I’m an AI prompt engineer. Here are 3 rules to get the best results using ChatGPT – and what
people get wrong.
- Give the chatbot a role
You need to tell the large language model what type of person they are or their job. If you tell it
to “act as a writer,” that will generate a different response from asking it to “act as a party
planner” or “act as a programmer.”
If you ask the chatbot: “act as a party planner and write this Python code,” it will produce a
worse result for what you want than if you ask it to do the same task as a developer. - Give it plenty of context best results using ChatGpt
A client I’m currently working with asked ChatGPT to take a document and turn it into a memo.
ChatGPT didn’t know what to do with that information. Its output was unclear and random.
You need to tell it how long the memo needs to be and give it certain rules to follow to
structure it.
For example, if you wanted the AI to read an article and summarize the places in the US that
were mentioned the most, it’s not enough to tell it to: “Read this article and tell me the three
places in the US that were mentioned most.”
Instead, you could prompt it: “Read this article over, take notice of all of the locations that are
mentioned, go through all of those locations, and find the ones that are in the United States.
From that list of locations in the United States, figure out the ones mentioned the most.” - Break steps into multiple prompts
Often, tasks are more complicated, and you have to give the chatbot step-by-step instructions
in multiple prompts.
With the previous example, you could tell it: “Read this through and give me a list of every
single place that’s mentioned along with the number of times it’s mentioned.”
Then, the next prompt could be: “Show me the ones in the United States.”
The prompt after that could be: “Organize it by number and then pick the three at the top,” or
whatever you want it to do.
How to Use ChatGPT Responsibly in Your Work best results using ChatGpt
Prompt engineers play a crucial role in maximizing the benefits and minimizing the risks of
utilizing AI. By promoting fairness, equity, transparency, and accountability in your work with
LLMs, this technology can be used responsibly.
Introduction to ChatGPT
ChatGPT is a conversational AI developed by OpenAI, based on the Generative Pre-trained
Transformer (GPT) models. GPT-4, the model’s latest iteration as of 2023, is a massive language
model trained on diverse internet text. However, it doesn’t know specifics about which
documents were in its training set, nor can it access any personal data unless explicitly provided
during the conversation.
ChatGPT’s ability to generate text is remarkable; it can answer questions, write essays,
summarize long documents, translate languages, simulate characters for video games, and
much more. It’s essential, however, to remember that it doesn’t understand the text like
humans do. It generates outputs based on patterns learned during its training.
Using ChatGPT effectively involves developing high-quality prompts. Be aware of the model’s
constraints: there’s a limit to the number of tokens (individual pieces of text, such as words or
punctuation) in each conversation, and you might occasionally receive an error message due to
a system glitch. Nonetheless, engaging with ChatGPT can be a highly intuitive and creative
process. best results using ChatGpt
Error, bias, and Other Failure Modes In LLMs best results using ChatGpt
Like all AI models, LLMs aren’t perfect. They can produce outputs that are incorrect, biased, or
otherwise problematic. These issues can manifest in a few ways:
Hallucination: LLMs sometimes generate factually incorrect information or details that aren’t
part of their training data. This is often referred to as hallucination.
Bias: LLMs can reflect the biases present in their training data, leading to discriminatory or
offensive outputs. Biases may span across gender, ethnicity, political ideologies and more.
Misinformation: LLMs don’t inherently know truth from falsehood, so they can sometimes
provide misleading or entirely wrong information.
Understanding these limitations is crucial when it comes to using ChatGPT responsibly and
effectively. Always use critical thinking skills to assess the responses AI provides.
Principles Of Responsible Prompt Engineering best results using ChatGpt
Prompt engineering refers to crafting prompts that guide LLMs, like ChatGPT, to produce the
desired outputs. As a responsible prompt engineer, you should consider the following
principles
Accuracy: Craft clear and precise prompts that guide the model toward generating accurate and
safe outputs. best results using ChatGpt
Unbiased: Strive to eliminate bias in your prompts to avoid perpetuating or exacerbating
existing biases.
Privacy: Ensure your prompts do not solicit sensitive or private information, respecting users’
privacy.
Evaluation: Regularly evaluate the performance of your prompts, considering not only the
technical performance but also the societal and ethical impacts.
The Broader Societal Implications Of LLMs best results using ChatGpt
LLMs have numerous benefits, such as boosting productivity, democratizing access to
information, enabling creative applications and more. As mentioned above, they also pose
challenges like spreading misinformation, perpetuating harmful biases, invading privacy,
disrupting jobs and undermining human autonomy.
ChatGPT is a powerful tool, but like all tools, it’s only as good as the person using it. You can
contribute to responsible innovation by promoting fairness, equity, transparency, and
accountability in your work with LLMs. Stay informed about the latest developments, engage in
discussions about AI ethics, and always strive for responsible AI practices in your work.