Creating a blog post documenting and reflecting on the following exercises from A People’s Guide to AI.

• When you hear the words "Artificial Intelligence", what are the first four things that come to your mind? (p.11)

1. Learning from data the way machines pick up patterns, like how Netflix guesses your next show.
2. Automation tasks getting handled without us, from spam filters to self-driving cars.
3. Bias and fairness the risks of algorithms reinforcing inequality if they’re built on messy or unfair data.
4. Human creativity the mix of wonder and unease when machines generate art, music, or writing that feels almost human.

• Answer the questions from the "Everyday AI Activity" on pages 23-28 of the

p23:

1. Smartphone keyboard predictive text and autocorrect are powered by AI.
2. Social media apps Instagram or TikTok recommend posts and videos based on what I interact with.
3. Search engines Google uses AI to rank results, finish queries, and even answer questions directly.

p24:

One afternoon I opened Instagram and it showed me a carousel of design portfolio tips right after I had been sketching ideas for my own website. I hadn’t searched for anything, so it felt like the app somehow read my mind. It surprised me how closely it matched what I was working on in real life.

p25:

p26:

1. Convenience & Efficiency AI makes tasks faster (e.g.spam filters sort email) Personalization services like Netflix or Spotify recommend content that fits our tastes, saving us time searching. Accessibility translation tools, speech recognition, and text-to-speech make communication easier across languages and for people with disabilities. Safety & Accuracy fraud detection in banking, handwriting recognition in check deposits, or self-driving car features reduce errors and risks. Expanded Knowledge AI-powered search engines and assistants give us quick access to massive amounts of information.
2. Privacy AI systems often track our habits, preferences, and locations to work well, which can feel intrusive. Control automated decisions (like what news or posts we see) can shape our choices without us realizing it. Fairness biased data can lead to unfair outcomes (e.g., facial recognition misidentifying people of color, or predictive policing targeting certain communities). Human Skills relying too much on AI (like spell check or GPS) may weaken our memory, critical thinking, or problem-solving abilities. Jobs & Roles automation can replace certain types of work, making employment less stable in some industries. Serendipity personalization can trap us in “filter bubbles,” where we only see what algorithms think we’ll like, not the unexpected or new.

p27:

p28:

1. What it looks like
   • Is it an app, a website, a device, or something else?
   • Describe its appearance in simple terms (e.g., a phone app with a dashboard, or a wearable wristband with a small screen).
2. Where it lives
   • On a smartphone? Smartwatch? Computer? Physical device in your home?
3. Inputs
   • What kind of data it collects (calendar info, mood check-ins, activity levels, etc.).
4. AI process
   • What the AI does with that data (pattern recognition, recommendations, predictions).
5. Outputs
   • What the user sees/receives (reminders, tips, recommendations, alerts).
6. Human role
   • How people interact with it (choose, accept, adapt, or ignore suggestions).

Using the code examples above, try running image classification on a variety of images.

• What does the model recognize properly? What does it not recognize? What other aspects of the image affect the classification, including but not limited to position, scale, lighting, etc.
• You are welcome to use the provided examples as they are or modify the code to create interactive experience that reflects your creativity and curiosity!

What the model recognizes properly:

Common, well-lit, centered objects with distinct features (e.g., robin, banana).

What the model struggles with:

Low-contrast objects against similar backgrounds (coffee mug).

Abstract or simplified drawings not strongly represented in training data (heart misclassified).

Other influencing factors:

Position & Scale: Larger, centered objects are classified more accurately.

Lighting: Even lighting improves accuracy; poor lighting reduces confidence.

Background: Clean backgrounds improve accuracy, cluttered or similar-colored backgrounds reduce it.

Hand-Drawn Heart

Result: hurricane (Confidence: 0.11)

Observation: DoodleNet misclassified the heart sketch as “hurricane.” This shows the limits of the doodle-based model, which sometimes confuses simple shapes with unrelated categories.

Abstract or ambiguous drawings lead to misclassification, especially when they resemble shapes in other categories.

Banana

Result: banana “You found banana!”

Observation: The model recognized the banana correctly in a real-time webcam scavenger hunt game. The clean white background and good lighting supported accurate classification.

The model works reliably for common fruits in ideal conditions and can be incorporated into interactive applications.