Intro to Amazon AI Services

© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Web Summit - Lisbon

November 2017

AI Services on AWSBoaz Ziniman – Technical Evangelist, AWS

@ziniman


The difficulty comes in writing software that will make sense

of the data

© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Artificial Intelligence

A system or service which can perform tasks that usually require human intelligence


25,000 skills


And a few more examples…

Fraud detection Detecting fraudulent transactions, filtering spam emails, flagging suspicious reviews, …

Personalization Recommending content, predictive content loading, improving user experience, …

Targeted marketing Matching customers and offers, choosing marketing campaigns, cross-selling and up-selling, …

Content classification Categorizing documents, matching hiring managers and resumes, …

Churn prediction Finding customers who are likely to stop using the service, free-tier upgrade targeting, …

Customer support Predictive routing of customer emails, social media listening, …


Machine Learning On AWS Today


Amazon AI Ecosystem


General-purpose GPU compute applications.Features:• High Frequency Intel Xeon E5-2686v4 (Broadwell)• High-performance NVIDIA K80 GPUs, each with

2,496 parallel processing cores and 12GiB of GPU memory

• Supports GPUDirect™• Enhanced Networking• EBS-optimized by default at no additional cost

Amazon EC2 P2 Instances

Instance Size GPUs GPU Peer to Peer

vCPUs Memory (GiB)

Network Bandwidth*

p2.xlarge 1 - 4 61 1.25Gbpsp2.8xlarge 8 Y 32 488 10Gbpsp2.16xlarge 16 Y 64 732 20Gbps


General-purpose GPU compute applications.Features:• High Frequency Intel Xeon E5-2686v4 (Broadwell)• High-performance NVIDIA V100 GPUs, each with

5,120 GPU cores + 640 Tensor cores and 16GiB of GPU memory

• Supports GPUDirect™• Enhanced Networking• EBS-optimized by default at no additional cost

Amazon EC2 P3 Instances

Instance Size GPUs GPU Peer to Peer

vCPUs Memory (GiB)

Network Bandwidth*

p3.2xlarge 1 - 8 61 Up to 10Gbpsp3.8xlarge 4 Y 32 244 10Gbpsp3.16xlarge 8 Y 64 488 25Gbps


Expedia - Ranking Hotel Images

Leading online travel companyHow to improve hotels listing images?

• Use GPUs and deep learning to rank hotel images• Build a model based on 100K images

How to rank 10M images?• Use data parallelization across multiple GPUs on AWS• Improve ranking time from more than a week to 1 day


Amazon AI Ecosystem


One-ClickDeep Learning

AWS Deep Learning AMIsAmazon Linux & Ubuntu

Up to~40k CUDA cores

Apache MXNet

TensorFlow

Theano

Keras

Caffe

CNTK

Torch

Pre-configured CUDA drivers

Anaconda, Python3

Out-of-the-box Tutorials

+ CloudFormation template

+ Container Image

Available in the AWS Marketplace


Amazon AI Ecosystem


Can We Help CustomersPut Intelligence At The Heart Of Every Application & Business?


Amazon AI Ecosystem


PollyText-to-Speech

Artificial Intelligence Services on AWS


RekognitionImage Analysis



LexConversation Bots



PollyText-to-Speech



Amazon Polly

“Today in Seattle, WA it’s 11°F”

“Today in Seattle Washington it’s 11 degrees Fahrenheit”

Text In, Life-like Speech Out


A Focus On Voice Quality & Pronunciation

1. Automatic, Accurate Text Processing• “Today in Seattle, WA, it’s 11°F”• ‘"We live for the music" live from the Madison Square Garden.’



1. Automatic, Accurate Text Processing2. Intelligible and Easy to Understand



1. Automatic, Accurate Text Processing2. Intelligible and Easy to Understand3. Add Semantic Meaning to Text


https://www.w3.org/TR/speech-synthesis/

<speak>The spelling of my name is

<prosody rate='x-slow'><say-as interpret-as="characters">Boaz</say-as>

</prosody></speak>





• “Richard’s number is 2122341237“<say-as interpret-as="telephone">




• “Richard’s number is 2122341237“ • “My daughter’s name is Kaja.”





4. Amazon Polly-specific tags• dynamic range compression (drc)

<speak><prosody volume="loud">This text needs to be understandable and loud.<amazon:effect name="drc">This text also needs to be more understandable in a moving car.</amazon:effect></prosody></speak>





4. Amazon Polly-specific tags• dynamic range compression (drc)• whispered


Duolingo voices its language learning service Using Polly

Duolingo is a free language learning service where users help translate the web and rate translations.

With Amazon Polly our users benefit from the most lifelike Text-to-Speech voices available on the market.

Severin HackerCTO, Duolingo

”“ • Spoken language crucial for

language learning

• Accurate pronunciation matters

• Faster iteration thanks to TTS

• As good as natural human speech


High quality,through

best-in-classdeep learning

Deep functionality

Easy to use& thoughtfully

integrated

Built forproduction

Lowcost

Polly: Life-like Speech Service


RekognitionImage Analysis



Amazon Rekognition

Object and scene detectionFacial analysisFace comparisonCelebrity recognitionImage moderation

Amazon Rekognition: Images In, Rich Metadata Out


Object & Scene Detection


Facial Analysis


Smiling?

Facial Analysis

(Deep) Learning from a Masterpiecehttp://bit.ly/MonaLisaAI


Celebrity Recognition


Facial Search


Explicit NudityNudityGraphic Male NudityGraphic Female NuditySexual ActivityPartial Nudity

SuggestiveFemale Swimwear or UnderwearMale Swimwear or UnderwearRevealing Clothes

Image moderation


Amazon Rekognition Customers

Digital Asset ManagementMedia and EntertainmentTravel and HospitalityInfluencer MarketingSystems IntegrationDigital AdvertisingConsumer StorageLaw EnforcementPublic SafetyeCommerceEducation


DEMO




Deep functionality


integrated

Built forproduction

Lowcost

Rekognition: Search & Understand Visual Content


LexConversation Bots



Amazon LexAutomatic Speech RecognitionNatural Language Understanding

“What’s the weather forecast?”

WeatherForecast

Speech Recognition & Natural Language Understanding


Amazon LexAutomatic Speech RecognitionNatural Language Understanding

“What’s the weather forecast?”

“It will be sunny and 25°C”

WeatherForecast

Speech Recognition & Natural Language Understanding


Origin

Destination

Departure Date

✈ Flight Booking


Origin

Destination

Departure Date

✈ Flight Booking

“Book a flight toLondon”



AutomaticSpeech Recognition

Book Flight

London

Origin

Destination

Departure Date

✈ Flight Booking


Natural LanguageUnderstanding

Flight booking

London Heathrow



Book Flight

London

Origin

Destination

Departure Date

✈ Flight Booking



Flight booking

London Heathrow



Book Flight

London

Origin

Destination London Heathrow

Departure Date

✈ Flight Booking



Flight booking

London Heathrow



Book Flight

London

Origin Lisbon


Departure Date

✈ Flight Booking



Flight booking

London Heathrow



Book Flight

London

Origin Lisbon


Departure Date

✈ Flight Booking

Prompt

“When would you like to fly?”



PollyPrompt



Flight booking

London Heathrow



Book Flight

London

Origin Lisbon


Departure Date

✈ Flight Booking



“Next Friday”

Origin Lisbon


Departure Date

✈ Flight Booking


“Next Friday”

Origin Lisbon


Departure Date

✈ Flight Booking


Next Friday


“Next Friday”

Origin Lisbon


Departure Date

✈ Flight Booking


Flight booking

10/11/2017


Next Friday


“Next Friday”

Origin Lisbon


Departure Date 10/11/2017

✈ Flight Booking


Flight booking

10/11/2017


Next Friday


Origin Lisbon



✈ Flight Booking


Origin Lisbon



✈ Flight Booking

Confirmation

“Your flight is booked for next Friday”


Origin Lisbon



✈ Flight Booking

Confirmation

“Your flight is booked for next Friday”

“Your flight is booked fornext Friday”

Polly


Lex: Build Natural, Conversational Interactions In Voice & Text



Deep functionality


integrated

Built forproduction

Lowcost


Wrap up


Smile!@ziniman


Thank you!@ziniman

Intro to Amazon AI Services

Documents

Transcript of Intro to Amazon AI Services