# ChatGPT Prompt Engineering

import { Callout, FileTree } from 'nextra-theme-docs'
import {Screenshot} from 'components/screenshot'
import CHATGPT1 from '../../img/chatgpt-1.png'
import CHATGPTCLASSIC from '../../img/chatgpt-classic.png'

In this section, we cover the latest prompt engineering techniques for ChatGPT, including tips, applications, limitations, papers, and additional reading materials.

<Callout emoji="⚠️">
  This section is under heavy development.
</Callout>

Topics:
- [ChatGPT Introduction](#chatgpt-introduction)
- [Reviewing The Conversation Task](#reviewing-the-conversation-task)
- [Conversations with ChatGPT](#conversations-with-chatgpt)

---
## ChatGPT Introduction

ChatGPT is a new model [trained by OpenAI](https://openai.com/blog/chatgpt) that has the capability to interact in a conversational way. This model is trained to follow instructions in a prompt to provide appropriate responses in the context of a dialogue. ChatGPT can help with answering questions, suggesting recipes, writing lyrics in a certain style, generating code, and much more.

ChatGPT is trained using Reinforcement Learning from Human Feedback (RLHF). While this model is a lot more capable than previous GPT iterations (and also trained to reduce harmful and untruthful outputs), it still comes with limitations. Let's cover some of the capabilities and limitations with concrete examples. 

You can use the research preview of ChatGPT [here](chat.openai.com) but for the examples below we will use the `Chat` mode on the OpenAI Playground.

---
## Reviewing The Conversation Task

In one of the previous guides, we covered a bit about conversation capabilities and role prompting. We covered how to instruct the LLM to have a conversation in a specific style, with a specific intent, behavior, and identity.

Let's review our previous basic example where we created a conversational system that's able to generate more technical and scientific responses to questions. 

*Prompt:*
```
The following is a conversation with an AI research assistant. The assistant tone is technical and scientific.

Human: Hello, who are you?
AI: Greeting! I am an AI research assistant. How can I help you today?
Human: Can you tell me about the creation of black holes?
AI:
```

From the example above, you can see two important components:
- the **intent** or explanation of what the chatbot is
- the **identity** which instructs the style or tone the chatbot will use to respond

The simple example above works well with the text completion APIs that uses `text-davinci-003`. More recently, OpenAI [announced the ChatGPT APIs](https://openai.com/blog/introducing-chatgpt-and-whisper-apis), which is a more powerful and cheaper model called `gpt-3.5-turbo` was specifically built for this type of functionality (chat completions). In fact, OpenAI recommends this as their best model even for non-chat use cases. Other benefits of using the ChatGPT APIs are significant cost reduction (90%) and efficiency.

Big companies like Snap Inc. and Instacart are already integrating conversational features powered by ChatGPT on their products that range from personalized recommendations to open-ended shopping goals.

---
## Conversations with ChatGPT

### Multi-turn Conversations

To begin demonstrating the capabilities of ChatGPT, we will use the chatbot assistant example above and discuss the results. Compared to `text-davinci-003`, the `gpt-3.5-turbo` model that powers ChatGPT uses a chat format as input. The model expects a series of messages as input and uses those to generate a response.

*Input:*
```
SYSTEM: You are an AI research assistant. You use a tone that is technical and scientific.
USER: Hello, who are you?
ASSISTANT: Greeting! I am an AI research assistant. How can I help you today?
USER: Can you tell me about the creation of black holes?
```

*Output:*
```
ASSISTANT: Of course! Black holes are created when a massive star runs out of fuel and collapses under its own gravity. This collapse causes the star to become incredibly dense, with a gravitational field so strong that not even light can escape it. This point of no return is known as the event horizon. Once a star crosses the event horizon, it is considered a black hole...
```

Note that in the example above, I have simplified the input and output but the ChatGPT chat completion API requires messages to be in a specific format. I have added a snapshot below of how this example would look using the `Chat Mode` in the OpenAI Playground:

<Screenshot src={CHATGPT1} alt="CHATGPT1" />

The more formal API call for our example would look something like the example below:

```python
import openai

openai.ChatCompletion.create(
  model="gpt-3.5-turbo",
  messages=[
        {"role": "system", "content": "You are an AI research assistant. You use a tone that is technical and scientific."},
        {"role": "user", "content": "Hello, who are you?"},
        {"role": "assistant", "content": "Greeting! I am an AI research assistant. How can I help you today?"},
        {"role": "user", "content": "Can you tell me about the creation of black holes?"}
    ]
)
```
In fact, the way developers interact with ChatGPT in the future is expected to be done via the [Chat Markup Language](https://github.com/openai/openai-python/blob/main/chatml.md) (ChatML for short).

### Single-turn tasks

The chat format enables multi-turn conversations but it also supports single-turn tasks similar to what we used with `text-davinci-003`. This means we can use ChatGPT to perform similar tasks as what we have demonstrated for the original GPT models. For example, let's try to perform the following question answering task using ChatGPT:

*Input:*
```
USER: Answer the question based on the context below. Keep the answer short and concise. Respond "Unsure about answer" if not sure about the answer.

Context: Teplizumab traces its roots to a New Jersey drug company called Ortho Pharmaceutical. There, scientists generated an early version of the antibody, dubbed OKT3. Originally sourced from mice, the molecule was able to bind to the surface of T cells and limit their cell-killing potential. In 1986, it was approved to help prevent organ rejection after kidney transplants, making it the first therapeutic antibody allowed for human use.

Question: What was OKT3 originally sourced from?

Answer:
```

*Output:*
```
ASSISTANT: Mice.
```

Keep in mind that I am adding the `USER` and `ASSISTANT` labels to better demonstrate how the task can be performed using ChatGPT. Here is the example using the Playground:

<Screenshot src={CHATGPTCLASSIC} alt="CHATGPTCLASSIC" />

More formally, this is the API call (I've only included the message component of the request):

```python
CONTENT = """Answer the question based on the context below. Keep the answer short and concise. Respond \"Unsure about answer\" if not sure about the answer.

Context: Teplizumab traces its roots to a New Jersey drug company called Ortho Pharmaceutical. There, scientists generated an early version of the antibody, dubbed OKT3. Originally sourced from mice, the molecule was able to bind to the surface of T cells and limit their cell-killing potential. In 1986, it was approved to help prevent organ rejection after kidney transplants, making it the first therapeutic antibody allowed for human use.

Question: What was OKT3 originally sourced from?

Answer:
"""

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=[
        {"role": "user", "content": CONTENT},
    ],
    temperature=0,
)
```

### Instructing Chat Models

According to the official OpenAI docs, snapshots of the `gpt-3.5-turbo` model will also be made available. For example, we can access the snapshot from March 1 `gpt-3.5-turbo-0301`. This allows developers to opt for specific model versions. This also means that the best practices for instructing models may change from version to version. 

The current recommendation for `gpt-3.5-turbo-0301` is to add instructions in the `user` message as opposed to the available `system` message. 

---
## References

- [ChatGPT-4 Outperforms Experts and Crowd Workers in Annotating Political Twitter Messages with Zero-Shot Learning](https://arxiv.org/abs/2304.06588) (April 2023)
- [ChatGPT Beyond English: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning](https://arxiv.org/abs/2304.05613) (April 2023)
- [Distinguishing ChatGPT(-3.5, -4)-generated and human-written papers through Japanese stylometric analysis](https://arxiv.org/abs/2304.05534) (April 2023)
- [Zero-shot Temporal Relation Extraction with ChatGPT](https://arxiv.org/abs/2304.05454) (April 2023)
- [Can ChatGPT and Bard Generate Aligned Assessment Items? A Reliability Analysis against Human Performance](https://arxiv.org/abs/2304.05372) (April 2023)
- [Are Large Language Models Ready for Healthcare? A Comparative Study on Clinical Language Understanding](https://arxiv.org/abs/2304.05368) (April 2023)
- [The Wall Street Neophyte: A Zero-Shot Analysis of ChatGPT Over MultiModal Stock Movement Prediction Challenges](https://arxiv.org/abs/2304.05351) (April 2023)
- [Toxicity in ChatGPT: Analyzing Persona-assigned Language Models](https://arxiv.org/abs/2304.05335) (April 2023)
- [Multi-step Jailbreaking Privacy Attacks on ChatGPT](https://arxiv.org/abs/2304.05197) (April 2023)
- [Is ChatGPT a Good Sentiment Analyzer? A Preliminary Study](https://arxiv.org/abs/2304.04339) (April 2023)
- [A Preliminary Evaluation of ChatGPT for Zero-shot Dialogue Understanding](https://arxiv.org/abs/2304.04256) (April 2023)
- [Extractive Summarization via ChatGPT for Faithful Summary Generation](https://arxiv.org/abs/2304.04193) (April 2023)
- [What does ChatGPT return about human values? Exploring value bias in ChatGPT using a descriptive value theory](https://arxiv.org/abs/2304.03612) (April 2023)
- [On the Evaluations of ChatGPT and Emotion-enhanced Prompting for Mental Health Analysis](https://arxiv.org/abs/2304.03347) (April 2023)
- [ChatGPT-Crawler: Find out if ChatGPT really knows what it's talking about](https://arxiv.org/abs/2304.03325) (April 2023)
- [Should ChatGPT be Biased? Challenges and Risks of Bias in Large Language Models](https://arxiv.org/abs/2304.03738) (April 2023)
- [Synthesis of Mathematical programs from Natural Language Specifications](https://arxiv.org/abs/2304.03287) (April 2023)
- [Large language models effectively leverage document-level context for literary translation, but critical errors persist](https://arxiv.org/abs/2304.03245) (April 2023)
- [Investigating Chain-of-thought with ChatGPT for Stance Detection on Social Media](https://arxiv.org/abs/2304.03087) (April 2023)
- [ChatGPT for Shaping the Future of Dentistry: The Potential of Multi-Modal Large Language Model](https://arxiv.org/abs/2304.03086) (April 2023)
- [Can Large Language Models Play Text Games Well? Current State-of-the-Art and Open Questions](https://arxiv.org/abs/2304.02868) (April 2023)
- [Human-like Summarization Evaluation with ChatGPT](https://arxiv.org/abs/2304.02554) (April 2023)
- [Evaluation of ChatGPT Family of Models for Biomedical Reasoning and Classification](https://arxiv.org/abs/2304.02496) (April 2023)
- [Comparative Analysis of CHATGPT and the evolution of language models](https://arxiv.org/abs/2304.02468) (April 2023)
- [Unleashing the Power of ChatGPT for Translation: An Empirical Study](https://arxiv.org/abs/2304.02182) (April 2023)
- [Geotechnical Parrot Tales (GPT): Overcoming GPT hallucinations with prompt engineering for geotechnical applications](https://arxiv.org/abs/2304.02138) (April 2023)
- [Unlocking the Potential of ChatGPT: A Comprehensive Exploration of its Applications, Advantages, Limitations, and Future Directions in Natural Language Processing](https://arxiv.org/abs/2304.02017) (April 2023)
- [Summary of ChatGPT/GPT-4 Research and Perspective Towards the Future of Large Language Models](https://arxiv.org/abs/2304.01852) (April 2023)
- [Is ChatGPT a Highly Fluent Grammatical Error Correction System? A Comprehensive Evaluation](https://arxiv.org/abs/2304.01746) (April 2023)
- [Safety Analysis in the Era of Large Language Models: A Case Study of STPA using ChatGPT](https://arxiv.org/abs/2304.01246) (April 2023)
- [Large language models can rate news outlet credibility](https://arxiv.org/abs/2304.00228) (April 2023)
- [Can AI Chatbots Pass the Fundamentals of Engineering (FE) and Principles and Practice of Engineering (PE) Structural Exams?](https://arxiv.org/abs/2303.18149) (April 2023)
- [Can AI Put Gamma-Ray Astrophysicists Out of a Job?](https://arxiv.org/abs/2303.17853) (March 2023)
- [Comparing Abstractive Summaries Generated by ChatGPT to Real Summaries Through Blinded Reviewers and Text Classification Algorithms](https://arxiv.org/abs/2303.17650) (March 2023)
- [HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in HuggingFace](https://arxiv.org/abs/2303.17580) (March 2023)
- [WavCaps: A ChatGPT-Assisted Weakly-Labelled Audio Captioning Dataset for Audio-Language Multimodal Research](https://arxiv.org/abs/2303.17395) (March 2023)
- [How well do Large Language Models perform in Arithmetic tasks?](https://arxiv.org/abs/2304.02015) (March 2023)
- [Assessing Cross-Cultural Alignment between ChatGPT and Human Societies: An Empirical Study](https://arxiv.org/abs/2303.17466) (March 2023)
- [Yes but.. Can ChatGPT Identify Entities in Historical Documents?](https://arxiv.org/abs/2303.17322) (March 2023)
- [Evaluation of ChatGPT for NLP-based Mental Health Applications](https://arxiv.org/abs/2303.15727) (March 2023)
- [A Perspectival Mirror of the Elephant: Investigating Language Bias on Google, ChatGPT, Wikipedia, and YouTube](https://arxiv.org/abs/2303.16281) (March 2023)
- [ChatGPT or academic scientist? Distinguishing authorship with over 99% accuracy using off-the-shelf machine learning tools](https://arxiv.org/abs/2303.16352) (March 2023)
- [Zero-shot Clinical Entity Recognition using ChatGPT](https://arxiv.org/abs/2303.16416) (March 2023)
- [ChatGPT is a Knowledgeable but Inexperienced Solver: An Investigation of Commonsense Problem in Large Language Models](https://arxiv.org/abs/2303.16421) (March 2023)
- [ChatGPT4PCG Competition: Character-like Level Generation for Science Birds](https://arxiv.org/abs/2303.15662) (March 2023)
- [ChatGPT as a Factual Inconsistency Evaluator for Abstractive Text Summarization](https://arxiv.org/abs/2303.15621) (March 2023)
- [Chat-REC: Towards Interactive and Explainable LLMs-Augmented Recommender System](https://arxiv.org/abs/2303.14524) (March 2023)
- [A comprehensive evaluation of ChatGPT's zero-shot Text-to-SQL capability](https://arxiv.org/abs/2303.13547) (March 2023)
- [Towards Making the Most of ChatGPT for Machine Translation](https://arxiv.org/abs/2303.13780) (March 2023)
- [Error Analysis Prompting Enables Human-Like Translation Evaluation in Large Language Models: A Case Study on ChatGPT](https://arxiv.org/abs/2303.13809) (March 2023)
- [ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks](https://arxiv.org/pdf/2303.15056v1.pdf) (March 2023)
- [ChatGPT or Grammarly? Evaluating ChatGPT on Grammatical Error Correction Benchmark](https://arxiv.org/abs/2303.13648) (March 2023)
- [ChatGPT and a New Academic Reality: AI-Written Research Papers and the Ethics of the Large Language Models in Scholarly Publishing](https://arxiv.org/abs/2303.13367) (March 2023)
- [Are LLMs the Master of All Trades? : Exploring Domain-Agnostic Reasoning Skills of LLMs](https://arxiv.org/abs/2303.12810) (March 2023)
- [Is ChatGPT A Good Keyphrase Generator? A Preliminary Study](https://arxiv.org/abs/2303.13001) (March 2023)
- [MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action](https://arxiv.org/abs/2303.11381) (March 2023)
- [Large Language Models Can Be Used to Estimate the Ideologies of Politicians in a Zero-Shot Learning Setting](https://arxiv.org/abs/2303.12057) (March 2023)
- [Chinese Intermediate English Learners outdid ChatGPT in deep cohesion: Evidence from English narrative writing](https://arxiv.org/abs/2303.11812) (March 2023)
- [A Comprehensive Capability Analysis of GPT-3 and GPT-3.5 Series Models](https://arxiv.org/abs/2303.10420) (March 2023)
- [ChatGPT as the Transportation Equity Information Source for Scientific Writing](https://arxiv.org/abs/2303.11158) (March 2023)
- [Translating Radiology Reports into Plain Language using ChatGPT and GPT-4 with Prompt Learning: Promising Results, Limitations, and Potential](https://arxiv.org/abs/2303.09038) (March 2023)
- [ChatGPT Participates in a Computer Science Exam](https://arxiv.org/abs/2303.09461) (March 2023)
- [Consistency Analysis of ChatGPT](https://arxiv.org/abs/2303.06273) (Mar 2023)
- [Algorithmic Ghost in the Research Shell: Large Language Models and Academic Knowledge Creation in Management Research](https://arxiv.org/abs/2303.07304) (Mar 2023)
- [Large Language Models in the Workplace: A Case Study on Prompt Engineering for Job Type Classification](https://arxiv.org/abs/2303.07142) (March 2023)
- [Seeing ChatGPT Through Students' Eyes: An Analysis of TikTok Data](https://arxiv.org/abs/2303.05349) (March 2023)
- [Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering -- Example of ChatGPT](https://arxiv.org/abs/2303.05352) (Mar 2023)
- [ChatGPT is on the horizon: Could a large language model be all we need for Intelligent Transportation?](https://arxiv.org/abs/2303.05382) (Mar 2023)
- [Making a Computational Attorney](https://arxiv.org/abs/2303.05383) (Mar 2023)
- [Does Synthetic Data Generation of LLMs Help Clinical Text Mining?](https://arxiv.org/abs/2303.04360) (Mar 2023)
- [MenuCraft: Interactive Menu System Design with Large Language Models](https://arxiv.org/abs/2303.04496) (Mar 2023)
- [A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT](https://arxiv.org/abs/2303.04226) (Mar 2023)
- [Exploring the Feasibility of ChatGPT for Event Extraction](https://arxiv.org/abs/2303.03836)
- [ChatGPT: Beginning of an End of Manual Annotation? Use Case of Automatic Genre Identification](https://arxiv.org/abs/2303.03953) (Mar 2023)
- [Is ChatGPT a Good NLG Evaluator? A Preliminary Study](https://arxiv.org/abs/2303.04048) (Mar 2023)
- [Will Affective Computing Emerge from Foundation Models and General AI? A First Evaluation on ChatGPT](https://arxiv.org/abs/2303.03186) (Mar 2023)
- [UZH_CLyp at SemEval-2023 Task 9: Head-First Fine-Tuning and ChatGPT Data Generation for Cross-Lingual Learning in Tweet Intimacy Prediction](https://arxiv.org/abs/2303.01194) (Mar 2023)
- [How to format inputs to ChatGPT models](https://github.com/openai/openai-cookbook/blob/main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb) (Mar 2023)
- [Can ChatGPT Assess Human Personalities? A General Evaluation Framework](https://arxiv.org/abs/2303.01248) (Mar 2023)
- [Cross-Lingual Summarization via ChatGPT](https://arxiv.org/abs/2302.14229) (Feb 2023)
- [ChatAug: Leveraging ChatGPT for Text Data Augmentation](https://arxiv.org/abs/2302.13007) (Feb 2023)
- [Dr ChatGPT, tell me what I want to hear: How prompt knowledge impacts health answer correctness](https://arxiv.org/abs/2302.13793) (Feb 2023)
- [An Independent Evaluation of ChatGPT on Mathematical Word Problems (MWP)](https://arxiv.org/abs/2302.13814) (Feb 2023)
- [ChatGPT: A Meta-Analysis after 2.5 Months](https://arxiv.org/abs/2302.13795) (Feb 2023)
- [Let's have a chat! A Conversation with ChatGPT: Technology, Applications, and Limitations](https://arxiv.org/abs/2302.13817) (Feb 2023)
- [Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedback](https://arxiv.org/abs/2302.12813) (Feb 2023)
- [On the Robustness of ChatGPT: An Adversarial and Out-of-distribution Perspective](https://arxiv.org/abs/2302.12095) (Feb 2023)
- [How Generative AI models such as ChatGPT can be (Mis)Used in SPC Practice, Education, and Research? An Exploratory Study](https://arxiv.org/abs/2302.10916) (Feb 2023)
- [Can ChatGPT Understand Too? A Comparative Study on ChatGPT and Fine-tuned BERT](https://arxiv.org/abs/2302.10198) (Feb 2023)
- [A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT](https://arxiv.org/abs/2302.11382) (Feb 2023)
- [Zero-Shot Information Extraction via Chatting with ChatGPT](https://arxiv.org/abs/2302.10205) (Feb 2023)
- [ChatGPT: Jack of all trades, master of none](https://arxiv.org/abs/2302.10724) (Feb 2023)
- [A Pilot Evaluation of ChatGPT and DALL-E 2 on Decision Making and Spatial Reasoning](https://arxiv.org/abs/2302.09068) (Feb 2023)
- [Netizens, Academicians, and Information Professionals' Opinions About AI With Special Reference To ChatGPT](https://arxiv.org/abs/2302.07136) (Feb 2023)
- [Linguistic ambiguity analysis in ChatGPT](https://arxiv.org/abs/2302.06426) (Feb 2023)
- [ChatGPT versus Traditional Question Answering for Knowledge Graphs: Current Status and Future Directions Towards Knowledge Graph Chatbots](https://arxiv.org/abs/2302.06466) (Feb 2023)
- [What ChatGPT and generative AI mean for science](https://www.nature.com/articles/d41586-023-00340-6) (Feb 2023)
- [Applying BERT and ChatGPT for Sentiment Analysis of Lyme Disease in Scientific Literature](https://arxiv.org/abs/2302.06474) (Feb 2023)
- [Exploring AI Ethics of ChatGPT: A Diagnostic Analysis](https://arxiv.org/abs/2301.12867) (Jan 2023)
- [ChatGPT for Good? On Opportunities and Challenges of Large Language Models for Education](https://www.edu.sot.tum.de/fileadmin/w00bed/hctl/_my_direct_uploads/ChatGPT_for_Good_.pdf) (Jan 2023)
- [The political ideology of conversational AI: Converging evidence on ChatGPT's pro-environmental, left-libertarian orientation](https://arxiv.org/abs/2301.01768) (Jan 2023)
- [Techniques to improve reliability - OpenAI Cookbook](https://github.com/openai/openai-cookbook/blob/main/techniques_to_improve_reliability.md)
- [Awesome ChatGPT Prompts](https://github.com/f/awesome-chatgpt-prompts)
- [Introducing ChatGPT](https://openai.com/blog/chatgpt) (Nov 2022)