Home » Chat GPT: Three tips to give better instructions to AI bots

Chat GPT: Three tips to give better instructions to AI bots

by admin
Chat GPT: Three tips to give better instructions to AI bots

Anna Bernstein trains artificial intelligence.
Courtesy of Anna Bernstein

Anna Bernstein is a software developer at Copy.ai, a company that develops artificial intelligence (AI) tools for creating posts and emails.

Your job is to write “prompts” to train the bot to produce high-quality, accurate text.

Here are three tips on how to write those texts to get the best results with AI.

We are currently testing machine translations of articles by our US colleagues at Insider as an added service to our readers. This article has been automatically translated and reviewed by an editor. We welcome feedback at the end of the article.

This essay is based on conversations with Anna Berstein, a 29-year-old employee at generative AI firm Copy.ai in New York. The following text has been edited for length and clarity.

As a freelance writer and historical research assistant, I’ve spent a lot of time browsing libraries. Now I’m a Prompt Engineer, helping optimize the world‘s most advanced technology. Prompt Engineering specializes in making the most of programs based on artificial intelligence. The English term “prompt” stands for the command line in which the user communicates with the respective software – comparable to the input field of a search engine such as Google or Bing.

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, an AI tool that creates copy for blogs, sales emails and social media posts can create.

read too

business/chatgpt-plant-familienurlaub-in-costa-rica-das-ging-dort-schief/”>

Chat GPT is planning a family vacation in Costa Rica – and with the result it proves that it cannot replace travel experts

He mentioned that Copy.ai, which is based on OpenAI’s GPT-3 language model, had problems with the quality of its results. He asked me if I wanted to apply. I didn’t like the stress of freelancing – plus it seemed fascinating – so I accepted even though I had studied English and had no technical background.

See also  After IT disruption - Production starts again at Volkswagen - News

Beginning of my career as an AI whisperer

Soon after, I was offered a month-long contract to work on performing different types of tones. At first I hardly knew what I was doing. But then the founder explained to me that prompt engineering is like a magic spell: if you pronounce the spell a little wrong, something wrong can happen – and vice versa. On his advice, I managed to find a solution for better tone compliance, which led to an offer for a full-time position at the company.

Since then, the scope of my work has expanded. Now I help improve existing tools and develop new ones with the goal of getting the AI ​​to spit out the best answers for users.

read too

Chat GPT with “human-level performance”: This is what the new version GPT-4 can do

In practice, I spend my days writing text-based requests that I can’t reveal due to my non-disclosure clause in the contract. I feed them into the backend of the AI ​​tools so that they can, for example, create a high-quality, grammatically correct and factually correct blog post.

Working with AI is unpredictable

To do this, I design the text based on a user’s request. Put simply, a user types something like “Write a product description about a pair of sneakers,” which I get in the backend. My task then is to write commands that result in this request producing the best output:

  • Instruction or “Write a product description on this topic”.
  • Follow-up example, or “Here are some good product descriptions, write one like this about this”.

read too

This is Mira Murati, the woman behind the hyped chatbot Chat GPT

In addition to pure prompt engineering, I also advise on how the models behave, why they might behave the way they do, which model we should use, whether we can make a particular tool, and how we should go about it.

See also  Forex, Panetta: “The rate cut is ever closer, it is physiological to increase wages”

I love the “mad scientist” part of the job, where I can come up with a silly idea for a prompt and see that it actually works. As a poet, the task also suits my obsessive way of dealing with language. It’s a really odd intersection of my literary background and analytical thinking.

However, the work is unpredictable. New language models are constantly coming onto the market, which means that I have to adapt my defaults again and again. The work itself can be tedious. There are days when I’ll spend hours – sometimes even weeks – obsessively modifying and testing a single prompt just to get it working.

read too

Social media experts on Chat GPT: “It becomes dangerous when people can no longer distinguish whether an AI is at work”

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 found about AI is the idea that it’s sentient when it’s not. When she tries to talk about an AI, we freak out because we see so many of our fears in what she says. But that’s because it’s trained on our fears, shaped by creepy sci-fi depictions of AI.

Writing good prompts, while easy to learn, is difficult to master. Getting the AI ​​to do what you want takes a lot of experimentation. Over time I’ve developed strange strategies; some of my “prompts” are really wildly structured.

Here are some tips that can help you develop better AI prompts:

See also  Markets, Piazza Affari closes sharply up: Ftse Mib + 2.49%

1. Uses a synonym dictionary

Don’t give up on a concept just because your first formulation didn’t produce the desired result. Often the right word or phrase can reveal what you are doing.

read too

Stock traders say Chat GPT gives them outdated information and they waste too much time fact-checking

2. Watch your verbs

If you want the AI ​​to fully understand your request, you should make sure your prompt contains a verb. This should clearly state your intention. For example, “Rewrite that so it’s shorter” is more descriptive than “Summarize that.”

3. Chat GPT is great in intent so use that

Make it clear from the start what you intend to do and play with word choice, tense and procedure. Try “Today we’re going to write an XYZ” or “We’re trying to write an XYZ and would like your opinion.” It’s always useful to make your intention clear. Playing with different possibilities can make a big difference.

Did you like the translation?We look forward to your feedback by email [email protected]!

You may also like

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy