“Artificial Intelligence” is beginning to fundamentally change the ways we work and these changes are happening quickly. Even though the tools will become more powerful and easier to use over time, they will need new skills to use well. Now is the perfect time to begin developing those skills.
AI isn’t new
Artificial Intelligence (or, more specifically machine learning (ML)) is already working invisibly in your daily life: in your phone’s camera to improve your photographs, in text prediction and suggestions when you type, in facial and speech recognition on your devices, in your bank to identify fraud, in search engines, and in automated recommendations. These use classic machine learning techniques to recognize patterns and act on them. They recognize patterns in your actions and respond.
What’s different this time?
The latest crop of systems is based on Generative ML and Large Language Models (LLMs) — these can generate new content including text, images, and in some cases, video. You control them via a written prompt:
Prompt: Explain learning objectives to me in 50 words or less.
Response (Chat-GPT 3.5):
Learning objectives are clear, concise statements that define what learners should know or be able to do after completing a learning experience. They guide instructional design, helping educators focus on specific goals and assess the effectiveness of teaching methods in achieving desired outcomes.
That result isn’t bad, but it’s also not quite specific enough for learning designers. We’ll need to refine the prompt a few times until we get something useful.
Prompt: I'm a professional learning designer. Explain learning objectives to me in 50 words or less.
Response: Learning objectives are precise statements defining desired outcomes of a learning experience. Essential for instructional design, they guide content creation and assessment. Clear objectives align with organizational goals, enhance instructional focus, and enable effective evaluation of learners' mastery of skills or knowledge.
Prompt: What are the elements of an effective learning objective?
Response: An effective learning objective typically includes three key elements:
Behavior (Verb): Describes the observable and measurable action or behavior that learners should demonstrate.
Condition: Specifies the circumstances or context under which the behavior will be performed, providing clarity on the situation.
Criterion: States the standard or level of performance that indicates successful achievement of the objective, outlining what constitutes success.
Notice the changes from one prompt to the next. First, we tell it about the audience (or you could use “You’re a professional learning designer. Explain…”) This doesn’t give specific enough information, so we rephrase the question: “What are the elements of…”?
Prompt engineering in 60 seconds
This refinement practice is called prompt engineering. This applies anywhere, whether you’re generating text, images, and/or sound.?
You can think of a prompt as an instructional objective for the system:
Behavior —? “Explain learning objectives to me”
Condition — “I’m a learning designer” (as opposed to a businessperson, student, etc.)
Criterion — “in 50 words or less”
Then it’s a practice of iteration and refinement. You might rephrase the question, add a bit of information, and ask for a new answer. You could supply samples of writing, specific information about your problem, etc. and this will alter successive results.
How can you get started?
There are a few ways to get started for free:
Google Bard is one of the most advanced models out there and you can use it for free. You get a chat window where you can ask questions and see results. Use the buttons at the bottom to rate the answers (which can improve your results), share the response, and fact-check the answer.?
ChatGPT was the first to release a model with a chat interface and is also free to use (in its more basic form). A “Plus” membership ($20USD/month) gives access to its more advanced model.
If you’re a software developer, check out Github Copilot, the Cursor editor, or Google’s Duet AI. All of these help write code and fix bugs.
Do you read research papers? Take a look at Semantic Scholar. It’s an academic search engine that can also summarize papers, give you an overview of a research field, and set up a curated feed of papers you’ll find useful. (You’ll need an account for the ML-driven features, and there’s a generous starting allowance.)
In my experience, it’s easier to learn with a friend. Team-GPT adds a multi-person chat to the ChatGPT models so you and a friend can experiment together. It provides an excellent free course to help you become an expert. It also has a fantastic library of prompts so you can learn by example. You can get started for free, and it’s generally quite inexpensive (I spend less than $5/month typically).
Let’s begin
As a practical starting point, check out Team-GPT and their free course. Get a feel for the next generation of models with Google Bard. Then consider one of the more specialized systems.