Lesson 3: AI and ML Core Concepts

Question 1

Artificial Intelligence

Answer 1

the ability for a computer to simulate human cognitive behavior.
Computer vision: the process of interpreting the world visually. We train models on videos or images, typically to recognize some object, state, or interaction.

Question 2

Differential privacy

Answer 2

A technique in which individual data points are modified with some constant error term. The end result is that the data points can no longer be tied back to specific locations but the distribution of this data remains the same in aggregate. For example, a dataset of police incidents may use differential privacy to protect accidental disclosure of information, such as knowledge of a domestic disturbance. The dataset would have each point (in latitude and longitude) enclosed in a circle of radius r, where r might be approximately one city block. We choose a random point within the circle and define that point as the location of the incident. On net, we know how many crimes there are in a particular neighborhood, but we do not necessarily know at which houses the incidents occur.

Question 3

Homomorphic encryption

Answer 3

A style of encryption which allows a developer or data scientist to perform calculations on encrypted data without decrypting it first. The result of these calculations will also be in an encrypted form and the decrypted result will be exactly the same as if we performed all of the operations on unencrypted data. This ensures that data scientists can work on data sets while maintaining maximum privacy, as they will not see the unencrypted data at any time.

Question 4

Knowledge mining

Answer 4

the process of extracting knowledge from vast amounts of information. In Azure, Azure Cognitive Search is the primary tool. It tags information in documents, allowing for easy, detailed searches of those documents.

Question 5

What is the Acronym for Language Understanding Intelligent Service?

Answer 5

LUIS

Question 6

LUIS is an Acronym for which words?

Answer 6

Language Understanding Intelligent Service

Question 7

Language Understanding Intelligent Service (LUIS) is…

Answer 7

a service which accepts written or spoken inputs, processes those inputs, and allows developers to perform some action based on those inputs.

Question 8

Machine Learning

Answer 8

the process of combining algorithms and data to allow a computer to learn without human intervention.

Question 9

What is the Acronym for Natural language processing?

Answer 9

NLP

Question 10

Which words does the acronym, NLP, stand for?

Answer 10

Natural language processing

Question 11

Natural language processing (NLP) is…

Answer 11

is the process of interpreting written or spoken language. We train models based on written documents or audio clips of speech.

Question 12

Responsible AI

Answer 12

The Responsible AI project is intended to serve as a framework for promoting ethical behavior when working with and deploying artificial intelligence systems. It consists of six guidelines: Fairness, Reliability and Safety, Privacy and Security, Inclusiveness, Transparency, Accountability

Question 13

What are the six guidelines of Responsible AI?

Answer 13

Fairness, 
Reliability and Safety, 
Privacy and Security, Inclusiveness, 
Transparency, 
Accountability

Question 14

Fairness

Answer 14

One of the six guidelines of Responsible AI. It is an AI systems should treat all people fairly and not affect similarly situated groups in different ways.

Question 15

Reliability and Safety

Answer 15

One of the six guidelines of Responsible AI. Customers should be able to trust that AI solutions will perform reliably and safely within a clear set of parameters, as well as respond safely to unanticipated situations.

Question 16

Privacy and Security

Answer 16

One of the six guidelines of Responsible AI. AI systems should be secure and respect existing privacy laws.
Inclusiveness: AI systems should engage and empower people and use inclusive design practices to eliminate unintentional barriers.

Question 17

Transparency

Answer 17

One of the six guidelines of Responsible AI. People should know how AI systems work and how they interact with data to make decisions.

Question 18

Inclusiveness

Answer 18

One of the six guidelines of Responsible AI. AI systems should engage and empower people and use inclusive design practices to eliminate unintentional barriers.

Question 19

Accountability

Answer 19

One of the six guidelines of Responsible AI. Those who design and deploy AI systems are accountable for how their systems operate.

Question 20

Responsible ML

Answer 20

The Responsible ML project is an effort to control machine learning models and protect users. It consists of three key guidelines: Understand, Protect, Control.

Question 21

What are the 3 key guidelines of responsible ML?

Answer 21

Understand,
Protect,
Control.

Question 22

Understand

Answer 22

One of the 3 key guidelines of responsible ML. We should understand our models, including knowing which factors play a role and to what extent they affect the outcome of the model. This ties in quite closely to the Responsible AI guidelines of transparency and fairness.

Question 23

Protect

Answer 23

One of the 3 key guidelines of responsible ML. We want to use tools and processes which protect data privacy at all times, even during training.

Question 24

Control

Answer 24

One of the 3 key guidelines of responsible ML. We should create audit trails and track the lineage of our models. This will help us ensure that the right model is producing the expected results in a production environment.

Question 25

Strong Artificial Intelligence

Answer 25

a system which exhibits a human’s ability to generalize and solve a variety of problems, including learning how to solve new problems without direct human programming. As of today (2021), there are no strong AI systems in the world, and there remains debate in the academic community around whether strong AI is achievable.

Question 26

Super Artificial Intelligence

Answer 26

is a system which surpasses a human’s ability to generalize and solve problems. This is like strong AI, but the expectation is that the system is superior in every domain of problem-solving capability.

Question 27

Weak Artificial Intelligence

Answer 27

is a narrow application of intelligence, focused on solving one problem. An example of this is a system which counts people who enter or leave a bus. All artificial intelligence systems today are considered weak AI.