• Write drafts of anything – blog posts, essays, promotional material, lectures, scripts, short stories. All you have to do is prompt it. Basic prompts result in boring writing,but getting better at prompting is not that hard. AI systems are more capable as writers with a little practice and user feedback. 

• Make your writing better. You can paste your text into an AI and ask it to improve the content, check for grammar and improve paragraphing. Or ask for suggestions about how to make it better for a particular audience. Ask it to create five drafts in radically different styles. Ask it to make things more vivid or add examples.  

• Help you with tasks. AI can do things you don’t have the time to do. Use it to write summaries, create project templates, take meeting notes, and a lot more. Later in this module, you’ll have a chance to try out using an AI tool to help you complete a work task. 

• Unblock yourself. It’s very easy to get distracted from a task when you get stuck. AI can provide a way of giving yourself momentum. Ask it for ideas to help you get started. You often need to have a lot of ideas to have good ideas, and AI is good at volume. With the right prompting, you can also get it to be very creative. Or you can ask it for possible next steps in a project or a work schedule to keep you organized. The key is dialog. 

AI tools are being integrated directly into common office applications. Microsoft 365 applications now have the option to include an AI-powered “Co-pilot” to assist with various tasks as you work within documents, and Gemini is being integrated into Google’s G Suite applications. The implications of what these new innovations mean for writing are pretty profound. McMaster is currently not planning to activate Copilot for Microsoft 365 as it comes with a significant cost and the data security risks are still unknown.  

There are four big image generators most people use: 

1. Stable Diffusion: is open source and can be run from any high-end computer. It takes effort to get started, since you have to learn to craft prompts properly, but once you do it can produce great results. It is especially good at combining AI with images from other sources. Here is a guide to using Stable Diffusion (be sure to read both parts 1 and part 2).
2. DALL-E: is incorporated into Copilot (in creative mode) and Copilot image creator. This system is solid, but not as good as Midjourney.
3. Midjourney: is the best system as of mid-2023. It has the lowest learning-curve: just type in “thing-you-want-to-see –v 5.2” (the –v 5.2 at the end is important, it uses the latest model) and you get a great result. Midjourney requires Discord. Here is a guide to using Discord.
4. Adobe Firefly: is built into a variety of Adobe products, but it lags behind DALL-E and Midjourney in terms of quality. However, while the other two models have been unclear about the source images that they used to train their AIs, Adobe has declared that it is only using images it has the right to use. One of the major benefits of Firefly is generative fill – you can use it while editing an image in Photoshop to add something to or alter that image based on your prompting. 

Here are the first images that were created by each model when provided with the prompt: “Fashion photoshoot of sneakers inspired by Van Gogh” (each image is labelled with the AI model)  

An AI video generator is a web-based or standalone software that allows you to easily create video assets without needing prior video editing experience. These tools can assist with tasks like erasing video elements, creating green screens, using text to video to construct scripts from a URL or blog post, and more. It is now easy to generate a video with a completely AI generated character, reading a completely AI-written script, talking in an AI-made voice, animated by AI. It can also deepfake people. Runway v2 was the first commercially available text-to-video tool and is a useful demonstration of what is to come.

You can use Copilot to analyze or summarize documents or images. You can do this by attaching images or files using the web browser interface. Or you can open a PDF using the Microsoft Edge browser (right-click on a PDF and hover over the Open with option > choose Microsoft Edge) and click the Copilot icon in the top right corner of the browser to bring up Copilot in the sidebar. Then tell the tool what you’d like it to do (e.g., generate a summary). You can get more creative with your prompts to extract what you need. For example, summarize the key points from this document; give a 3-sentence summary highlighting the most important information in this report; extract a bulleted list of the 5 main takeaways from these meeting notes; describe the text that appears in the image. Using Copilot with your McMaster Microsoft license means that the data is not used to train large language models, so make sure you are signed in with your MacID@mcmaster.ca email and password.

Code Interpreter is a mode of GPT-4 that lets you upload files to the AI, allows the AI to write and run code, and lets you download the results provided by the AI. It can be used to execute programs, run data analysis, and create all sorts of files, web pages, etc. Though there has been a lot of debate since its release about the risks associated with untrained people using it for analysis, many experts testing Code Interpreter are impressed, one paper even suggesting it will require changing the way we train data scientists.

Claude 3 is excellent for working with text, especially PDFs. It’s possible to post entire books into the tool. You can also give it several complex academic articles and ask it to summarize the results, with reasonable results! You can then interrogate the material by asking follow-up questions: what is the evidence for that approach? What do the authors conclude? And so on.

Similarly, Gemini 1.5 Pro has a 128K-token context window. A limited group of developers and enterprise customers can try it with a context window of up to 1 million tokens, but this is a computationally intensive process that requires further optimizations.

It’s currently not recommended to use AI as a search engine. The risk of hallucination is high (an explanation of hallucinations is provided in the “What’s the catch” tab of this module). However, there is some evidence that AI can provide more useful answers than search when used carefully, according to a recent pilot study. Especially in cases where search engines aren’t very good, like tech support, deciding where to eat, or getting advice, Copilot is often better than Google as a starting point. This is an area that is evolving rapidly, but you should be careful about these uses for now.

What’s more exciting is the possibility of using AI to help us learn.You can ask the AI to explain concepts and get very good results. You can use structured prompts to work with AI as an automated tutor. Because we know the AI could be hallucinating, you would be wise to double-check any critical data against another source.

Clark, P. (2023, May 23). Dream bigger: Get started with Generative Fill. Adobe Blog. https://blog.adobe.com/en/publish/2023/05/23/future-of-photoshop-powered-by-adobe-firefly

Ethan Mollick [@emollick]. (2023a, April 5). There are big categories of common problems that, in retrospect, were never good applications for Google search. Bing AI, even with occasional inaccuracies, is just better for things like: tech support, Deciding what to do/where to eat How-to advice, Getting started advice https://t.co/9gIBxq86It [Tweet]. Twitter. https://twitter.com/emollick/status/1643718474668097538

Ethan Mollick [@emollick]. (2023b, June 15). There is a lot of excitement for AI to be a universal tutor. And it shows real promise, but there are some important problems that need to be solved. To get a sense of how good it is, try this prompt (in GPT-4): Https://chat.openai.com/share/ec1018ec-1d86-4160-b587-354253c7d5cb More in our paper: Https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4475995 https://t.co/X8kpg08DEr [Tweet]. Twitter. https://twitter.com/emollick/status/1669434927761313807

Ethan Mollick [@emollick]. (2023c, July 11). Every field of professional education needs to be working on a paper like this right now. This one tests Code Interpreter’s ability to do data science (90% on exams, the field is “on the verge of a paradigm shift”) Then it suggests how to change training. Https://arxiv.org/pdf/2307.02792v2.pdf https://t.co/OnUk22ZZ06 [Tweet]. Twitter. https://twitter.com/emollick/status/1678615507128164354

Gartenbert, C. (2024, February 16). What is a long context window? Google Blog. https://blog.google/technology/ai/long-context-window-ai-models.

Gunnell, M. (2022, April 11). How to use Discord: A beginner’s guide. PCWorld. https://www.pcworld.com/article/540080/how-to-use-discord-a-beginners-guide.html

Mollick, E. (2023a, September 16). A quick and sobering guide to cloning yourself. One Useful Thing. https://www.oneusefulthing.org/p/a-quick-and-sobering-guide-to-cloning

Mollick, E. (2023b, September 16). How to Use AI to Do Stuff: An Opinionated Guide. One Useful Thing. https://www.oneusefulthing.org/p/how-to-use-ai-to-do-stuff-an-opinionated?utm_medium=reader2

Mollick, E. (2023c, September 16). On-boarding your AI Intern. One Useful Thing. https://www.oneusefulthing.org/p/on-boarding-your-ai-intern

Mollick, E. (2023e, September 16). Setting time on fire and the temptation of The Button. One Useful Thing. https://www.oneusefulthing.org/p/setting-time-on-fire-and-the-temptation 

Mollick, E. (2023f, September 16). What AI can do with a toolbox… Getting started with Code Interpreter [Now called Advanced Data Analytics]. One Useful Thing. https://www.oneusefulthing.org/p/what-ai-can-do-with-a-toolbox-getting 

Mollick, E. (2023g, September 16). What happens when AI reads a book ??. One Useful Thing. https://www.oneusefulthing.org/p/what-happens-when-ai-reads-a-book 

Mollick, E. & L. Mollick. (2024). Student Exercises. More Useful Things: AI Resources. https://www.moreusefulthings.com/student-exercises  

OpenAI. (2023). ChatGPT (GPT-4) Friendly Tutor Explains Concepts.  [Large language model].              https://chat.openai.com/share/ec1018ec-1d86-4160-b587-354253c7d5cb.  

OpenAI. (2024). Data Controls FAQ. https://help.openai.com/en/articles/7730893-data-controls-faq  

Prateek K. Keshari [@prkeshari]. (2023, July 9). 20 mins and 3 prompts later, ChatGPT code interpreter gives me 2 branded downloadable html files. Result ? https://t.co/NPMrW72g2A [Tweet]. Twitter. https://twitter.com/prkeshari/status/1678155933606637568 

Stokes, J. (2022, September 29). Stable Diffusion 2.0 & 2.1: An Overview. Johnstokes.Com. https://www.jonstokes.com/p/stable-diffusion-20-and-21-an-overview 

 Xu, R., Feng, Y., & Chen, H. (2023). ChatGPT vs. Google: A Comparative Study of Search Performance and User Experience (arXiv:2307.01135). arXiv. https://doi.org/10.48550/arXiv.2307.01135