Turbocharge your project

Based Learning with ai

Welcome, Educators!

​

​

Who do we have in the room?

​

-Name

-Role

-What does artificial intelligence make you think of?

Empower underserved groups (especially girls, women) to solve problems in their communities using technology

Diversity in AI/CS

• Women and minorities are underrepresented in CS/IT/AI/technology
• Why it matters:
• Diverse engineering teams create better (and less biased!) products
• Economic equity: access to stable, high-paying jobs
• Social justice: underrepresented groups are systemically discouraged from full participation
• Economic need: underproduction of workers is a national competitive disadvantage

​

CS and AI are everywhere.

​

When you think STEM, think CS.

Credit: Marie DesJardins

When You Think CS, Think AI, Machine Learning, Data Science

Credit: Marie DesJardins

Agenda

Introduction

to AI

What ai technologies do you use?

Did you think of these?

AI has three basic parts:

Inputs - dataset

Finds patterns with learning algorithm

prediction!

Source: MIT Media Lab

numbers

​

​

​

sound

​

​

pictures

​

text

Banana

Flour

Ice cream

Yogurt

Apple

Avocado

Black beans

Lentils

17

25”

3.444

19%

​

Finds patterns with learning algorithm

To make predictions

Data can be…

(inputs)

What kind of data does your household create each day via technology?

Every Google search, words you type into emails

​

Every question you ask Alexa

​

Connected devices - when you turn on lights, Air conditioner

​

Taps you make on your cell phone

​

Anything you purchase online

​

Who you are connected to on social sites

​

Songs you listen to

​

Steps you take

​

Restaurants you look up

Data you generate...

Example: instagram ai predicts which advertisement you’d like to see

Inputs - dataset

• Past clicks
• Brands you follow
• Words in captions

Finds patterns with learning algorithm

Predicts what advertisement you might click on

Source: MIT Media Lab

Does AI stop at just making a prediction?

Inputs - dataset

Finds patterns with learning algorithm

prediction

Source: MIT Media Lab

AI technologies use predictions to do things

Uses data

(inputs)

Finds patterns with learning algorithm

To make predictions

Source: MIT Media Lab

Actions or decisions

Example: instagram ai decides which advertisement to show you

Inputs - dataset

• Past clicks
• Brands you follow
• Text in captions

Finds patterns with learning algorithm

Predicts what advertisement you might click on

Source: MIT Media Lab

Action - Shows you a 👟 ad from your favorite store

You try: youtube AI...

What inputs or data would AI consider?

Finds patterns with learning algorithm

What is it gathering data to predict?

What actions or decisions can it make?

Train an AI model

To find patterns

​

machinelearningforkids.co.uk

Click: Get started > Try it now > Copy template > UK Newspaper headlines

Click: Learn and test > Train new model

​

1. Visit those websites, get new headlines and test if the model works

Consider: how could we make the model more confident?

to make predictions

Inputs - dataset

Finds patterns with learning algorithm

prediction!

Actions or decisions

You try: change the Newspaper model

1) How can we get this model to make correct decisions with more accuracy?

​

Let’s do that.

(back to projects, Train)

2) It finds patterns between 4 newspapers.

​

Let’s change the pattern by adding in another newspaper.

Retrain it and test.

​

How have we changed its predictions?

Inputs - dataset

Finds patterns with learning algorithm

prediction!

Actions or decisions

How can we tell if something uses ai?

This alarm clock is set to go off every morning at 6:30am.

​

No AI

Google Maps needs data to predict the best route since it can’t program every route everywhere and uses changing starting/ending points and changing traffic or rail conditions.

​

AI

Choose your side

Directions

#2 uses AI

#1 uses AI

Which tech uses AI? Choose your side of the room.

Roomba

Vacuum

Hyundai Sonata

Remote control cars

Spell checker

Email text predictor

microwave

Shrimp, jelly, sausage pizza

Warning: This is a trick question ;)

Prosthetic - sensing objects

Prosthetic - responds to user

Big Takeaway: AI technologies make predictions to do things

Uses data

(inputs)

Finds patterns with learning algorithm

To make predictions

Source: MIT Media Lab

Actions or decisions

Agenda

Explore

Data

AI uses different types of data to make different predictions

These predictions can lead to different actions or decisions

Your challenge: choose an AI technology that interests you

Answer these questions (you can do research!)

​

• What types of data is it using?
• What patterns might it be programmed to find in the data?
• What predictions is it making?
• What actions or decisions does it make based on its predictions?

Data Exploration with tourist tool

Technology with AI can be “narrow” still: It only is good at what it has learned.

Let’s train a model to predict what a tourist should do when visiting Detroit:

OR

​

Accounts

🤔 How could you make the dataset better?

Bias in Machine Learning

• Gender bias: speech recognition,

Google Translate

• Racial bias: Amazon face recognition, credit modeling
• “[The] algorithm used a seemingly race-blind metric: how much patients would cost the health-care system in the future. But cost isn’t a race-neutral measure of health-care need.”
• Collection bias: Disastrous chatbot: Tay
• Design/algorithmic bias: HireVue
• Dangers of proprietary, unvalidated ML

Credit: Marie DesJardins

Brainstorm: find a meaningful

problem

ICebreaker

Take as many candies as you “need.”

Icebreaker

For every candy, think of 1 problem - a time you have recently felt frustrated, or angry, or thought about something you would like to improve.

Why is this important to you? What value does it have for you?

Icebreaker

Now take turns: Share with your small group two of your ideas that you’re most interested in or really care about.

One person should record each idea on a sticky note.

Choosing a problem to work on

Now your group has several problems brainstormed.

Let’s see if they point to any problem you would be interested in working on!

Put your idea on hold.

Do these categories help you think of any other related problems or angles?

Some innovation categories

Time to think about solutions! Solutions can:

• Improve something that already exists
• Reduce the cost of something that already exists
• Raise awareness about a problem and help to cause changes in people’s behavior
• Apply an existing approach to a new situation, or even
• Invent a completely new solution, technology, or way of doing things

Consider how your problem could be solved with AI

Your brainstormed problem and solution

Answer these questions about your possible solution

​

• What types of data will it use?
• What patterns might it be programmed to find in the data?
• What predictions will it make?
• What actions or decisions will it make based on its predictions?

**Remember: You can work on solving a piece of a problem with your invention

Inspiration - past problems chosen:

• An invasive species is taking over a lake
• Danger of drowning in unsupervised pools
• People brushing teeth for too little time
• Recognizing child abuse
• Parents have trouble understanding what a newborn needs (from their cry)
• And many more!

​

choose ~2 problems to keep in mind

What you care about or love

What the world or community needs

Problems that can be solved with AI

Sweet spot

Agenda

Agenda

How to start planning a solution

• An invasive species is taking over a lake
• Danger of drowning in unsupervised pools
• People brushing teeth for too little time
• Recognizing depression
• Parents have trouble understanding a newborn’s cries

Pick one. How can you use ML to start creating a solution?

What data could you use to train your model?

• Want some help? Use this dataset to make a model that predicts flood severity

​

http://tiny.cc/technovationresa

Research your

idea

Researching to refine your problem

Find statistics about the problem online

• Are there multiple places this problem occurs?
• How many people are impacted?
• Are there other organizations or companies addressing the problem you can learn from?

Create a survey to collect information from people in your community

Identify an expert you’d like to interview. Create a diagram of what you know and what you want to know. Then, write out the questions you want to ask them.

Identify 2 competitors and think about how your idea is different

Plan your

invention

​

This family drew a paper prototype.

After watching the video, can you begin to answer:

• How would a person use their invention?
• What different parts does invention have?

​

A paper prototype is a hand-drawn model of your invention.

It's a plan that shows the different parts of your idea, how your invention will work and move, and what materials you need.

​

1. Draw your paper prototype

• How will your invention gather or use data to make decisions?
• What decisions will your invention make?
• What actions do you want your invention to take?

2. Answer these questions to make your plan more complete.

• How will your invention look and act?
• What will it do?
• Who will use it?
• Does it need special materials?
• If you build an invention that takes actions outside of your computer... You might also need additional materials. These could be sensors + Raspberry Pi, Microbit or Arduino or other materials like LEDs or buzzers (if you want your invention to take actions)
• If your invention just works on your computer or phone, you might not need any other materials

3. Plan it out

Agenda

Make your own

ai model

Machinelearningforkids.co.uk

Add a new project

You can use number or text data w/out an account. For images and sound I have student logins.

Create a model that makes predictions

View video: http://tiny.cc/wayneresa

Connecting Scratch!

Level of experience with scratch

How would you rate yourself on a scale of 1-10?

• Discuss options for Scratch (or other coding)

Example

Air Quality Alert Model on ML4K

• Images for testing - alert, alert, clear, clear
• Chatbot to respond, app to recognize and label.

Agenda

The

Pitch

Presentations

What is your problem? How would you like to solve it?

Share your ML Model

• What was the goal?
• Were you able to do anything in Scratch?
• What do you want to do next?

Agenda

Where to go

From here

Educators Can…

• Provide “launch” or inspirational experiences, not “weeder” courses
• Connect learning with tech to real life
• Encourage learning from failure. Debugging and figuring out learning on your own is hard. Failure is ok!
• Provide opportunities for students to collaborate in their learning
• …while including assessments to measure individual learning and performance
• Teach the students who are least like you, not just those who are most like you
• (Not many of them will become CS educators…)

Educators can…

• Teach about the positive and negative impacts of technology
• Talk to students about ethics and professional responsibility
• Guide students to be resilient and solution-oriented
• Help students to learn to communicate and work with people who think differently than they do, or are from a different background

Credit: Marie DesJardins

Try it out

Sign up for a free mentor account on https://curiositymachine.org/ to use or remix full curriculum.

Includes opportunity to include hardware if desired.

​

What sort of settings are you interested to try it in?

Technovation Girls equips young women ages 10-18 with the skills to become tech entrepreneurs and leaders. With support from volunteer mentors, girls work in teams to code mobile apps that address real-world challenges that impact them.

Thank you! before you go...

1. Which part(s) of the program increased your understanding of AI significantly? Write 2:

​

⃞ Introduction to AI ⃞ Ideation ⃞ Make a Model ⃞ Other

​

2. What surprised you the most after taking this session?

​

3. How might what you learned today influence your future plans or intentions?

​

4. How many people do you think you might try AI with this spring or next fall?

​

___ Students this spring ___ Parents this spring ___ Students in the fall ___ Parents in the fall

appendix

What decisions does your invention make?

Some inventions make a single decision. Others make many more!

Think about what decisions your invention makes.

What are all the actions it could take?

Inventions make decisions to help them decide what actions to take.

Does your invention make decisions about people?

​

If yes, could your invention’s decisions hurt people or groups?

​

Either way, have you talked about your invention with your community or the people you’re trying to help?

How will your invention gather representative data?

Data is representative when it reflects the characteristics of the population on which the invention is being used.

If your data is not representative, the invention might make mistakes.

What if your invention makes a mistake?

Example: Google Maps ai finds “Best” routes to take

Inputs - dataset

• Current location
• Destination
• Mode (walk, car, bike, public transport)
• Traffic

Finds patterns with learning algorithm

Predicts what path is best to get you to your destination fastest

Source: MIT Media Lab

Example: Google Maps ai Finds “Best” routes to take

Action - Shows you

​

the best route(s)

Inputs - dataset

• Current location
• Destination
• Mode (walk, car, bike, public transport)
• Traffic

Finds patterns with learning algorithm

Predicts what path is best to get you to your destination fastest