熱門創新服務專題
Unlocking New Todays: Artificial Intelligence and Data Platforms on AWS (Level 200)
Speaker: Paul Yung, Head of Territory Development HKT, AWS
Unlocking New Todays - Artificial Intelligence and Data Platforms on AWS
1. Unlocking New Todays
Artificial Intelligence and Data Platforms on AWS
Paul Yung - Head of Territory Development, HKT
pyung@amazon.com
Track 4
15:05pm - 15:45pm
Level: 200
11. Stanford University- Improving Health Outcomes
• Early detection of diabetic complications
• Leading cause of blindness in adults
• Catch it early enough, prevented 90% of
time (糖尿病性視網膜病變)
“Before AWS, we couldn’t
even attempt these
projects….AWS makes
research liberating.”
Jason Su
Stanford University Student
13. Systems see which items have
been taken or returned
Machine learning understands
in-store and purchase patterns
Combination of computer vision,
sensor fusion, and deep learning
Just Walk Out Technology
21. “The future is here,
it’s just not evenly distributed yet”
William Gibson
Author of Neuromancer (神經漫遊)
22. We want to democratize AI
We want to lower the costs and barriers to
Machine Learning
We want to put machine learning in the hands
of every developer and data scientist
23. Aggressive migration
New data created on AWS
Data
Training Prediction
PBs of existing data
The Challenge For Artificial Intelligence: SCALE
24. Tons of GPUs
Elastic capacity
Training
Prediction
Pre-built images
The Challenge For Artificial Intelligence: SCALE
Aggressive migration
New data created on AWS
Data
PBs of existing data
25. Tons of GPUs and CPUs
Serverless
At the Edge, On IoT Devices
Prediction
The Challenge For Artificial Intelligence: SCALE
Tons of GPUs
Elastic capacity
Training
Pre-built images
Aggressive migration
New data created on AWS
Data
PBs of existing data
26. Frameworks &
Infrastructure
AWS Deep Learning AMI
GPU
(P3 Instances)
MobileCPU IoT (Greengrass)
Vision:
Rekognition Image
Rekognition Video
Speech:
Polly
Transcribe
Language:
Lex Translate
Comprehend
Apache
MXNet
PyTorchCNTK Keras
Caffe2
& Caffe
TensorFlow Gluon
AWS ML | A Full Stack
Application
Services
Platform
Services
Amazon Machine
Learning
Mechanical
Turk
Spark &
EMR
Amazon
SageMaker
AWS
DeepLens
Amazon confidential
28. Image Recognition and Analysis powered by
Deep Learning allows search, verification, and
organization millions of images
Potential Use Cases
Searchable Image Library Detect Inappropriate Content in Images
Face-based User Verification Sentiment Analysis
Facial Recognition Celebrity Identification
Vision Service - Amazon Rekognition Image
29. Rekognition Image: Object & Scene Detection
Object and scene detection makes it easy for you to add
features that search, filter, and curate large image libraries
Identify objects and scenes and provide confidence scores
Flower
Arrangement
Chair
Coffee Table
Living Room Indoors
Furniture
Cushion
Vase
Maple
Villa
Plant
Garden
Water
Swimming Pool
Tree
Potted Plant
Backyard
Patio
33. Amazon Rekognition Video
Video Analysis
Object & Activity
Detection
Person
Tracking
Face
Recognition
Real-time Live
Stream
Content
Moderation
• Fully managed, scalable, and easy-to-use video analysis service
• Deep learning-based -> Continuously improving
• Integrated with Amazon S3, Amazon Kinesis Video Streams, AWS
Lambda – get started with video analysis right out-of-the-box
34. Turn text into lifelike speech using
deep learning technologies to
synthesize speech that sounds like a
human voice
Potential Use Cases
Content Creation Education & E-learning
Mobile & Desktop Applications Customer Contact Center
Internet of Things (IoT) Accessibility
Speech Service - Amazon Polly
Price $4.00 per 1 million characters for speech requests
35. Let’s take a listen…
「To be or not to be」莎士比亞齣戲《哈姆雷特》第三幕第一場
36. “Today in Taipei, it’s > 16°F”
‘"We live for the music" live from the Madison Square Garden.’
1. Automatic, Accurate Text Processing
Polly: A Focus On Voice Quality & Pronunciation
37. Polly: A Focus On Voice Quality & Pronunciation
“She sells sea shells by the sea shore”
1. Automatic, Accurate Text Processing
2. Intelligible and Easy to Understand
38. 1. Automatic, Accurate Text Processing
2. Intelligible and Easy to Understand
3. Add Semantic Meaning to Text
Polly: A Focus On Voice Quality & Pronunciation
“Richard’s number is 2122341237“
“Richard’s number is 2122341237“
Telephone Number
39. 1. Automatic, Accurate Text Processing
2. Intelligible and Easy to Understand
3. Add Semantic Meaning to Text
4. Customized Pronunciation
Polly: A Focus On Voice Quality & Pronunciation
“My daughter’s name is Kaja.”
“My daughter’s name is Kaja.”
40. FM Wakayama is using Amazon Polly as an AI Announcer
FM Wakayama is a public radio station that
broadcasts out of Wakayama, Japan.
Listeners seemed to think
the news were read out by
a real person
Akimasa Yamaguchi
Station Manager, FM Wakayama
”
“ • Amazon Polly is used for late-
night and early-morning radio
broadcasts.
• Amazon Polly voices can be
leveraged for Public Service
Announcements during times of
emergency, in many languages
• “In the future, Amazon Polly may
save someone’s life.”
41. Amazon Transcribe (Preview)
Automatic conversion of speech into accurate,
grammatically correct text
Support for
telephony
audio
Timestamp
generation
Intelligent
punctuation and
formatting
Recognize
multiple
speakers
Custom
vocabulary
Multiple
languages
e .g. Episode of "Game of Thrones” (63 mins), cost @ USD 1.512
42. Amazon Translate (Preview)
Natural and fluent language translation
Real-time
translation
Batch
analysis
Automatic
language
recognition
Low cost
Price @$ 1 5 pe r m il l ion charact e rs
46. Amazon Comprehend
Discover insights and relationships in text
Entities
Key Phrases
Language
Sentiment
Amazon
Comprehend
Social media posts, emails,
web pages, documents,
phone transcriptions
Input Output
47.
48.
49.
50.
51.
52.
53. STORM
WORLD SERIES
STOCK MARKET
WASHINGTON
LI BRARY OF
NEWS ARTICLES
Amazon
Comprehend
Discover insights and relationships in text
Amazon Comprehend
e . g . 3 0 0 d o c u m e n t @ 1 M B e a c h
C o m p r e h e n d - 4 5 m i n s t o c o m p l e t e @ U S D 1 . 8
54. Frameworks &
Infrastructure
AWS Deep Learning AMI
Apache
MXNet
PyTorchCNTK Keras
Caffe2
& Caffe
TensorFlow Gluon
AWS ML | A Full Stack
Application
Services
Platform
Services
Amazon confidential
56. Support for all major frameworks
AWS Deep Learning AMI
Apache
MXNet
Torch
CNTKKeras
Theano
Caffe2
& Caffe
TensorFlow
Amazon EC2
AnacondaIntel MKL
Nvidia CUDA &
cuDNN
Python2 &
Python3
• Get started quickly with easy-to-launch tutorials
• Hassle-free setup and configuration
• Pay only for what you use – no additional charge
for the AMI
• Accelerate your model training and deployment
• Support for popular deep learning frameworks
58. Train model in the cloud Run model at the edge
AWS
Greengrass
AWS IoTAmazon
EC2
59. Amazon EC2 P3 Instances
• Up to 8 x NVIDIA Tesla V100 GPUs (Latest
generation)
• 5,120 Tensor Cores., 128GB GPU memory
• 1 PetaFLOP of computational performance –
14x faster than P2
• NVLink 2.0, 300 GB/s GPU-to-GPU
communication – 9X better than P2
T h e f a s t e s t , m o s t p o w e r f u l G P U i n s t a n c e s i n t h e c l o u d
60. Frameworks &
Infrastructure
AWS ML | A Full Stack
Application
Services
Platform
Services
Amazon Machine
Learning
Mechanical
Turk
Spark &
EMR
Amazon
SageMaker
AWS
DeepLens
Amazon confidential