ChatGPT 101: What Is Generative AI (and How to Use It)

If you've spent any time online in the first few weeks of 2023, you may have already picked up on the buzz surrounding ChatGPT. But what exactly is this latest tool in the world of generative artificial intelligence (AI)? 

In this article, we'll introduce ChatGPT and the technology that powers it. We'll also cover its features and limitations, as well as present some tips on how it should (and potentially should not) be used.

What is ChatGPT?

ChatGPT is a chatbot from OpenAI that enables users to "converse" with it in a way that's meant to mimic natural conversation. As a user, you can ask questions or make requests in the form of prompts, and ChatGPT will respond. The intuitive, easy-to-use, and free tool has already gained popularity as both an alternative to traditional search engines and as a tool for AI writing, among other things.  

What does"GPT" stand for?

The "GPT" in ChatGPT is short for generative pre-trained transformer. In the field of AI, training refers to the process of teaching a computer system to recognize patterns and make decisions based on input data, much like how a teacher gives information to their students, then tests their understanding of that information. 

A transformer is a type of neural network trained to analyze the context of input data and weight the significance of each part of the data accordingly. Since this type of model learns context, it's commonly used in natural language processing (NLP) to generate text similar to human writing.  (In AI, a model is a set of mathematical equations and algorithms that a computer uses to analyze data and make decisions. Read more about some common machine learning models here.)

While older AI chatbots could answer questions with detailed responses, ChatGPT uses a dialog format, which allows it to answer follow-up and clarifying questions, as well as recognize and reject inappropriate or dangerous requests (such as questions about illegal activity). 

course

ChatGPT Teach-Out

This Teach-Out introduces learners to artificial intelligence and explains how large language models and chatbots like ChatGPT work. You will better ...

4,444 already enrolled

Learn More

Average time: 1 month(s)

Learn at your own pace

How does ChatGPT work?

ChatGPT uses natural language processing (NLP), an AI technology that deals with understanding, analyzing, and generating human-like language. The large language model (LLM) was trained using a combination of two major inputs:

1. A massive volume of sample text pulled from web pages and program code before the end of 2021

2. Conversations provided by real humans, who demonstrated the desired responses to provided prompts, then ranked outputs from the model based on the quality of the response  

Did you know?

Providing occasional feedback from humans to an AI model is a technique known as reinforcement learning from human feedback (RLHF). Leveraging this technique can help fine-tune a model by improving both safety and reliability.

guided project

ChatGPT for Beginners: SciFi Writing with Dall-e

This 1-hour long project-based course is for anyone interested in exploring the capabilities of cutting-edge AI technology. In it you will learn how to ...

8 already enrolled

BEGINNER level

Learn More

Average time: 1 month(s)

Learn at your own pace

Skills you'll build:

Critical Thinking, Creative Writing, AI communication

Features and limitations of ChatGPT (and other generative AI)

ChatGPT represents an exciting advancement in generative AI, with several features that could help accelerate certain tasks when used thoughtfully. It also comes with limitations. Understanding both the features and limitations is key to leveraging this technology for the greatest impact.

Features of ChatGPT

Some of the features ChatGPT include the ability to:

• Generate human-like text that mimics the style and structure of the input data

• Generate a response to a given prompt or input text. This could include writing a story or answering a question. 

• Generate text in multiple languages

• Modify the style of generated text (formal or informal, for example)

• Ask clarifying questions to better understand the intent of input data

• Respond with text that is consistent with the context of a conversation, such as offering follow-up instructions or understanding references made to previous questions

Other generative AI models can perform similar tasks with images, sounds, and video.

Limitations and risks of ChatGPT

While ChatGPT is a powerful tool, it has its limitations. To start, these types of transformer models lack common sense reasoning ability. This can translate to a limited ability to handle complexity, nuance, and questions around emotions, values, beliefs, and abstract concepts. These limitations can manifest in many ways:

• It does not understand the meaning of the text it generates. While some output from ChatGPT may sound humanlike, the model isn't human. This has a few implications. It may have a limited ability to handle nuance, ambiguity, or things like sarcasm or irony. Perhaps more problematic is the fact that it can generate text that sounds plausible but is incorrect or even nonsensical. What's more, it can't verify the veracity of its output.   

• It can generate biased, discriminatory, or offensive text. A language model like ChatGPT is only as good as its input data. This model was trained on large amounts of text data from across the internet, including biased input. If the data used to train the model is biased, this can show up in the generated text. 

• Responses can be rooted in outdated information. The model has limited knowledge of events after 2021 and is not connected to the wider internet. If you're using ChatGPT to produce code, for example, it could be pulling from outdated examples that no longer meet modern cybersecurity standards. 

• Output can be formulaic. ChatGPT can generate text that's similar to existing text and is known to overuse certain phrases. This can mean text that reads as flat and unimaginative, or in more extreme cases, could constitute plagiarism or a copyright infringement.  

• The tool isn't always available. The exploding popularity of ChatGPT has led to some capacity issues. When the servers get overloaded, you may get a message that "ChatGPT is at capacity."

guided project

Machine Learning with ChatGPT: Image Classification Model

In this 1-hour project, you will learn how to build a machine learning model using ChatGPT. We will use the MNIST database which is a large database of ...

4 already enrolled

INTERMEDIATE level

Learn More

Average time: 1 month(s)

Learn at your own pace

Skills you'll build:

Artificial Neural Network, Python Programming, Machine Learning, keras

6 ways to use ChatGPT to augment your work

With a thorough understanding of the features, limitations, and risks associated with generative AI, it becomes a powerful tool to save time and energy. New use cases are emerging every day; here are just a few ways you might incorporate ChatGPT (or another generative AI model) into your day-to-day work:

1. Conduct research. 

While ChatGPT lacks some of the critical functionality of Google and other search engines (particularly in making attribution easy to see and access), it can be a useful starting point for conducting research. Here are two examples:

2. Beat writer's block.

If you've ever experienced difficulty getting started on a writing project while staring at a blank page, you're not alone. Depending on the source of your writer's block and what you're trying to achieve, you may be able to use ChatGPT to overcome inertia:

• Get your creativity flowing with generated writing prompts.

• Build an outline or structure from key points you want to include.

• Generate a first paragraph to build upon. (You can go back and revise or delete it later.)

• Find that word that's on the tip of your tongue.

  

3. Strengthen an existing piece of writing.

You can input an existing piece of text into ChatGPT and ask it to identify uses of passive voice, repetitive phrases or word usage, or grammatical errors. This could be particularly useful if you're writing in a language for which you're not a native speaker.

4. Write and debug code.

ChatGPT isn't just for use with text as prose. You can also use to:

• Explain the functionality of unfamiliar input code

• Generate boilerplate code to build off of

• Itentify bugs within existing code 

• Add comments to existing code for improved readability and collaboration

• Identify edge cases where your code might fail

5. Generate keywords for blog posts or marketing campaigns.

ChatGPT's use of a transformer model (the "T" in ChatGPT) makes it a good tool for keyword research, as it can generate related terms based on context and associations (compared to the more linear approach of more traditional keyword research tools). You can also input a list of keywords and classify them based on search intent.

6. Summarize long documents or explain complex concepts.

ChatGPT can quickly summarize the key points of long articles or sum up complex ideas in a way that's easier to understand. This could be a time saver if you're trying to get up to speed in a new industry or need help with a tricky concept while studying.

5 tips for using ChatGPT responsibly

Take advantage of generative AI tools thoughtfully and responsibly by following these five guidelines:

1. Always review and edit generated text for accuracy and quality.

2. Treat generative AI as a starting point rather than a finished product.

3. Use it for repetitive or time-consuming tasks that don't require creativity or originality.

4. Don't use any sensitive or private information as input data.

5. Leverage it in conjunction with other tools and techniques, including your own creativity, emotional intelligence, and strategic thinking skills.

Learn more about the technology behind ChatGPT

For a broad introduction to the fundamentals of machine learning, consider the Machine Learning Specialization from DeepLearning.AI and Stanford. Get familiar with key concepts in AI as you learn to build and train real-world machine learning models. 

specialization

Machine Learning

#BreakIntoAI with Machine Learning Specialization. Master fundamental AI concepts and develop practical machine learning skills in the beginner-friendly, 3-course program by AI visionary Andrew Ng

4.9

(9,538 ratings)

167,246 already enrolled

BEGINNER level

Learn More

Average time: 3 month(s)

Learn at your own pace

Skills you'll build:

Decision Trees, Artificial Neural Network, Logistic Regression, Recommender Systems, Linear Regression, Regularization to Avoid Overfitting, Gradient Descent, Supervised Learning, Logistic Regression for Classification, Xgboost, Tensorflow, Tree Ensembles, Advice for Model Development, Collaborative Filtering, Unsupervised Learning, Reinforcement Learning, Anomaly Detection