatGPT and Google’s Bard can be trained to adopt specific personas, such as a customer service representative or a subject matter expert. But it can go further. If you feed the chatbot with the right information and context, the persona it produces can be quite helpful in a variety of different tasks. This can be everything from testing ideas to even helping someone understand a complex topic they have no experience or knowledge of.
In such a case, you can train your chatbot to help you learn how to code! So let’s take a look at how you can create your own chatbot persona for specific tasks by first by making one together. Keep in mind though, generative AI is a rapidly evolving technology so it’s important to keep up with any changes and techniques!
So with that out of the way, let’s start.
Deciding the purpose
This is the first, but most important step when crafting the persona. You first need to have a fair understanding of what the persona is supposed to do. Or better said, what you want it to do. Going in blind without a clear objective would not only cause you to not get the most out of the chatbot persona but also will likely lead you to generate information that isn’t actionable. Both you want to avoid.
Setting the persona
This step is quite honestly either the easiest or the hardest depending on your personal knowledge of the subject matter at hand. But let’s try some advertising copy. I like some MadMen-style advertising so I’m going to pick Eugene M. Schwartz.
Example Prompt: Please take on the persona of the famous copywriter, Eugene Schwartz.
When you enter this, depending on the chatbot you use, it will likely be a short introduction of the person that you requested the chatbot to act as and a bit about them – such as their work, and some bio information – which is quite valuable as we move on.
Refining the persona
Another trick that one can do is to have the LLM analyze the target’s writing style. By doing so, not only can you assist the chatbot in getting more than a surface-level understanding of what you’re about to ask of it, but it will also provide some reference for it to pull from when it’s time to generate content.
Assigning a Task
Of course, the next step will be assigning a task. In the case of this blog, we’ll ask the chatbot to provide some eye-catching headlines. That’s something that Eugene Schwartz was the master of. He wrote an entire book on the subject, and even though Google Adwords didn’t exist in his time, the concepts he employed are still with us. So when doing this task, you must know your platform. Be very clear to the chatbot.
Take the following as an example:
Today we will be selling sunscreen and I will need the following: Please generate five headlines of no more than 30 characters, five descriptions of no more than 90 characters, and one long headline of 90 characters.
Take it a step further by asking the AI to now provide six variations of what was generated.
This can also be taken much further if you feed the chatbot a company name and some branding messaging. Let’s try a made-up company together.
Please generate five headlines of no more than 30 characters, five descriptions of no more than 90 characters, and one long headline of 90 characters.
Company name: Sunaway
Company tagline: We keep the rays away
Target Audience: Beachgoers.
Using ChatGPT, this was the result:
As you can see, large language models are quite powerful if you are able to optimize your prompts, and they will only become more so once domain/industry-specific LLMs become more mainstream. Though these aren’t made to replace humans behind the keyboard, these LLMs are poised to supercharge productivity and even can help small agencies and companies swing as hard as the big fish in their industry.
So it’s becoming important to keep up with any and all changes associated with LLMs. And the best place to do this is at ODSC West 2023 this October 30th to November 2nd. With a full track devoted to NLP and LLMs, you’ll enjoy talks, sessions, events, and more that squarely focus on this fast-paced field.
Confirmed sessions include:
- Personalizing LLMs with a Feature Store
- Understanding the Landscape of Large Models
- Building LLM-powered Knowledge Workers over Your Data with LlamaIndex
- General and Efficient Self-supervised Learning with data2vec
- Towards Explainable and Language-Agnostic LLMs
- Fine-tuning LLMs on Slack Messages
- Beyond Demos and Prototypes: How to Build Production-Ready Applications Using Open-Source LLMs
- Automating Business Processes Using LangChain
- Connecting Large Language Models – Common pitfalls & challenges
What are you waiting for? Get your pass today!