AI 101

By Brandon Leshchinskiy

1
Watch Mashable video about Google’s AI-based personal assistant:
https://www.youtube.com/watch?v=JvbHu_bVa_g

2
With the right data and the right model,
machine learning can solve many problems.

But finding the right

data and training the
right model
can be difficult.
3
AI

ML

Deep
Learning 4
Car photo © Google; kitten photo © source unknown. All rights reserved.
This content is excluded from our Creative Commons license. For more information, 5
see https://ocw.mit.edu/help/faq-fair-use/.
 AI can be general or narrow.

Terminator image © Skydance Media; car photo © Google. All rights reserved.
This content is excluded from our Creative Commons license. For more 6
information, see https://ocw.mit.edu/help/faq-fair-use/.
 AI can be general or narrow.

Terminator image © Skydance Media. All rights reserved.

This content is excluded from our Creative Commons license. For more 7
information, see https://ocw.mit.edu/help/faq-fair-use/.
 AI can be general or narrow.

Terminator image © Skydance Media; car photo © Google. All rights reserved.
This content is excluded from our Creative Commons license. For more 8
information, see https://ocw.mit.edu/help/faq-fair-use/.
 AI can be general or narrow.

© source unknown. All rights reserved. This content is excluded from our Creative Commons license. 9
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 AI can be general or narrow.

Terminator image © Skydance Media; car photo © Google. All rights reserved.
This content is excluded from our Creative Commons license. For more 10
information, see https://ocw.mit.edu/help/faq-fair-use/.
 Typical “narrow” tasks include vision,
language processing, and planning.

Photo of eye © source unknown; smart speaker and map detail © Google. All rights reserved. This content is excluded
11
from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 Typical “narrow” tasks include vision,
language processing, and planning.

Photo of eye © source unknown. All rights reserved. This content is excluded from our 12
Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 Typical “narrow” tasks include vision,
language processing, and planning.

Photo © Google. All rights reserved. This content is excluded from our Creative Commons license. 13
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 Typical “narrow” tasks include vision,
language processing, and planning.

© Google. All rights reserved. This content is excluded from our Creative Commons license. 14
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 Typical “narrow” tasks include vision,
language processing, and planning.

Photo of eye © source unknown; smart speaker and map detail © Google. All rights reserved. This content is excluded
from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/. 15
There are many ways to build AI,
including expert systems and tree search.

16
There are many ways to build AI,
including expert systems and tree search.

17
There are many ways to build AI,
including expert systems and tree search.

18
There are many ways to build AI,
including expert systems and tree search.

19
AI

ML

20
 Machine learning can perform many tasks,
i.e. classification, clustering, and regression.

Cat photo © source unknown. All rights reserved. This content is excluded from our Creative 21
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 Machine learning can perform many tasks,
i.e. classification, clustering, and regression.

© source unknown. All rights reserved. This content is excluded from our Creative 22
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
Machine learning can perform many tasks,
i.e. classification, clustering, and regression.

23
Machine learning can perform many tasks,
i.e. classification, clustering, and regression.

24
 Machine learning can perform many tasks,
i.e. classification, clustering, and regression.

Cat photo © source unknown. All rights reserved. This content is excluded from our Creative
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/. 25
 There are three types of learning:
supervised, unsupervised, and
reinforcement learning.

Images © source unknown. All rights reserved. This content is excluded from our Creative
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/. 26
 There are three types of learning:
supervised, unsupervised, and
reinforcement learning.

Images © source unknown. All rights reserved. This content is excluded from our Creative 27
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 There are three types of learning:
supervised, unsupervised, and
reinforcement learning.

Photo © source unknown. All rights reserved. This content is excluded from our Creative 28
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 There are three types of learning:
supervised, unsupervised, and
reinforcement learning.

Images © source unknown. All rights reserved. This content is excluded from our Creative
29
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 There are three types of learning:
supervised, unsupervised, and
reinforcement learning.

Images © source unknown. All rights reserved. This content is excluded from our Creative 30
Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
AI

ML

Deep
Learning 31
With the right data and the right model,
machine learning can solve many problems.

But finding the right

data and training the
right model
can be difficult.
32
 1. Define a problem.

Cat photo © source unknown; dog photo © Getty images. All rights reserved. This content is excluded from our
33
Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
Golden retriever photo © Getty Images; other photos © source unknown. All rights reserved. This content
is excluded from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.

2. Find data.

Cat
Dog
Cat

Cat Do g Dog
34
Golden retriever photo © Getty Images; other photos © source unknown. All rights reserved. This content
is excluded from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.

3. Clean data.

Cat
Dog
Cat

Cat Do g Dog
35
Golden retriever photo © Getty Images; other photos © source unknown. All rights reserved. This content
is excluded from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.

3. Clean data.

Dog Cat
Cat

Cat Dog Dog

36
4. Choose a model.
Dog Cats
s
Always Always

Sometimes Sometimes

37
5. Train the model.

38
 5. Train the model.

Cat

© source unknown. All rights reserved. This content is excluded from our Creative Commons license.
39
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 5. Train the model.

Cat

© source unknown. All rights reserved. This content is excluded from our Creative Commons license. 40
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 5. Train the model.

Dog

© source unknown. All rights reserved. This content is excluded from our Creative Commons license.
41
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 5. Train the model.

Dog

© source unknown. All rights reserved. This content is excluded from our Creative Commons license. 42
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
 6. Test the model.

Cat

© source unknown. All rights reserved. This content is excluded from our Creative Commons license.
43
For more information, see https://ocw.mit.edu/help/faq-fair-use/.
7. Deploy the model.

 44
Golden retriever photo © Getty Images; other photos © source unknown. All rights reserved. This content
45
is excluded from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.
AI

ML

Deep
Learning 46
1.Goal?

47
1.Goal?
2.Training data?

48
1.Goal?
2.Training data?
3.Model?

49
1.Goal?
2.Training data?
3.Model?
4.Accuracy?
50
1.Goal?
2.Training data?
3.Model?
4.Accuracy?
51
AI 101
By Brandon Leshchinskiy

52
1.Goal?
2.Training data?
3.Model?
4.Accuracy?
53
MIT OpenCourseWare

https://ocw.mit.edu/

RES.6-013 AI 101 Fall 2021

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

54