Auto ML

Question 1

Kendra

Kendra TL;DR?

Answer 1

Intelligent search service powered by ML

Question 2

Kendra

How do you search Kendra?

Answer 2

Designed to mimic interacting with a human expert: natural language

Question 3

Kendra

Kendra supported Data Sources?

Answer 3

S3, Confluence, RDS, OneDrive, Salesforce, lots of commercial products, not a lot of AWS

Question 4

Textract

Textract TL;DR?

Answer 4

ML service autoamtically extracts text, handwriting, and data from scanned docs

Question 5

Translate

Translate TL;DR?

Answer 5

Language translation using ML / neural machine

Question 6

Polly

How can you increase the correctness of Polly for your company?

Answer 6

Lexicons: list of words and how they should be pronounced (just written out)

Question 7

Polly

Example of an entry in a Lexicon?

Answer 7

“k8s” -> “kubernetes”

Question 8

Polly

How can you make Polly sound more human?

Answer 8

SSML: Speech Synthesis Markup Language

Question 9

Polly

Examples of SSML?

Answer 9

Set volume, rate, and pitch in-line with text to speak.

Question 10

Lex

What’s the big thing that organizes all the interactions and data-gathering for an action?

Answer 10

An Intent: holds Slots, each Slot is one value like in a web fill-out form

Question 11

Lex

What happens if Lex doesn’t get all the Slot values it needs?

Answer 11

Smart enough to prompt for each missing Slot with separate prompts.

Question 12

Lex

What happens when all Slots are filled?

Answer 12

Lex invokes a Lambda function to fulfill the Intent

Question 13

Lex

Different types of media you can use to interact with Lex?

Answer 13

Text or voice – chat bot on a web page with TXT-like conversation, or talk to it (uses Polly)

Question 14

Personalize

What is Personalize?

Answer 14

Provide personalized recommendations to humans

like Amazon.com uses this

Question 15

Mechanical Turk

What is Mechanical Turk?

Answer 15

Croudsource to perform simple human tasks

Question 16

Mechanical Turk

How does it work?

Answer 16

You set jobs, descriptions, and a ton of input data, get back results

Question 17

Mechanical Turk

Use Case for Mechanical Turk for AI/ML?

Answer 17

Use to add Labels to a large corpus of training data you want to use with a ML model

Question 18

Mechanical Turk

What other system does Mechanical Turk integrate with for AI use cases?

Answer 18

Augmented AI (A2I)

Question 19

Augmented AI (A2I)

What is A2I?

Answer 19

Human oversight of ML predictions

Question 20

Augmented AI (A2I)

What benefit does A2I provide for ML workloads?

Answer 20

Low confidence predictions by an ML model can be sent to A2I for a human review

Question 21

Augmented AI (A2I)

Who (humans) do the reviews in A2I?

Answer 21

Your own workforce, contractors, or send it to Mechanical Turk