What is generative AI?
You may have heard about AI in the media this year.
AI is an abbreviation for “artificial intelligence”. We use artificial intelligence all the time in our daily lives. If you open your phone with Face ID, access content on a social media or streaming app feed, use spell check, get frustrated with an ad seeming to follow you around, use a smart home device, or the maps app to check the traffic or route you need to take, you’ve used AI.
The kind of AI that has captured the popular imagination this year is called generative AI. This is AI that creates something new: new text, new pictures, new code, or new video. Programs mentioned in the popular press include ChatGPT, Bing AI, DALL-E, Midjourney, Lensa, and Bard.
The big change that happened in early 2023 was the easy availability of a type of generative AI called a Large Language Model. A Large Language Model is a type of algorithm that uses a lot of data – this could be all of Wikipedia, or all of a large code database like Stack Overflow. When asked to generate content, the model uses the data to predict what it should generate; what is most likely to be an acceptable response. AI companies work to finesse these models so that they give a good answer most of the time. Open AI, who made ChatGPT, liken the process to ‘training a dog’.
Large Language models are why people are worried about plagiarism. AI like ChatGPT and Bing AI can be used to generate sensible, literate answers to questions, and because these answers are generated, not copied, it is very hard for teachers – and even for automated anti-plagiarism tools -- to “pick up” when students are presenting AI generated content as their own.
What can Generative AI Produce?
Please note: Generative AI is evolving rapidly, and many products offer more than one of these functions.
Text
Outputs writing in response to a prompt. Can analyse text or images to look for patterns or offer advice. Can be asked to output text in different styles and voices, and to incorporate references and quotes. Can be asked to “improve” text for the user and will do so.
Images
Outputs an image in response to a prompt. Can work in a variety of styles, add, or remove elements from an image, including expanding on an existing image. Referred to as a ‘deepfake’ when a person’s image is realistically mapped into a situation that they have never been in.
Analysis and Code
Outputs analysis of data (e.g., graphs, qualitative analysis of words) and/or computer code. This analysis can include the analysis of a person’s study or work habits to aid self-reflection, or the analysis of a piece of work and advice for improvement. This analysis may also be of large groups of data such as student data or workforce data, such as resume-ranking programs.
Design
Outputs visual design such as layout, such as automatically arranging content, choosing colour and font schemes, creating images. A common addition to slideshow or graphic design programs.
Video and Audio
Outputs a video and/or audio clip. This may be entirely AI generated, generated based on an image or series of images, or a ‘deepfake’ where a person’s image or voice is mapped onto a real or digital clip that shows them realistically doing or saying something that has never occurred. Music can also be generated via AI.
Organisation
Emergent technology that is being woven into multimodal tools that allows an AI to offer organisation assistance such as automatically adding follow-up meetings/to-dos to a person’s calendar based on what is said in a meeting, automatically write summaries or minutes of meetings, or undertaking administrative tasks such as planning and booking a travel itinerary.
How can Generative AI be used in schools?
Not all uses of AI in school contexts will be permitted by all sectors and/or schools. Please note that the Board of Senior Secondary Studies (BSSS) does not provide technology to schools and students, nor does it set policies about the acceptable use of technology in sectors and schools.
Teachers can use generative AI to:
Generate lesson and curriculum plans
Generate ideas for innovative activities and assessments
Analyse curriculum materials as part of constructing an assessment
Create exemplars, datasets, examples, arguments and other lesson-specific texts
Visualise ideas for students
Quickly format or transform resources
Differentiate class materials
Generate polished text from a broad/draft description
Students can use generative AI to:
Get specific, personalised feedback
Differentiate instructions or explanations, including making documents accessible for students with disability
Format or transform assessment work (e.g., slideshows for a speech)
Visualise and ideate (e.g., test concepts in a design class)
Summarise articles, texts, and videos
Change the lexical density of text for ease of understanding
Converse with an AI to finesse understanding
Brainstorm or discuss ideas for a specific task
Create a scaffold for a specific task
Create a study plan or calendar
Generate study advice and hints
Activities
Watch
Watch this short documentary by Vox media. It discusses how AI art is generated, and provides an introduction to the way generative AI works, and introduces terms such as 'latent space'.
Generate
Use an AI image generator, and prompts of your choice, to generate an AI image. Please note that the examples below were selected because they are easy to access without logging in, or use logins people may already have.
Dall-E III
If you have a paid account for ChatGPT, you can ask ChatGPT to generate an image.
If you do not wish to sign up to a generator but you are okay with your Microsoft ID being associated with the images, you can use Bing's image generator: https://www.bing.com/images/create
If you use Canva, use the "Magic Media" function in the left-hand menu. You may need to scroll down
Dall-E II
If you signed up to ChatGPT, you can use OpenAI's Dall-E with the same account: https://openai.com/product/dall-e-2
Other
If you do not wish to sign up to a generator at all, DeepAI's image generator is free to use for a limited number of styles: https://deepai.org/machine-learning-model/text2img
Paste a copy of your image and prompt into your workbook file.
Dall-E II generated with the prompt: " realistic oil painting landscape by Raphael"
Reflect
Read or view one or more of the articles linked below. There is no expectation that you will interact with all of them. These articles represent a range of points of view -- some positive, some negative.
I would like to learn more about ChatGPT in general: https://www.youtube.com/watch?v=BWCCPy7Rg-s
I would like to read about generative AI and the arts: https://theconversation.com/chatgpt-dall-e-2-and-the-collapse-of-the-creative-process-196461
I would like to read about generative AI and the sciences: https://www.cigionline.org/articles/chatgpt-strikes-at-the-heart-of-the-scientific-world-view/
I would like to read about generative AI and linguistics/philosophy: https://www.nytimes.com/2023/03/08/opinion/noam-chomsky-chatgpt-ai.html
I would like to read about how people are using generative AI in their jobs now: https://www.abc.net.au/news/science/2023-01-25/chatgpt-midjourney-generative-ai-and-future-of-work/101882580
I would like to learn about how theorists believe AI could positively affect businesses: https://www.ted.com/talks/andrew_ng_how_ai_could_empower_any_business
I would like to watch a video about how to use ChatGPT for learning English (aimed at students of English as an Additional Language): https://www.youtube.com/watch?v=-IBxIXBaei8
Now write a short reflection in your workbook file without using AI; how do you see this paradigm shift interacting with your teaching, or with the academic disciplines that you work within?