September 15 :: Teach A Machine

WOVEN
AI & Digital Literacy
Session 1 of 4  ยท  Teach a Machine
UHAA  ยท  United Harbor Area Association  ยท  Work in pairs, one laptop per pair
โš“ Career Spotlight
AI and Data Technician

The harbor's automated terminals use AI cameras to recognize every container and keep the yard running without crashes. AI and data technicians are the people who train and check those systems. Today you do the same thing they do: teach a machine to recognize what it sees. You can start training for this work right here at LA Harbor College in Wilmington.

๐ŸŽฌ Watch first  ยท  What is machine learning?

A quick look at how machines learn from examples. You are about to do exactly this.

๐ŸŽฏ Today's Mission

Train your own AI to tell three things apart, then find a way to fool it.

By the end you can
  • Build and train an AI model with your own examples
  • Test it and read its confidence
  • Explain why the examples you pick decide how smart it is
โ˜…
๐Ÿค– Main Challenge
Teachable Machine
  1. Go to teachablemachine.withgoogle.com. Click Get Started, then Image Project, then Standard image model. No login needed.
  2. Make 3 classes. Pick three things to tell apart. Rock, paper, scissors works great.
  3. For each class, hold the object to the webcam and press Hold to Record for about 15 seconds. Move it around a little.
  4. Click Train Model. Wait a few seconds.
  5. Test it. Show the webcam each thing. Watch the guess and the confidence bar.
  6. Break it. Show it something you never trained, your face or a pen. It still guesses confidently. That is the whole lesson: AI only knows what you showed it.
โญ Finished early? Level up

Quick wins:

  • Add a 4th class, or add more examples in different lighting and angles. Watch accuracy change.

Or pick one bigger project. All stay inside Teachable Machine:

  • Flawed on purpose. Train a model with only close-ups of one object, then show it a different one. Watch it get it wrong with full confidence. That is biased data, exactly what we dig into in Session 3.
  • Full-body controller. Start a new Pose Project. Train poses like jump, lean left, and lean right, then watch it track your body live. Ask a teacher to help wire it into a Scratch game to go further.
  • Posture helper. Pose Project. Train "sitting up straight" and "slouching," then make the screen flash a reminder when you slouch. A real assistive-tech idea.
  • Gesture messages. Train a few hand signs, each with its own message. Make the sign, the message appears. This is how communication tools for people with limited mobility begin.
๐Ÿ—ฃ๏ธ Quick Share
  • One thing your model got right.
  • One thing that fooled it, and why.
๐Ÿ Wrap Up

You just trained real machine learning. The examples you chose were the skill, not the buttons. Before you go: close your tabs and return the laptop.

Next time: when AI gets it wrong. You will spot deepfakes and fake media, and learn the quick checks that catch them.

๐Ÿงญ Career Connections
AI and data jobs are coming to the harbor. Here is how to get one.
๐Ÿ”Ž Find it
  • Your UHAA mentors are your first stop. They can connect you to local tech and career programs.
  • LA Harbor College in Wilmington has tech and data pathways, close to home.
๐Ÿ“ Apply
  • Put this workshop on your resume: trained and tested a machine learning model.
  • Entry level means they train you. Apply anyway.
๐ŸŽค Interview
  • Tell the story of how you fooled your model and what you learned.
  • Lead with curiosity and willingness to learn.
โญ Your first stop: UHAA

UHAA is here to help you take the next step toward these careers, from building a resume to making real local connections. Ask your mentors, that is what they are here for.

Built by Woven Learning & Technology