I’m an AI prompt engineer. Here are 3 ways to use ChatGPT to get the best results.
-
Anna Bernstein is a Prompt Engineer at Copy.ai, a company that makes AI tools for generating blog posts and emails.
-
Your job is to write prompts to train the bot to produce high-quality, accurate writing.
-
Here are three tips on how to write prompts to get the best results with AI.
This essay is based on conversations with Anna Berstein, a 29-year-old prompt engineer at New York-based generative AI firm Copy.ai. The following has been edited for length and clarity.
When I was a freelance writer and historical research assistant, I spent a lot of time scrolling through microfiche in libraries. Now I’m a fast engineer helping to optimize the most advanced technology in the world.
My journey into prompt engineering began in the summer of 2021 when I met a guy at a jazz bar who was then working for Copy.ai, which makes an AI tool that can create copy for blogs, sales emails, and social media posts .
He mentioned that Copy.ai – which runs on OpenAI’s GPT-3 language model – had some issues with the quality of its outputs and asked if I wanted to try to be a fast person. I didn’t like the stress of freelancing – and it seemed intriguing – so I accepted, even though I was an English major and didn’t have a technical background.
Soon after, I was offered a one-month contract to work on performing different keys. At first I hardly knew what I was doing. But then the founder explained that prompting is like writing a spell: if you get the spell slightly wrong, something wrong can happen – and vice versa. On his advice, I managed to find a solution for better audio fidelity, which led to a full-time position at the company.
Since then, the scope of my work has grown. I now help improve existing tools and create new ones with the goal of getting the AI to spit out the best answers for users.
In practice, I spend my days writing text-based prompts — which I can’t disclose due to my non-disclosure agreement — that I put in the AI tools’ backend so they can do things like z correctly and accurately.
I do this by designing the text around a user’s request. In very simple terms, a user types something like “Write a product description about a pair of sneakers”, which I receive in the backend. So it’s my job to write prompts that can make this query generate the best output by:
-
Instruction or “Write a product description about it”
-
Example-Following, or “Here are some good product descriptions, write one like this along with them”
Besides the pure prompt engineering part of my job, I also advise on how the models behave, why they behave the way they do, which model to use, whether we can make a particular tool or not, and what approach we should take to it.
I love the “mad scientist” part of the job, where I’m able to come up with a really stupid idea for a command prompt and see that it actually works. As a poet, the role also feeds into my obsessive nature with approaching language. It’s a really strange intersection of my literary background and my analytical thinking.
However, the job is unpredictable. New language models are coming out all the time, which means I have to keep recalibrating my prompts. The work itself can be tedious. There are days when I’ll compulsively modify and test a single command prompt for hours – sometimes even weeks – just to get it working.
At the same time, it’s exciting not knowing what’s coming next.
Aside from people at parties not understanding my job, one of the big misconceptions I’ve noticed about AI is the idea that it’s sentient when it’s not. When it tries to talk about being an AI, we freak out because we reflect so many of our fears in what it says. But that’s because it’s trained on our fears, informed by creepy, sci-fi-like depictions of AI.
While writing good prompts is easy to learn, it’s actually difficult to master. Getting the AI to do what you want requires trial and error, and over time I’ve come up with weird strategies. Some of my prompts have a really wild structure.
Here are some tips that can help you develop better prompts:
1. Use a thesaurus. Don’t abandon a concept just because your first formulation didn’t produce the desired result. Oftentimes, the right word or phrase can pave the way for what you do.
2. Pay attention to your verbs. If you want the AI to fully understand your request, make sure your prompt includes a verb that clearly states your intent. For example, “rewrite to be shorter” is more effective than “compress”.
3. ChatGPT is great on purpose so take advantage of that. Be clear from the start of what you are trying to do, and play around with phrasing, tenses, and approaches. You can try “Today we’ll write an XYZ” or “We’re trying to write an XYZ and want your input”. It’s always useful to have an intention screen over what you’re doing, and playing around with different approaches can make a big difference.
Read the original article on Business Insider
