AI: Training your Robot Companion |
Student Workbook Pages Printable Edition |
�Developed by the MIT Media Lab & i2 Learning
Table of Contents
Day 1
Day 2
Day 3
Day 4
Day 5
2
AI or Not?
3
Item | AI or Not? | Reason |
Toaster | | |
Robot Dog | | |
Automatic Door | | |
Car | | |
AI or Not?
4
Item | AI or Not? | Reason |
Face Filter | | |
Robot with Remote | | |
Virtual Assistant | | |
Self-Driving Car | | |
AI or Not?
5
Item | AI or Not? | Reason |
Toaster | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Robot Dog | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Automatic Door | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Car | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
AI or Not?
6
Item | AI or Not? | Reason |
Face Filter | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Robot with Remote | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Virtual Assistant | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Self-Driving Car | AI Not AI | This ( is / is not ) an example of AI because: ꠛ perceives environment ꠛ learns over time ꠛ makes plans on its own ꠛ is intelligent ꠛ interacts with its environment |
Ethical Dilemmas
B. Climate Catastrophe. In the city of Winterfell ran a river. The river once was the pride of the area, but for many years now it had become a wasteland where everything from trash, industrial waste, and unwanted junk was dumped. The city raised a significant amount of money to clean up the river, and found only one company that would be able to return the river to nearly its original quality. The city wants to hire this company, until someone brings to their attention that a different branch of the company is deeply involved in drilling for oil, and has been responsible for the damage of many ecosystems in the areas where it drills, including at least one major oil spill. Paying for this company to clean the Winterfell river would almost certainly support them to damage other environments elsewhere. What should the people of Winterfell do?
7
Ethical Dilemmas
C. Big Brother. A school wants to install new artificial intelligence “chatbots” with cameras in each space in their building, including classrooms, hallways, and stairwells. These devices have the capability to listen to and recognize human speech, connect with the other devices in the building, and take pictures and video. The devices are marketed as a way to help teachers and students and keep everyone safe. However, some people think that installing the chatbots will invade everyone’s privacy and will be a distraction. What are the pros and cons of the school installing these devices? Should the school install them?
8
Ethical Dilemmas
D. A Rock and a Dry Place. Two people were walking in the desert on their way back to the nearest oasis. They had with them just their clothes and water carriers, no food, no animals, nor any equipment. Upon stopping for some water they realized that they only had enough water for one of them to reach the oasis. One of the wanderers was an old and wise, a well respected elder who had lived a great life of honor. The second traveler was younger, but strong in body and spirit. The wanderers had to decide what to do with their remaining water - what should they do?
E. Best of the Worst. Two young men, brothers, had got into serious trouble. They were secretly leaving town in a hurry and needed money. Karl, the older one, broke into a store and stole a thousand dollars. Bob, the younger one, went to a retired old man who was known to help people in town. He told the man that he was very sick and that he needed a thousand dollars to pay for an operation. Bob asked the old man to lend him the money and promised that he would pay him back when he recovered. Really Bob wasn't sick at all, and he had no intention of paying the man back. Although the old man didn't know Bob very well, he lent him the money. So Bob and Karl skipped town, each with a thousand dollars. Which is worse, stealing like Karl or cheating like Bob?
9
Pizza Recipe Algorithm
Write an “algorithm” to make a pizza:
10
Question #1: | Question #2: |
Answer 1: | Answer 1: |
Answer 2: | Answer 2: |
Question #3: | Question #4: |
Answer 1: | Answer 1: |
Answer 2: | Answer 2: |
Combo # | Combo # | Combo # | Combo # | Combo # | Combo # | Combo # | Combo # |
Q1: | Q1: | Q1: | Q1: | Q1: | Q1: | Q1: | Q1: |
Q2: | Q2: | Q2: | Q2: | Q2: | Q2: | Q2: | Q2: |
Q3: | Q3: | Q3: | Q3: | Q3: | Q3: | Q3: | Q3: |
Q4: | Q4: | Q4: | Q4: | Q4: | Q4: | Q4: | Q4: |
Pizza: | Pizza: | Pizza: | Pizza: | Pizza: | Pizza: | Pizza: | Pizza: |
Cat/Dog Classifier
The next few slides can be printed if students would like to train their classifier by taking pictures of the cat / dog examples.
All images labelled: Cat <#> are training images for the cat category.
All images labelled: Dog <#> are training images for the dog category.
All images labelled: Test <#> are testing images students use to evaluate their models.
All unlabelled images are extra images students can use to further train and test their models.
13
Cat Training Set
Cat 15
14
Cat 1
Cat 2
15
Cat 3
Cat 4
16
Cat 5
Cat 6
17
Cat 7
Cat 8
18
Cat 9
Cat 10
19
Cat 11
Cat 12
20
Cat 13
Cat 14
21
Dog 7
22
Dog Training Set
Dog 1
Dog 2
23
Dog 3
Dog 4
24
Dog 5
Dog 6
25
Test 7
26
Test
Set
Test 1
Test 2
27
Test 3
Test 4
28
Test 5
Test 6
29
30
Extra Cats and Dogs
31
32
33
34
35
36
Cat/Dog Classifier
How many images are included?
How are the images similar?
How are the images different?
How many images are included?
How are the images similar?
How are the images different?
37
Cat/Dog Classifier Training Data
38
Cat/Dog Classifier Test Data
39
Image | Classification (Cat or Dog?) | Confidence Score | Was the classifier correct? |
| | | |
| | | |
| | | |
| | | |
Cat/Dog Classifier Test Data
40
Image | Classification (Cat or Dog?) | Confidence Score | Was the classifier correct? |
| | | |
| | | |
| | | |
Cat/Dog Classifier Test Data
41
Image | Classification (Cat or Dog?) | Confidence Score | Was the classifier correct? |
| Cat Dog | | Yes No |
| Cat Dog | | Yes No |
| Cat Dog | | Yes No |
| Cat Dog | | Yes No |
Cat/Dog Classifier Test Data
42
Image | Classification (Cat or Dog?) | Confidence Score | Was the classifier correct? |
| Cat Dog | | Yes No |
| Cat Dog | | Yes No |
| Cat Dog | | Yes No |
Cat/Dog Classifier Reflection
Which class did your classifier work better on? (Circle one)
Cats Dogs
Why do you think that is?
With your group, use the extra cat dog images, whether printed or from online, to re-curate your training dataset.
43
Cat/Dog Classifier Reflection
Which class did your classifier work better on? (Circle one)
Cats Dogs
The accuracy rate per class is calculated as the number of test set images from a class that were correctly labelled over the total number of test set images of the class.
44
Number of Cats Labelled as Cats
Number of Cats in Test Set
=
=
%
Number of Dogs Labelled as Dogs
Number of Dogs in Test Set
=
=
%
#
#
#
#
#
#
Cat/Dog Classifier Reflection
Which class did your classifier work better on? (Circle one)
Cats Dogs
Why do you think that is?
With your group, use the photos on the tables to re-curate your training dataset.
45
46
Pizza Delivery App | Areas of Interest | ||||
Tastiness | Healthiness | Speed | Healthiness | ||
Stakeholders | App User (Pizza Eater) | | | | |
App User’s Family | | | | | |
Maker of the Pizza | | | | | |
App User’s Teacher | | | | | |
47
Ethical Matrix Topic: ____________ | Areas of Interest | ||||
______________ | ______________ | ______________ | ______________ | ||
Stakeholders | ______________ | | | | |
______________ | | | | | |
______________ | | | | | |
______________ | | | | | |
Exploring Word Embeddings
This is a two-dimensional graph. Two-dimensional means that it has two axes: an x-axis and a y-axis.
We put seven words on the two-dimensional graph. Which words from the word bank to label these axes.
48
Word bank: age (young-old), size (small-large), color (red-purple), habitat (sea-land)
X-axis: ______________________
Y-axis: ______________________
Type your answer here
Type your answer here
Exploring Word Embeddings
Open the ‘Explore word analogies’ website and look at the ‘Gender analogies’ graph.
All of the roughly parallel lines are analogies. For example:
1. queen is to she like king is to he
2. aunt is to uncle as actress is to actor
This page is just for reading. Look at the graph, then go to the next page.
49
Exploring Word Embeddings
All of these analogies go along the x-axis. Add an analogy that goes along the y-axis by thinking of a word that completes: girl is to woman as _____ is to queen.
Add the pair “girl” and your new word.
Can you think of any other vertical pairs?
50
???
Exploring Word Embeddings
Now, the graph axes are labeled young-old for the x-axis and boy-girl for the y-axis. Place the following words on the graph.
51
young
old
boy
girl
Exploring Word Embeddings
Now, the graph axes are labeled animal-plant-old for the x-axis and mountain-ocean for the y-axis. Place the following words on the graph.
52
animal
plant
ocean
mountain
Exploring Word Embeddings
Open an ‘Empty’ graph on the ‘Explore word analogies website.
Make the axes young - old and boy-girl.
Place mother, grandmother, son, grandfather, father, and daughter on the graph.
Compare your graph to the one on the website. Are there any differences?
53
Type your answer here
Exploring Word Embeddings
Open an ‘Empty’ graph on the ‘Explore word analogies website.
Make the axes animal - plant and mountain - ocean.
Place deer, coconut, tortoise, fish, pig, seaweed on the graph.
Compare your graph to the one on the website. Are there any differences?
54
Type your answer here
Exploring Word Embeddings
Now, change the y-axes to say male-female instead of boy-girl.
How did the graph change?
55
Type your answer here
Exploring Word Embeddings
Besides problems with accuracy, word embeddings are also known to have some biases such as gender and age bias.
Starting from an ‘Empty’ graph, make the axes young - old and male - female.
Place the following words on the graph: nurse, doctor, scientist, dancer, teacher, actor, artist.
56
What problems do you want your AI project to solve?
What tasks do you want your AI project to help with?
What are some ways your AI project could accomplish its task?
What aspects of Artificial Intelligence will your project use?
Final AI Project Planner
| Where? | What? |
Who? | Why? | How? |
Final AI Project Planner
| What is the main goal of the project? | Where will it be used? |
Who will use it? Who will it help? | How does it work? | Why is this project important? |
Final AI Project Planner
| Where will it be used? Choose ALL that apply:
| What is the main goal of the project? Draw your project here: |
Who will use it? Choose ALL that apply:
| Why is this project important? Choose ALL that apply:
| How does it work? Choose ALL that apply:
|
Task - What will your project accomplish?
Task - What will your project accomplish?
Our project will help with .
Our project will accomplish by
First ____________________________________
Then ___________________________________
And finally _______________________________
(who)
(goal)
(goal)
Technology - What hardware (robot) and software (scratch program, AI algorithm) will your project use?
Technology - What hardware (robot) and software (scratch program, AI algorithm) will your project use?
Our robot will _____________________________________________________
_________________________________________________________________.
Our scratch program will ____________________________________________
_________________________________________________________________.
Our AI algorithm will (choose one: identify images / use voice commands).
Our AI algorithm will use data such as __________________________________
_________________________________________________________________.
Technology - What software (scratch program, AI algorithm) will your project use?
Our scratch program will _____________________________________________
_________________________________________________________________.
Our AI algorithm will use (image recognition / text classification) to classify ______ _________________________________________________________________.
We will train our algorithm with data such as ______________________________ _________________________________________________________________.
We will test our model by ____________________________________________.
Ethics - who are the stakeholders for this project? Are there any conflicts of interest?
(description)
(stakeholders)
(positive impacts)
(negative impacts)
Project Description
Directions: Use Sparki to brainstorm project ideas. Your password is:
Project title Project creators .
Our project will .
The main stakeholders for our project are .
The most vulnerable stakeholders are .
The benefits of our project are .
The possible harms of our project are .
Project Rubric
Directions: Use Jibo and the AI Playground to make an example of AI that supports physical, emotional, mental, or social wellbeing.
As you work on your project, check off the following goals.
Ask for help if you get stuck.
How? .
Drag checkmarks from here
Project Progress
Directions: As you work on your project, paste screenshots and recordings of your project here. You can also use this slide to post questions and ask for help from the teachers and classmates.
My questions:
Ethics - how are people affected by your project?
The group(s) who benefit the most from our project are _____________________.
Possible negative impacts of our project include __________________________.
We will track our project’s impact by ___________________________________.
Stakeholders (circle all that apply)
Users Community members Children The company selling it
Government Other technology companies
Ethics - who are the stakeholders for this project? Are there any conflicts of interest?
Stakeholders (circle all that apply)
Users Community members Children The company selling it
Government Other technology companies
Areas of Interest (circle all that apply)
Profit (making money) Easy to use Affordability (cheap to buy)
Safety Privacy Efficiency (works well)
Task - Task Analysis
| Goal 1: | Goal 2: | Goal 3: | Goal 4: | Goal 5: | Goal 6: |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | |
Small Steps
Technology- Task Analysis
| Goal 1: | Goal 2: | Goal 3: | Goal 4: | Goal 5: | Goal 6: |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | |
Small Steps
Ethics - Task Analysis
| Goal 1: | Goal 2: | Goal 3: | Goal 4: | Goal 5: | Goal 6: |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | |
Small Steps
Task - Task Analysis
| Goal 1: Decorate Robot | Goal 2: Create “Scenery” for Project | Goal 3: Explain goals of project | Goal 4: Explain why this project is important | Goal 5: Create other aspects of model project | Goal 6: |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | |
Small Steps
Technology- Task Analysis
| Goal 1: Build Scratch Program for project | Goal 2: Use dataset to train machine learning AI algorithm | Goal 3: Eliminate bias from the dataset | Goal 4: Connect Robot to Scratch Program and AI algorithm | Goal 5: | Goal 6: |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | |
Small Steps
Ethics - Task Analysis
| Goal 1: Identify Stakeholders | Goal 2: Identify Areas of Interest | Goal 3: Identify conflicts of interest | Goal 4: Explain steps taken to minimize ethical concerns | Goal 5: Explain overall societal impact project will have | Goal 6: |
| | | | | | |
| | | | | | |
| | | | | |
Small Steps
Peer Feedback Prep
Peer Feedback Prep
The task (what the project does/who uses it etc.)
The technology (how to use scratch and AI technology to complete project goals)
The ethics (who are the stakeholders and what are the conflicts of interest?)
Scratch Program AI algorithm Ethical Matrix Robot Use Presentation Materials Other: ______________
In writing In conversation
Feedback Form
Area of Focus/Specific Question:
Feedback:
Area of Focus/Specific Question:
Feedback:
Area of Focus/Specific Question:
Feedback:
We got feedback - now what?
What pieces of feedback did you receive? | What are you going to do to address that feedback? | Who will take care of this? |
| | |
| | |
| | |
| | |
Final To-Do List
What task needs to be completed? | What specific actions need to be done to complete this task? | Who will take care of this? |
| | |
| | |
| | |
| | |