GETTING STARTED WITH

AI IN THE CLOUD
Getting Started with
AI in the Cloud
There is no doubt that Artificial Intelligence (AI) is a game-changer
for businesses. The ability to take millions of records, isolate
patterns, and predict future outcomes enables organizations
to make data-driven decisions. In this ebook, you will come to
understand the value of leveraging AI, why it makes sense to run
AI in the cloud, explore what Amazon’s AWS and Microsoft’s Azure
cloud businesses are delivering to make it easy for you to build AI
solutions, and, finally, what skills you need to develop AI solutions
on cloud platforms.

Why AI is important
The importance of artificial intelligence to business has been
celebrated by many people in the past few years. This means that
the first wave of big data technologies is now entering the second
machine era.

As more companies consider building AI into their products, they

face questions about what kinds of analytics to use, how to set up
their AI systems, and the optimal environment to implement such
systems in.

2 | www.simplilearn.com
The corporate world needs to keep pace with today's rapidly
changing technologies. Artificial intelligence is already replacing
workers in warehouses or factories, and now businesses must also
start training robots and AI to be part of the company's everyday
operations.

The goal of this eBook is to illustrate how to get started with AI in

the cloud. There is a specific focus in the text on tools that enable
quick success. Often, these tools do not require programming
skills. These tools are a gateway that opens up the power of AI.
When you are comfortable with the technology, you can then roll
up your sleeves and start coding with Python.

Why it makes sense to run

AI in the cloud
The purpose of AI is to analyze patterns in data and then make
predictions based on those patterns. At the same time, the
earliest cloud technologies have been leveraged to store extensive
amounts of data. It is for this reason that it makes so much sense
to have AI running in the cloud. If your data is in the cloud, why
not have your AI scripts also running in the cloud? The closer your
programs are to the data, the more efficient your scripts will run.

At its core, there are three levels of AI:

Artificial Intelligence

Machine Learning

Deep Learning

AI is a general term used for computer learning. You can consider

it a blanket term when talking about tracking patterns and
practices with a computer.

Machine learning (ML) uses mathematical principles and data

analysis to extrapolate the behavior of unknown, unstructured
data to make predictions about what the future will bring. ML is a
group of methods that help systems become better at answering
complex questions over time. It is a set of techniques that not only
can gather and analyze large amounts of data but can also use

3 | www.simplilearn.com
these data to make predictions.

Deep learning (DL) is a way of having multiple computers learn

how to think through problems using statistics. In a sense, DL
mimics human intelligence. There are now two major players in
the game—Google, with its recent acquisition of DeepMind; and
Microsoft, through its acquisition of several startups over the past
few years, including Maluuba and Lobe.

Each of these approaches to AI can be easily applied to cloud

services to solve various problems. For example, Google uses
DeepMind, a deep learning solution, to manage the heating and
cooling for all of its data centers worldwide.

AI in AWS
Amazon’s AWS is arguably the largest cloud provider and growing
at an exponential rate. However, AWS does provide leading AI
solutions. AWS breaks out the following services you can use to
build settlements in their cloud:

ML Services: Build, train, and deploy ML fast

AI Services: Easily add intelligence to your applications

Frameworks: Choice and flexibility with broad framework

support

Compute

Analytics and Security

Learning Tools: Dive deep on ML with AWS DeepRacer and

DeepLens

4 | www.simplilearn.com
The Power of Machine Learning
on AWS
Machine learning on AWS is a high-performance computing
service. This service enables Amazon EC2 instances with 16 cores
(up to 2TB for each center) and enough memory (256GB) to store
up to 14 petabytes of data at a time.

Similarly, each machine is connected to one or more Amazon's

EC2 Instances with up to 16 cores. You can deploy a fleet of these
machines, called a cluster, on AWS's network. When used on AWS,
a machine learning cluster only needs one IP address, so you can
be sure that all the machines will work together as one.

AWS EC2 instances use the TensorFlow library. This is useful for
building machine learning algorithms on AWS servers. It can be
used to create models of hierarchical data structures and feed
the models with sample data. AWS is currently running a massive
amount of RAM for TensorFlow on their Elastic Compute Cloud
(EC2), but that's just the tip of the iceberg—it has more than 8
Terabytes of data on its AWS S3 (filesystem) cluster. The AWS
costs start at $0.05 a month for the essential S3 bucket. They
don't let you access the data directly (this is one of their big selling
points), but they do have a read/write filesystem.

AI Services running on AWS

AWS pre-trained AI Services provide ready-made intelligence
for applications and workflows. AI Services integrates with
requests to address standard use cases such as personalized
recommendations, modernizing a contact center, improving safety
and security, and increasing customer engagement. Because AWS
uses the same deep learning technology that powers Amazon.
com and ML Services, developers get quality and accuracy from
continuously-learning APIs.

5 | www.simplilearn.com
There is a broad selection of AI tools you can use, including the
following:

Recommendations: Personalize experiences for your

customers with the same recommendation technology used
at Amazon.com

Forecasting: Build accurate forecasting models based on

the same machine learning forecasting technology used by
Amazon.com

Image and Video Analysis: Add image and video analysis

to your applications to catalog assets, automate media
workflows, and extract meaning

Advanced Text Analytics: Use natural language processing to

extract insights and relationships from unstructured text

Document Analysis: Automatically extract text and data from

millions of documents in just hours, reducing manual efforts

Voice: Turn a book into lifelike speech to give voice to your

applications

Conversational Agents: Easily build conversational agents

to i mprove customer service and increase contact center
efficiency

Translation: Expand your reach through efficient and cost-

effective translation to reach audiences in multiple languages

Transcription: Easily add high-quality speech-to-text

capabilities to your applications and workflows

Enterprise Search: Add natural language search capabilities

to your apps so your end users can more easily find the
information they need

Fraud Detection: Identify potentially fraudulent online activities

based on the same technology used at Amazon.com

Code: Automate code reviews and identify your most

expensive lines of code.

AWS wants to make it easy for you to get started with their
AI services. For this reason, their AI tools do not require
programming experience.

6 | www.simplilearn.com
Frameworks running on AWS
The most popular framework running on AWS for AI is TensorFlow.

AWS also supports the following frameworks:

PyTorch

Apache MXNet

Chainer

Theano

Gluon

The Microsoft Cognitive Toolkit

Horovod

Keras

Even the best AI developers are challenged when it comes to

building solutions on AWS. For this reason, AWS has developed
Amazon Machine Image (AMI) to accelerate implementation.
Essentially, an AMI is a container configured for your AI needs.
There are two AI AMIs you can choose from:

CONDA AMI: The Conda-based AMI is available in Ubuntu,

Amazon Linux, and Windows 2016 versions for developers who
want pre-installed pip packages of deep learning frameworks
in separate virtual environments

BASE AMI: The Base AMI is available in Ubuntu, and Amazon

Linux versions for developers who want a clean slate to set up
private deep learning engine repositories or custom builds of
deep learning engines

AWS provides step by step instructions for installing AMIs here

7 | www.simplilearn.com
AWS AI and ML Learning Tools
The best way to learn is through gaming. To this end, AWS has two
tools you can use that let you learn and play with AI at the same
time. They include:

AWS DeepComposer: AWS DeepComposer MIDI-compatible

keyboard to compose melodies as input for your ML generated
compositions

AWS DeepRacer: AWS DeepRacer is a fully autonomous

1/18th-scale race car designed to help you learn about
reinforcement learning through autonomous driving

AWS DeepLens: AWS DeepLens is a deep learning-enabled

video camera for developers. Integrated with Amazon
SageMaker and many other AWS services, it enables you to
get started with deep learning in less than 10 minutes through
sample projects and practical, hands-on examples

You can sign up to signup for DeepComposer here, buy

DeepRacer here and buy DeepLense here

All three come with great “how-to” guides, setup videos, sample
code and connection with additional AI enthusiasts.

8 | www.simplilearn.com
AI in Azure
The ultimate goal of AI is to help replicate human-driven activities.
It may sound scary—like the movie Terminator (and Elon Musk will
have you think that AI will take over the world)—but, there is an
optimistic side to AI. Today, AI is being used to filter junk email,
correct your grammar in Word, collect together your favorite im-
ages, and to automate tedious tasks. What AI is doing is freeing up
time for you to work on other, potentially more exciting topics.
In this article, you will learn about the tools and services Microsoft
Azure offers to help accelerate AI adoption.

9 | www.simplilearn.com
Microsoft Azure AI Services
AI is not new to Microsoft. If you use Office 365, then you have
seen AI appearing in tools such as Outlook (the “Focus” tab for
email), Word (grammar check-in now powered by AI), and Pow-
erPoint (AI fuels the new design recommendation action). Indeed,
Microsoft was the first company to reach human parity for image
recognition, voice-to-text recognition, and translation.

The challenge for many developers is getting easy access to AI

tools and adding AI functionality into their apps. Microsoft’s Azure
AI services help to reduce these barriers. Azure Services focus on
three key areas:

Cognitive Services: Core tools that offer human-like activities

such as image recognition, speech-to-text services, voice
services, and search

Frameworks: Support for the most popular AI frameworks

including Pytorch, Tensorflow, Keras, scikit-learn, and Onnx to
train data models

Azure Services: Building block services that accelerate support

for AI solutions including Azure Data Bricks, Azure Machine
Learning, and Machine Learning VMs

An overview of the services for Azure can be found here.

Where to use Azure AI Services

The power of AI comes through training models to identify
patterns. In many ways, AI is really only about solving complex
pattern recognition. The data used to train AI often consists of
documents, images, video, audio, and handwritten notes. Many
companies that are more than ten years old will likely have vast
archives of paper content. Sorting through the material to find
patterns is of extreme value.

An example scenario that Microsoft has developed is an AI tool

that can help sort through the paper documentation of the
recently released 34,000 pages from the investigation of the
assassination of US President John F. Kennedy (JFK). The website

10 | www.simplilearn.com
where you can search through the content can be found here.

The JFK example provides a demonstration in which AI can be

used to search typed letters and facial recognition for images, and
convert handwritten notes into searchable terms. Also, the demo
includes a graph to visualize how related terms are connected.

It is evident when looking at the JFK AI Demo site that the

same tools used on the website can provide a benefit to many
businesses. For instance, an insurance company that often has
millions of scanned policies can now quickly detect if fraud is
occuring if relationships between documents can be rapidly
identified. To kick start your work, Microsoft has made the code
available for the JFK site on GitHub

To understand how to use the GitHub files, check out the 1-hour
class that explores how to implement the JFK demo.

Training your AI
To achieve success with AI, you need to educate the tool to
identify patterns. Training AI can be a daunting task if you are new
to the technology. To help guide you, the Azure team introduced
another tool that you can use to train your AI correctly. The tool is
available at www.customvision.ai

The following example will train your AI to identify the difference

between a cat and dog. To do this, follow these steps:

1. Go to www.customvision.ai and create an account.

2. After creating your account, you will be taken to a page that

lists all of your projects. If the account is new, then there will
be no projects. Let’s go ahead and create the first project.

3. Select the “New Project” button. Give your project a name

(CatDog) and a description (identifying the difference between
Cats and Dogs).

4. At this time, you do not have a Resource. Resources are used

in Azure to monitor and manage the service you are creating.
Select “Create New” and fill out the fields as follows:

a. Name: CatDog

11 | www.simplilearn.com
 b. Subscription: Free

c. ResourceGroup: SimplilearnResource

d. Kind: CustomerVision.Training

e. Location: South Central US

f. Price Tier: $0

5. All of the data you enter will be used to manage the service in
Azure Cloud.

6. The next screen will ask what tags you want to create to train
the AI. Create labels for Cat and Dog.

7. Now, you will begin to train the AI. You will need ten images
of cats and ten pictures of dogs to do so. Upload each image
and, as you upload the image, select whether tag to determine
if the image is a cat or a dog.

8 After you have trained the AI, you can now start to upload
images and have the AI tell you if it believes the image is a cat
or a dog.

Behind the scenes, all of the work you are doing with the
customvision.ai tool is integrated with Microsoft Azure. You can
view the Azure Portal and see the new services you have created.

Wrapping Up
The tools listed above are here to kick start your ability to be
successful with AI. The next step is to start building solutions that
are more complicated and have practical value to you. This may be
for your business, community work, or simply a hobby. The bottom
line is that AI is now accessible to many more people due to cloud
tools and services.to support them.

A common misconception is that more web workers will require

more resources. However, this isn't the case at all. With several

12 | www.simplilearn.com
 websites being hosted on Azure, there are very few resources
being used to run them all. Rather, you are using only the
resources necessary to run the backend, creating a highly available
and reliable architecture. This provides more resources to the
front-end developers in terms of modules and components, which
makes framework deployment efforts much easier.

In short, what you can expect is that Azure Functions will continue
to expand in scope and capability as developers explore the full
potential of serverless applications.

Getting Started
If this eBook has piqued your interest in developing more advanced skills—and
really get your career going—Simplielarn’s comprehensive Artificial Intelligence
Engineer Master's Program will provide you with everything you need to know
to become an expert in this exciting and rapidly growing field. Co-developed
with IBM, this program is a blend of Artificial Intelligence, Data Science, Machine
Learning, and Deep Learning. It enables the real-world implementation of
advanced tools and models, while providing in-depth knowledge of AI concepts,
including Data Science, Python Programming, and Deep Learning with tools like
Keras and TensorFlow. The seven-course Blended Learning program wraps up
with an AI Capstone Project, giving you the opportunity to apply the skills and
knowledge you have gained to solve an industry-aligned problem with dedicated
mentoring sessions from experts. Upon completion, learners will receive a
Master’s Certificate which will help you stand out to existing or potential
employers in the exciting field of Artificial Intelligence.

13 | www.simplilearn.com
Founded in 2009, Simplilearn is one of the world’s leading providers of online training for Digital
Marketing, Cloud Computing, Project Management, Data Science, IT Service Management,
Software Development and many other emerging technologies. Based in Bangalore, India, San
Francisco, California, and Raleigh, North Carolina, Simplilearn partners with companies and
individuals to address their unique needs, providing training and coaching to help working
professionals meet their career goals. Simplilearn has enabled over 1 million professionals and
companies across 150+ countries train,certify and upskill their employees.

Simplilearn’s 400+ training courses are designed and updated by world-class industry experts.
Their blended learning approach combines e-learning classes, instructor-led live virtual
classrooms, applied learning projects, and 24/7 teaching assistance. More than 40 global
training organizations have recognized Simplilearn as an official provider of certification
training. The company has been named the 8th most influential education brand in the world by
LinkedIn.

For more information, visit www.simplilearn.com.

© 2009-2020 - Simplilearn Solutions. All Rights Reserved. | The certification names are the
trademarks of their respective owners.

14 | www.simplilearn.com