Prompt engineering just got automated, so is it still worth learning this field?
Remember when prompt engineering for AI chatbots was a huge buzz, hundreds of courses sprouted like mushrooms on the internet? And some even regarded it as a profession and called themselves prompt engineers.
Although I never regarded myself as a prompt engineer, I was one of those cool guys wannabes who downloaded a bunch of prompt engineering guides to get the most out of popular language and image models like ChatGPT and Stable Diffusion. A few examples of articles are the following:
What I didn’t know back then was how fast these AI models can evolve. In just a matter of months, prompts for the language and image models were driven by natural language and no longer bound to specific rules.
Today, one of the leading AI companies, Anthropic, is taking prompt engineering to a new level by letting you generate production-ready prompts by simply describing what you want to achieve.
Prompt Generator is a powerful tool designed to simplify the process of creating effective prompts for AI models like Claude. It addresses the common challenge faced by users who struggle with the “blank page problem” when attempting to create prompts that obtain high-quality responses from AI.
By allowing users to generate prompts simply by describing their desired task, the Prompt Generator removes the need for extensive knowledge in prompt engineering. The tool incorporates best practices such as separating input data from instructions and employing techniques like “chain of thought” to guide the AI model in generating coherent and relevant outputs.
The first thing you’ll notice when you open the console dashboard is the new option for prompting.
When you click on the “Generate a prompt” button, it will open a dialog box that lets you describe the task you want to do.
Prompt: Summarize documents into 10 bullet points max
I clicked on the “Generate a prompt” and got this result:
Please carefully read the following document:
<document>
{{DOCUMENT}}
</document>
After reading through the document, identify the key points and main ideas covered in the text. Organize these key points into a concise bulleted list that summarizes the essential information from the document. The summary should have a maximum of 10 bullet points.
Your goal is to be comprehensive in capturing the core content of the document, while also being concise in how you express each summary point. Omit minor details and focus on the central ideas.
Please output your final bullet point summary inside <summary> tags.
Okay, that’s looking pretty good. I usually have the same prompt idea but I am always too lazy to type all those instructions. This prompting assistant is already proving to be handy.
Clicking on the “Start editing” button will take you to a Workbench tab where you can further improve the prompt.
Under the model settings section, you can choose the most suitable Claude model for your application.
The Temperature slider allows you to control the randomness of the generated prompts.
Lastly, the Max tokens to sample specify the maximum number of tokens from 1k to 4k to be included in the generated prompts.
When you click on the “Run” button, it will generate a response based on the parameter settings and your sample text prompt.
You can save or rename your prompts and view them on the left side panel of the dashboard. I was expecting my prompts to be accessible in the https://claude.ai/chats UI but unfortunately, there’s no way to access them there.
If you want to learn more about prompt design, check out this link.
If you’re interested in analyzing the underlying prompt and architecture, Anthropic provided a Google Colab notebook for you to explore.
There, you can easily run the code to have Claude construct prompts on your behalf.
The full metaprompt code is available on the prompt generator Google Colab notebook.
On sign-up, you will be rewarded with a $5 free credit.
It’s very easy to consume this number of credits but should be enough to do test runs before deciding to buy a subscription plan.
Anthropic offers three tiers for its free and paid plans.
The introduction of Anthropic’s new Prompt Generator feature is a very welcome change that will greatly benefit users who don’t want to spend excessive time crafting effective prompts. This tool simplifies the process of creating prompts and empowers users to leverage the capabilities of AI models like Claude more easily.
However, there are a couple of limitations to the current implementation of the Prompt Generator. Firstly, the feature is currently exclusive to the Anthropic Console dashboard and not available in the main chat section. It would be really cool if I could access the Prompt Generator directly when prompting the Claude chatbot.
Secondly, the fact that the Prompt Generator is not free, even for paying customers, makes it less accessible to everyone. Anthropic, can you make it a free feature for paying subscribers, please?
‍
Software engineer, writer, solopreneur