Ever wonder how your favorite streaming service reads your mind and suggests the perfect movie? Or how your phone’s camera can spot your dog in a photo full of cats? That’s not magic, it’s Artificial Intelligence! But ‘AI’ can feel like a complicated, technical term.
Watch
- What is Machine Learning (2 min.)
- Generative AI explained in 2 minutes (2 min.)
To make sense of it, let’s use a metaphor: A futuristic kitchen:
As you can see from the image, each concept in AI builds upon the last. (click on triangle for more details)
The Kitchen (AI) is the whole field, everything related to making machines smart.
AI is the broad field focused on building systems that can perform tasks that typically require human intelligence like recognizing speech, making decisions, or generating content.
The Learning Styles (Machine Learning) are the different ways the kitchen can learn. It can follow a labeled Cookbook (Supervised), organize a messy Pantry on its own (Unsupervised), or learn by trial-and-error like a Robot Chef (Reinforcement).
Machine Learning is one of the main ways we achieve AI. Instead of programmers writing explicit, step-by-step instructions for every single possible situation, ML systems are “trained” to learn from large amounts of data. They identify patterns in the data and use those patterns to make predictions or decisions about new, unseen data. Think about Netflix recommendations or facial recognition on your phone.
The Chef’s Intuition (Deep Learning) is a more advanced technique, using complex ‘neural networks’ to figure out recipes from massive amounts of information, like just looking at a photo of a dish.
Deep Learning is a type of machine learning that uses layered neural networks to process large amounts of data and it’s especially good at handling complex patterns in images, text, or speech
The Food Replicator (Generative AI) is the most revolutionary appliance. Instead of just identifying food, it creates brand new dishes from scratch based on your instructions.
Generative AI refers to AI systems (usually based on deep learning) that can create new content like text, images, music, or video by learning from examples and generating outputs that resemble them.
LLMs are a specific kind of generative AI trained on massive text datasets to predict and generate language in a human-like way. They’re the engines behind tools like ChatGPT and Claude.
Activity
But first, let’s acknowledge we’re using metaphors, like chefs, kitchens, and cookbooks, to help explain AI. Even the title of this lesson “What’s Under the Hood?” is a metaphor. That’s intentional. Generative AI is complex and abstract, and metaphors help us connect abstract concepts to things we already understand.
But here’s the catch:
Metaphors do more than illustrate. They can also frame how we perceive systems, what kinds of questions we ask about them, and what we assume they can (or should) do.
Saying “AI understands you” or “the model is a chef” or that we can “see under the hood” can be helpful, but it can also be misleading. These systems don’t understand, plan, or think the way people do. And they aren’t mechanical like a car engine, they are mathematical. They generate responses by predicting what’s likely to come next based on patterns in massive datasets, not lived experience or reasoning.
When it comes to generative AI discourse, there’s a fair amount of metaphor that can shape our expectations about it. Stay curious and ask:
“What does this metaphor help me see more clearly?”
“And what might it oversimplify or obscure?”
- Visit ChatGPT. If you already have an account, great, but you can use ChatGPT for this activity without signing up. Type the following prompts in the “ask anything” text box:
- Prompt 1: “Give me a recipe for a dish that symbolizes resilience”
- Prompt 2: “What would the same recipe look like if it was written by a homesick astronaut?”
- Prompt 3: “Now rewrite that as if the dish is a conversation between ingredients”
Discussion
After using ChatGPT, how did your expectations going in shape the way you interpreted its response? Did the output feel better or worse than you anticipated? Were you surprised, disappointed, impressed? Why?
14 replies on “thing 2: AI 101: What’s Under the Hood? ”
This was fun! I did enjoy the outcomes, especially since I didn’t know what to expect. The last prompt turned out to be really enjoyable especially with the ‘dialog’ between the ingredients also noting their emotions as they ‘spoke’. It is interesting to see how each outcome resulted from the prompts given.
Interesting and impressive!
Wow! The ChatGPT exercise was crazy. Very creative. I could not have come up with anything like that. I wonder if ChatGPT’s responses were the same for each of us when using identical prompts? If not, were they similar?
I’m not sure what I expected with these prompts. The first prompt definitely gave me something that was easy to read and usable. The second and third prompt seemed a little ridiculous to me, and I was surprised at how fast I got results and how thorough they were. I’ve never considered having ingredients in a recipe have a conversation and I’m not sure that I need to! But I can see this being a valuable creative tool.
When I first read the prompt I thought it would give me a recipe for a food that is hard to damage or can survive in different conditions. What I was actually given was a recipe for “Three Sisters Stew”, which is created with three plants that are planted together and help each other grow. I thought it was an interesting interpretation of “resilience”. I was impressed the second prompt changed all of the measurements from tablespoons and cups to packets, cubes, and pouches like they would actually have in space.
I was surprised by how the language returned by these prompts reflected patterns I see most in quickly produced internet content. The second prompt returned a script that read like a TikTok video, and the third used parenthetical adverbs/descriptions in a way that was popularized by instant messaging conversations. The content read clean and tight, but shallow.
This exercise was very interesting. It seems we all got different recipes (mine was Braised Lentils with Roasted Root Vegetables & Herbed Yogurt) based on our past search history possibly? The reasoning for the recipe was solid. After the second prompt it turned into “Starboard Lentil Stew with Roasted Root Veg & Zero-G Yogurt Swirl
Captain’s comfort food log, Sol Cycle 84” – which was super fun with the comments added in and language changes in the preparation guide. The third prompt (A Conversation in the Pot: The Story of Resilience) made this just hilarious and entertaining!
This exercise highlights the creative abilities of generative AI. I did not expect the responses to be so descriptive and nuanced. From that vantage point, the responses impressed me. Makes me wonder if any of us will actually create one of these concoctions.
It wouldn’t let me use ChatGPT in a private window without logging in, so I had to use a regular browser window. It gave me a pretty solid beef stew recipe, and then had a section after the recipe where it went through the symbolism for the choices. It generated a lot of text for all three answers very quickly. It asked for feedback each time in a whimsical way, which matched the whimsy of the prompts. It wouldn’t be how I want to look for a recipe, but could be fun for idea generation.
WOW! impressive, the third prompt was amazing, I really enjoyed the poem at the end. Even the second prompt was great. It is a little scary but interesting and fun!
Since I am not unfamiliar with ChatGPT, these responses were very much in line with the expectations I had going into the activity. While the prompts are not the typical things that I would refer to the system for, the length and floweriness of the responses, especially the first two, were consistent in style with responses I am familiar with ChatGPT giving out. I was a little surprised by the brevity of the last response, as mine gave me little blurbs from each ingredient that was more like mini monologues rather than a conversation amongst them. I am curious about what made it give us different initial recipes (if using it as a guest instead of in an account). Mine was “Braised Lentils with Roasted Root Vegetables.”
That was a fun exercise. The answers were creative and interesting. I wasn’t expecting to receive something so imaginative from an AI program.
I have used ChatGPT before, but not in such a creative sense, more for practical means. I first prompt response was about what I expected. I thought it was interesting how it could change the recipe’s backstory based on perspective and even make the ingredients converse with each other. Very creative!
The Chat GPT replies were about at the level I expected. “What under the hood” topic title was somewhat misleading. I don’t think we really learned what’s under the hood. 🙂