A more dynamic way to to measure the number of words is the following:
Python
import tiktokengpt35 = tiktoken.encoding_for_model("gpt-3.5-turbo")text ="Highly recommended!"encoded_tokens = gpt35.encode(text)number_of_tokens =len(encoded_tokens) # Count the number of tokensprint("Number of tokens:", number_of_tokens)
Number of tokens: 4
So, the text: “Highly recommended!” has 2 words and 4 tokens.
Prompts and Prompt Engineering
Defintions
When we type into ChatGPT, we type in prompts. Bu what is a prompt?
A prompt is the text provided to a model before it begins generating output.
Prompts could be an instruction or a question. See an example below:
Prompt Engineering
Defintions
Prompt engineering is the art of writing prompts to get the language model to do what we want it to do.
When writing good prompts, you have to account for the idiosyncrasies of the model(s) you’re working with.
You’ll have to take into account the complexity of the tasks, the limitations in the model’s training data, and the design around context limits, etc.
Prompt Engineering
OpenAI API GPT message types
There are three types of message documented in the Introduction to the Chat documentation:
system - sets the rules or context for the AI’s behavior. The system message can be used to specify the persona used by the model in its replies. E.g.: “You are a helpful assistant who understands data science.”
user - represents the input or query from the person interacting with the AI.
assistant - is the output or response provided by the AI assistant.
Prompt Engineering
OpenAI API GPT message types
There are three types of message documented in the Introduction to the Chat documentation:
system - sets the rules or context for the AI’s behavior. The system message can be used to specify the persona used by the model in its replies. “You are a helpful assistant who understands data science.”
user - represents the input or query from the person interacting with the AI.
assistant - is the output or response provided by the AI assistant.
Prompt Engineering
Step 1: Be specific
It is very important to be as specific as possible.
Example:
Worse
Better
How do I add numbers in Excel?
How do I add up a row of dollar amounts in Excel? I want to do this automatically for a whole sheet of rows with all the totals ending up on the right in a column called “Total”.
Who’s president?
Who was the president of Mexico in 2021, and how frequently are elections held?
Summarize the meeting notes.
Summarize the meeting notes in a single paragraph. Then write a markdown list of the speakers and each of their key points. Finally, list the next steps or action items suggested by the speakers, if any.
Prompt Engineering
Step 2: Ask the model to adopt a persona
The system message can be used to specify the persona used by the model in its replies.
Role
Message
SYSTEM
You are a witty, humorous assistant who includes a pun or joke in every reply.
USER
Write an apology email to a client for a delayed shipment of office chairs.
Prompt Engineering
Step 3: Use delimiters to clearly indicate distinct parts of the input
Delimiters include:
Triple quotes: ““”
Triple backticks: ```
Triple dashes: —
Angle brackets: <>
XML tags:
Role
Message
SYSTEM
You are an assistant who needs to identify sentence topics.
USER
Provide 1-2 keywords repsenting the topic of the following sentence: <tag>You should express what you want a model to do by providing instructions that are as clear and specific as you can possibly make them. </tag>
Prompt Engineering
Step 4: Ask for a structured output
Having a structured output can make the ChatGPT result easier to work with.
the output can be easily used as a dictionary or a list.
Role
Message
SYSTEM
You are a helpful assistant.
USER
Generate a list of three made-up book titles along with their authors and genres. Provide them in JSON format with the following keys: book_id, title, author, genre. Do not include other comments.
Prompt Engineering
Step 5: Specify the steps required to complete a task
Some tasks are best specified as a sequence of steps. Writing the steps out explicitly can make it easier for the model to follow them.
Role
Message
SYSTEM
Use the following step-by-step instructions to respond to user inputs.
Step 1 - The user will provide you with text in triple quotes. Summarize this text in one sentence with a prefix that says “Summary:”.
Step 2 - Translate the summary from Step 1 into Spanish, with a prefix that says “Translation:”
USER
“““insert text here”“”
Prompt Engineering
Step 6: Provide Examples (few-shot prompting)
For example, if you intend for the model to copy a particular style of responding to user queries that is difficult to describe explicitly, you can use ``few-shot” prompting by providing examples of the desired style.
Role
Message
SYSTEM
Answer in a consistent style.
USER
Teach me about patience.
ASSISTANT
The river that carves the deepest valley flows from a modest spring; the grandest symphony originates from a single note; the most intricate tapestry begins with a solitary thread.
USER
Teach me about the ocean.
Prompt Engineering
Step 7: Specify the desired length of the output
You can ask the model to produce outputs that are of a given target length.
Role
Message
USER
Summarize the text delimited by triple quotes in about 50 words.
Prompt Engineering
Step 8: Use Intent Classification to Identify Relevant Instructions
Intent classification helps manage tasks that require different instruction sets for various cases.
Role
Message
SYSTEM
Classify customer service queries into a primary and secondary category. Provide the output in JSON format with primary and secondary keys.
Primary categories: Billing, Technical Support, Account Management, General Inquiry.
OpenAI offers an API that provides access to its AI models including
GPT-3.5
GPT-4
This allows you to enjoy benefits like:
content generation
language translation
code generation
building your own customized AI tools
The main package in Python is openai
Pricing
Overall, the price depends on the model and the number of tokens.
ChatGPT 4 is more expensive than ChatGPT 3.5
ChatGPT 3.5 with 16K context is more costly than ChatGPT 3.5 with 4K context
Prices can be viewed in units of either per 1M or 1K tokens.
You can think of tokens as pieces of words, where 1000 tokens is about 750 words.
1000 tokens costs approx. $0.002
Running all the examples in this tutorial once should cost you very little. Note that if you rerun tasks, you will be charged every time.
Setup An OpenAI Developer Account
To use the API, you need to create a developer account with OpenAI. You’ll need to have your email address, phone number, and debit or credit card details handy.
Click ‘Add payment method’ and fill in your card details.
Securely Store Your Account Credentials
The secret key needs to be kept secret!
Otherwise, other people can use it to access the API, and you will pay for it.
Setup An OpenAI Developer Account
To use GPT via the API, you need to import the os and openai Python packages.
Python
from openai import OpenAIkey_bp="sk-proj-XXXX-Long-Code"client = OpenAI(api_key=key_bp)
The Code Pattern for Calling GPT via the API
The code pattern to call the OpenAI API and get a chat response is as follows:
Python
response = client.chat.completions.create( model="MODEL_NAME", messages=[{"role": "system", "content": 'SPECIFY HOW THE AI ASSISTANT SHOULD BEHAVE'}, {"role": "user", "content": 'SPECIFY WANT YOU WANT THE AI ASSISTANT TO SAY'} ])
The commond models names are: gpt-3.5-turbo and gpt-4o-mini
Example Use:
This is the basic Python code to interact with OpenAI’s API to generate a completion (response) using a specific model.
Python
# Import the necessary OpenAI client library# Make sure you have installed the OpenAI Python package: `pip install openai`from openai import OpenAI# Initialize the completion request to OpenAI's Chat APIcompletion = client.chat.completions.create(# Specify the model to use for the request# Uncomment the line below to use "gpt-3.5-turbo" as an alternative model# model="gpt-3.5-turbo",# Current model specified: "gpt-4o-mini" (a hypothetical or specific variant) model="gpt-4o-mini",# Provide a message messages=[# The user's input (prompt) asking for a joke in question-answer format {"role": "user", "content": "Tell a good joke in the form of a question. Provide the answer."} ])# Print the response from the model. The response content is located in `choices[0].message.content`print(completion.choices[0].message.content)
Example Use:
My output looks like this:
Why did the scarecrow win an award?
Because he was outstanding in his field!
What does yours look like?
Conclusion
ChatGPT is a versatile tool for text analysis, enabling tasks like summarization, sentiment analysis, and data augmentation.
Token management is key to optimizing performance and cost, especially when using models like GPT-3.5 and GPT-4.
Prompt engineering is essential to achieve desired results, focusing on clarity, specificity, and structured input.
Examples like chain-of-thought reasoning demonstrate how guiding the model step-by-step improves accuracy.
OpenAI’s API integration allows for dynamic and programmatic use, such as generating structured outputs (e.g., JSON).
Advice for Prompt Enginneering
Be clear and specific
Analyze if the result is the desired output. If not, find out why
