What Is Generative AI? | Meaning & Examples
Since ChatGPT was released in 2022, generative AI has become ubiquitous. It’s an integral part of the Google search experience, and it’s embedded in most of the tools that students and professionals use every day.
Now that generative AI is literally everywhere, understanding how it works is a crucial part of digital literacy, critical thinking, and academic integrity.
Generative AI is an artificial intelligence technology that generates original media, such as text, images, or videos. To generate this media, generative AI tools (e.g., ChatGPT, Gemini, or Davinci AI) need prompts from human users. For example, you can prompt generative AI tools to draft a cover letter or create a song in a particular style.
Types of generative AI tools
Generative AI tools can be classified by the types of content they generate. The table below outlines some common uses for generative AI in 2024.
Content type | Description | Tools |
---|---|---|
Text | Chatbots, text generators, and AI writing tools generate new content based on a prompt from a human. The prompt can ask for an answer to a question, a summary, a translation, or a piece of writing. Chatbots are increasingly integrated into search engines (e.g., Google) to enhance search results. | Gemini, ChatGPT, QuillBot Paraphraser, QuillBot Summarizer, DeepL Translator |
Code | Generative AI can produce code in various programming languages (e.g., JavaScript or Python). Chatbots such as ChatGPT have this functionality, but there are also tools that specialize in code. | Open AI Codex, GitHub Co-Pilot, IBM watsonx Code Assistant, Code WP |
Images | These tools can generate an image based on a prompt from a human (e.g., “Create a picture of a Siamese cat eating pizza”) or alter an image based on the user’s instructions (e.g., “Make my grandma look like a Simpson’s character”). | DALL-E 3, Davinci AI, Canva |
Video | Generative AI tools can also make videos, such as the 2024 holiday commercial for Coca-Cola. | Clip Creator, Pictory, Simplified, Movavi Video Editor |
Audio | These tools create music and speech. The music-focused tools can generate a variety of lyrics, instrumental sounds, and rhythms based on the user’s prompt (e.g., “Write a happy holidays song in the style of disco”). | Suno, LANDR, ElevenLabs, Soundraw |
History of AI
A basic understanding of the history of AI provides insight about how generative AI works. The two main precursors to generative AI—logic programming and machine learning—are still prevalent in a variety of contexts, such as applying for a mortgage and getting movie recommendations from a streaming app.
Logic programming
Logic programming is the principle behind expert systems, which are tools that determine your eligibility for anything from a mortgage to financial aid or even a refund for a defective product. Expert systems were officially introduced at Stanford University in 1965 but became prolific in the 1980s.
Expert systems interpret data and make “decisions” in the same way as human experts (but more consistently and efficiently). For example, each time you check your credit score, a credit bureau’s expert system interprets all of the data that determines your total score (e.g., all of the payments you’ve made since you started your credit history).
Logic programming involves training an expert system with detailed facts and if-then rules from human experts about how each piece of data affects the “decision” or outcome. To teach expert systems how to make accurate or fair decisions, the human experts who program them have to anticipate many possible scenarios and variables.
Machine learning
Machine learning (ML) tools go further than expert systems because they learn from data and make new data based on what they’ve learned. Their “decisions” or outputs aren’t limited to a programmed range of possibilities (e.g., all the possible credit scores). Just like expert systems, ML tools receive initial data from human experts.
One of the earliest versions of ML was a computer program that played checkers, which was developed at IBM in the 1950s. Beginning in 2006 with Netflix recommendations, ML became prevalent in a wide variety of apps and technologies. Now, ML powers facial recognition, fraud detection, video game NPCs, and more.
For example, after you listen to an album or playlist on Spotify, if you don’t choose something else, the app plays similar-sounding songs and artists. It uses your listening history data to predict other songs you’d enjoy out of a vast range of possibilities.
How does generative AI work?
Generative AI combines ML with another AI technology—natural language processing (NLP), which enables computers to analyze parts of speech (e.g., nouns, verbs, and adjectives) and generate human language.
Generative AI models are trained to predict what should come after a prompt based on all of the data it’s seen so far (e.g., all the words that were scraped from the internet to train it).
When a user types a prompt in ChatGPT or another text-based tool, it predicts what word should come next based on the relationships and patterns it’s seen in the text that was used to train it. After that, the tool uses an iterative process to continue generating the next word.
ML and NLP are used simultaneously during this process. ML enables ChatGPT to generate words that are associated with the words in the prompt (similar to how Spotify generates songs similar to the previous play). NLP enables it to interpret how the words in the prompt work together and to generate a grammatically coherent response.
Strengths and limitations of generative AI
Generative AI is rapidly expanding and evolving. To use generative AI effectively, it’s important to understand the strengths and limitations.
Strengths
- Efficiency: Generative AI tools can jumpstart the brainstorming process when you’re experiencing writer’s block. Busy professionals can also quickly generate presentations, memos, and other forms of business writing.
- Accessibility: Generative AI enhances a wide variety of assistive technologies and learning accommodations, such as language translation and text to speech. Educators can use AI to adapt reading passages to students’ reading levels.
- Consistency: ChatGPT and other text-based tools can be used to ensure that workplace documents have a consistent style and structure across different writers.
Limitations
- Accuracy: When an AI receives instructions it hasn’t seen before, it produces hallucinations (made up details that aren’t based on fact). If a factual error is repeated often enough on the internet, it’s likely to appear in AI responses.
- Bias: Generative AI can also generate content with any of the biases (culture, gender, or otherwise) that are present in the training datasets. AI can only follow inclusive language guidelines to the extent that they’re reflected in the initial dataset.
- Reliability: Tools such as ChatGPT usually don’t identify the sources of their information. Therefore, always cross-check any facts that you want to use in writing assignments with credible sources.
Best practices for writing with generative AI
Generative AI can be a useful time-saver for different stages of the writing process. However, it should never be used to create an entire piece of writing from start to finish, which is a form of plagiarism. Educators have become increasingly vigilant with AI detectors—tools that evaluate the originality of a piece of writing.
The following best practices will help you maximize AI’s time-saving potential for different stages of the writing process while maintaining your academic integrity.
Brainstorming and Planning
Before you begin writing, ChatGPT, etc. can generate ideas for body paragraph topics if you provide the assignment description and topic in the prompt. Always cross-check those ideas with credible sources.
For example, if ChatGPT shows you three drawbacks of fracking, use each drawback as an online search phrase, and confirm that scholarly sources or other experts agree.
Also review ideas for different possible angles (e.g., multiple sides of an argument) to make an informed decision about your main idea.
Drafting
During the drafting stage, you might prompt the AI to generate drafts of sentences or paragraphs, but these should never appear wholesale in a finished essay. Instead, review the output for ideas about how you’ll write your own hook, topic sentences, and so forth.
You should always make your own adjustments to any content that the tool generates.
Revising
You might also use ChatGPT or tools like the QuillBot Paraphraser to explore ways to reword sentences or short passages. Avoid prompting these tools to revise an entire essay. Instead, work with short passages and provide clear and specific prompts (e.g., “I’d like you to revise this thesis statement to include formal word choices”).
Editing
The QuillBot Grammar Checker and other AIs can efficiently flag mechanical errors. However, you should also be an active participant in this process. Look for patterns of error and reflect on ways to avoid similar types of errors in future writing assignments. Also cross-check grammar and punctuation suggestions with the style guide you’re using.
Frequently asked questions about generative AI
- What is the difference between AI and generative AI?
-
The difference between AI and generative AI is that traditional AI follows specific rules to perform a task, but generative AI creates new content.
The first type of AI included programs that made decisions based on rules (in the same way as a human expert), such as determining a person’s credit score.
A later development, machine learning AI, classifies or predicts outcomes based on patterns. An example of machine learning is Netflix recommendations that are based on your previous viewing habits.
Generative AI tools combine machine learning with natural language processing technology. They learn from underlying patterns and use that information to “decide” what details to include in a paragraph, image, or other output. Examples of generative AI tools that create new content include ChatGPT and Gemini.
Generative AI tools are useful for brainstorming, prewriting, and paraphrasing, but they should never be used for writing entire assignments.
QuillBot’s free AI Detector can help you ensure that the writing you submit for class assignments is based on your own writing voice and ideas.
- How do you use generative AI for brainstorming and prewriting?
-
To use generative AI for brainstorming and prewriting, choose a generative AI tool (e.g., Gemini or ChatGPT), and type a prompt.
In your prompt, provide a brief description of the writing assignment and topic, and ask the tool to generate ideas for body paragraph topics. Place the description of your writing assignment in curly brackets.
After the generative AI tool responds to your prompt, review the list of body paragraph topics, and select the ones that you’d like to research further.
Use keywords from each item on the list as search phrases in an academic database or a search engine (e.g., smartphones and student distractions).
Then, research multiple ideas from the generative AI response in order to choose a main idea (e.g., your main argument) and body paragraph topics that are based on critical thinking.
When you’re in drafting stages of your writing task, QuillBot’s free Grammar Checker can help you avoid errors. QuillBot’s free Citation Generator can also help you create flawless citations for your outside sources.
QuillBot’s free AI Detector can help you ensure that the writing you submit for class assignments is based on your own writing voice and ideas.
- Can AI write a book?
-
Yes, AI can write a book. However, a book written by AI probably wouldn’t be a very good book.
Unlike humans, AI is limited when it comes to creativity (e.g., with figurative language), themes, and tone of voice.
Despite this, AI can be helpful to human authors. For example, AI can suggest writing strategies to put writers in the right mindset.
Many publishers also use AI detectors to check that what they publish is high-quality, human-written work.
If you’d like to see how AI detectors work, you can try QuillBot’s free AI Detector.
- How can you check if an image is AI generated?
-
There are a few ways you can check if an image is AI-generated.
First, review the image for anything that doesn’t look quite right. AI-generated images often have distorted text, patterns, or human features (especially faces and hands).
Second, check for metadata information. Some AI image generators use specific filenames or imprint a watermark on their images.
Third, understand how AI detectors work and how you can use them to analyze the probability that an image was generated by AI.
And if you need help detecting texts generated by AI, QuillBot’s free AI Content Detector is one option.