Image licensed from Shutterstock

AI has created a field that never existed before — prompt engineering.

To work as a prompt engineer, you must know how to create prompts that generate desirable responses from AI models such as those used by ChatGPT.

The salary isn’t bad.

Artificial intelligence company, Anthropic, has a role for a “prompt engineer and librarian” with a salary range between $175,000 and $335,000.

And that’s not the only company looking for prompt engineers! Let’s see in more detail what prompt engineers do, how you can become a prompt engineer and what are the requirements for this role.

What does a prompt engineer do?

A prompt engineer is someone who develops and refines AI models using prompt engineering techniques. This is like teaching a model how to do something by giving step-by-step instructions or “prompts.”

Prompt engineers work with large language models like GPT-3 (or the new GPT-4), which can generate human-like responses to text prompts. Their work focuses on designing prompts that generate desirable responses from language models, as well as enhancing the models to provide more accurate and relevant text outputs.

Here are some tasks a prompt engineer does:

  • Optimize language models using established techniques and tools
  • Write prose to test AI systems for quirks (identify AI’s errors and hidden capabilities)
  • Review and analyze datasets to identify patterns and trends in language and develop new prompts
  • Develop and maintain documentation for language models (examples, instructions, best practices, etc)
  • Train language models on new data sets, and monitor model performance to identify areas for improvement
  • Collaborate with data scientists/software engineers to integrate language models into software applications and systems.

As you can see, prompt engineers don’t write code all day, but they still need some programming knowledge to be able to work with datasets, develop and fine-tune language models, and collaborate with data scientists and software engineers.

How to learn prompt engineering

You don’t need to know coding to start learning prompt engineering. In fact, in the following link, you can find 4 free prompt engineering courses to join the top 1% of ChatGPT users.

After learning the basics, follow the steps below to continue your path as a prompt engineer.

  • Learn the basics of programming: As a prompt engineer, you’ll need to work with datasets and understand basic programming concepts. Python can be a good language for this.
  • Learn natural language processing (NLP) and machine learning (ML) concepts: Prompt engineers need to know concepts in both NLP and ML such as text preprocessing, feature engineering, model training, and optimization.
  • Practice developing prompts and fine-tuning language models: Learn to use prompt engineering techniques to generate text outputs from language models. Test different prompt types and fine-tune language models to improve performance.
  • Create a portfolio of prompt engineering projects to showcase your expertise.

The prompt engineering job that pays $335k

Anthropic pays prompt engineers up to $335k. The company specializes in developing general AI systems and language models, which may explain the high salary. It’s no surprise that Google has invested nearly $400 million in this company.

Here are some of the requirements for this job:

  • Excellent communicator, and love teaching technical concepts and creating high-quality documentation that helps out others
  • High-level familiarity with the architecture and operation of large language models
  • Basic programming skills and would be comfortable writing small Python programs
  • Stay up-to-date and informed by taking an active interest in emerging research and industry trends.

Note that the field of prompt engineering is less than 2 years old, so the prompt engineer role may differ from one company to the next, and, just like any other job, the salary for prompt engineers also varies.

The salary for similar jobs in other companies might not be as generous as Anthropic. Here’s an estimation by Indeed for an AI prompt engineer role at Boston Childrens Hospital.

Screenshot

If we compare both job postings, we’ll see that the requirements are different. In the second job posting, specifically require people with 5 years of engineering/coding experience with at least two-year experience in AI and NLP and a bachelor’s degree in computer science, artificial intelligence, or a related field (healthcare preferred).

Again, the requirements may be higher or lower in some companies, so if you don’t have any experience, a project portfolio is probably the best way to get ahead of other applicants.

If you enjoy reading stories like these and want to support me as a writer, buy me a coffee.