The “Intelligent” Cloud:

How AI is changing the face of Cloud Computing

Artificial Intelligence (AI) is changing the face of Cloud Computing, from creating increased demand for specialized compute intensive workloads for deploying Machine Learning (ML), and Deep Learning (DL) applications; enabling developers to create “Intelligent” applications leveraging simple cloud-based AI APIs; and leveraging AI to monitor and manage large scale Data Centers.  To find out more about the opportunities that AI is creating for Cloud Service Providers, and the broader Ecosystem of Cloud Vendors, Developers and Solution Providers explore the following topics:

How AI is changing the face of Cloud Computing

Artificial Intelligence (or AI for short) is having a dramatic impact on Cloud Computing, from creating increased demand for specialized Cloud-based compute intensive workloads for deploying Machine Learning (ML), and Deep Learning (DL) applications; enabling developers to create “Intelligent” applications leveraging simple cloud-based AI APIs; and leveraging AI to monitor and manage large scale Data Centers.

The 3 Cloud-enabled breakthroughs that are accelerating AI solutions

So why is it that Artificial Intelligence has suddenly become so important across so many different domains?

The current “revolution” in Artificial Intelligence is the direct result of three key (cloud-enabled) breakthroughs that have accelerated the development of practical AI solutions, they are: the emergence of affordable parallel processing; the availability of Big Data; and access to improved (Machine Learning) algorithms.

Let’s take a look at each of these three (Cloud-Computing) Enablers in more detail:

Firstly, the “Emergence of High Performance Parallel Processing”

Traditional processor architectures and platforms often required weeks to calculate the option range of a Neural Network (Deep Learning algorithm).  These days clusters of modern Graphic Processor Units (GPUs) and/or specialized high end Processors such as the Intel Xeon Phi can accomplish this much faster.   A new generation of AI “optimized” processor architectures such as the Intel Nervana platform and Google’s Tensor Processing Unit (TPU) is now being developed to support the emerging Cloud-based AI-as-a-Service features and capabilities.

Not every organization has the capital and skills needed to manage the large-scale computing platforms that needed to run advanced Machine Learning. Public Cloud Service Providers such as Amazon Web Services (AWS Machine Learning), Google Cloud Platform ML Engine, Microsoft Azure ML  and the IBM Watson Developer Cloud, offer Data Scientists and Developers a scalable infrastructure optimized for Machine Learning at a lower cost than setting up and configuring their own on-premise environment.

Reference: Machine Learning Cloud Comparison: AWS, Google, IBM, Microsoft;

Reference: Machine Learning Cloud Comparison: AWS, Google, IBM, Microsoft; TechTarget (2017)

Secondly, the “Availability of Big Data”.

In 2020, the digital universe is expected to reach 44 Zettabytes (1 Zettabyte is equal to 1 Billion Terabytes). Data valuable for enterprises, especially unstructured data from IoT devices and non-traditional sources, is projected to increase both in absolute and relative sizes.

Reference: Deloittes Tech Trends 2017: The Kinetic Enterprise (2017)

Reference: Deloittes Tech Trends 2017: The Kinetic Enterprise (2017)

AI needs “Big Data” and “Big Data” Analytics needs AI

In order to develop “intelligence”, Deep Learning algorithms require access to massive amounts of data. Only when computers can be trained with this kind of input can they expand their capabilities.  Once again, Cloud Computing is enabling access to “Big Data”.  Centralized storage and processing of “Big Data” in Data Centers is powering the current generation of AI applications.

Conversely, AI is essential to the next wave of “Big Data” analytics. It is a vital tool for achieving a higher maturity and scale in data analytics; and for allowing broader deployment of advanced analytics.

The impending “flood” of Data

Looking ahead to 2020 in the new world of connected people and things, there will be a literal flood of data being generated on a daily basis.  For example, each self-driving car may generate up to 4,000 GB of data per day, and a connected (smart) factory could generate over 1 million GB per day!

The role of Edge Computing Devices

Not all of this data will be stored and/or pushed up to the cloud.  Most of it will be processed locally by specialized “Edge” computing devices.

These “Edge” devices will also be able to leverage AI analytics as a result of specialized chipsets such as the Intel Arria 10 FPGA (Field Programmable Gate Array) for real-time Deep Learning Inferencing, the Intel Movidius Myriad Vision Engine, and the Intel GNA Speech Recognition Engine.

And finally, “Access to improved (Artificial Intelligence) Algorithms”

All of the major Cloud Service Providers offer access to advanced AI capabilities such as Natural Language Processing, Image Recognition, and general purpose Machine Learning algorithms.  There is also an emerging market for big data sets and algorithms that can be used for AI applications.

For example, Algorithmia is focused on building a marketplace for algorithms that are accessible through a simple, scalable API.  Algorithmia’s goal is to make applications smarter by building a developer community around smarter applications.  More than 30,000 developers already have accessed their growing library of 3,000 algorithmic microservices.

Kaggle is a platform for predictive modelling and analytics competitions in which companies and researchers post data and statisticians and data miners compete to produce the best models for predicting and describing the data. This crowd-sourcing approach relies on the fact that there are countless strategies that can be applied to any predictive modelling task and it is impossible to know at the outset which technique or analyst will be most effective.

Kaggle has over 500,000 registered users, and the Kaggle community spans 194 countries. It is the largest and most diverse data community in the world. In 2017, Google acquired Kaggle, which will be maintained as a distinct brand, but become part of the Google Cloud AI initiative.

Which comes "first": Digital - Mobile - Cloud - AI?

Over the past few years, technology vendors and industry pundits have been espousing the mantras of Digital First, Mobile First, Cloud First and more recently AI First as a strategic direction, all within the context of Big Data, and with the business imperative of Digital Transformation.  Is this just marketing speak, or is it part of an evolution of information technology? … and which is the “first” among firsts?

Although “AI First” has become the new mantra for vendors such as Google, Microsoft, IBM, Intel and others; it doesn’t mean that the other elements such as Cloud, Mobile, Digital and Big Data are no longer relevant.  In fact, they are all synergistically inter-related.

Big Data owes its very existence to the Digital first world, and the new (real-time) forms of data that are being generated by social media, mobile applications, and increasingly through Internet of Things devices … and it is this Big Data that is feeding and enriching the value of Machine Learning and Deep Learning algorithms.

Mobile is important because it has rapidly become the dominant form of computing and communications interaction; acting both as a generator and consumer of Big Data.

Cloud is the platform which enables all of these technologies to come together.  Cloud compute resources are powering the new AI workloads; Cloud provides the central point for Big Data storage and access; Cloud provides the API access for the next generation of “Smart” mobile applications; and Cloud Services have become the operational engine for deploying new AI enabled business models.

And of course, AI powered Big Data Analytics is the driving force in creating operational business value and new business models of organizations that have embraced Digital Transformation.

AI is the new Electricity

 “Just as electricity transformed almost everything 100 years ago; today I have a hard time thinking of an industry that AI won’t transform in the next several years.”

– Andrew Ng (Co-founder of Coursera, former Chief Scientist at Baidu, and former member of the Google Brain team).

Cognification:  The process of making objects smarter and smarter by connecting, integrating sensors and building Artificial Intelligence software into them (both locally and cloud-based).

Just as Power Stations and Transmission Lines distributed Electrical Energy to drive machines; Compute Intensive Data Centers, Cloud Based Services and the Internet will make the same machines intelligent, creating a new wave of productivity growth.

AI as a Service (AI-a-a-S)

In the past leveraging AI required expert knowledge and access to super-computer level processing power.  Nowadays anyone with an interest in AI can “spin up” an AI environment complete with advanced tools and be ready to start complex analysis (for example the Microsoft Data Science Virtual Machine).  Leveraging AI-as-a-Service offerings from Microsoft, Google, Amazon, IBM or Apple, with just a few lines of code, any developer can integrate advanced AI capabilities such as: Image recognition; Natural Language Processing (speech recognition); Big Data insights; and predictive analytics solutions.

For business users, vendors such as Intel provide access to pre-packaged cognitive solutions (Intel Saffron), that can identify patterns in data about people, places and things.  This “Democratization of AI” as Microsoft calls it, is rapidly expanding the breadth and depth of AI enabled solutions that are being deployed, which in turn drives a virtuous cycle of even greater demand for AI related cloud-based compute resources and associated services and solutions.

AI as a Tool for Automating IT Operations

Artificial Intelligence will have a profound impact on the IT industry. The same Machine Learning algorithms and models that learn from existing medical reports, and help doctors with diagnosis in the Healthcare industry, can also be applied to improve IT operations. Enterprise IT deals with enormous amounts of data acquired from server, operating system, network, and applications logs. These data sets can be used to develop ML models that assist System Administrators, DevOps Teams, Security and IT Support staff to diagnose and even automatically respond to issues.

Artificial Intelligence will have an enormous impact on the role of IT Support Professionals. From System Log Analysis through to Capacity Planning, and Data Center Energy Utilization.

Many of the issues that are currently escalated to IT Professionals will be able to be (automatically) tackled by intelligent algorithms.  For time-critical activities such as responding to a security breach, AI automation may eventually become the best available solution.

AI-Driven Cloud Infrastructure Monitoring and Management

AI has now become an important tool for (real-time) monitoring and managing Cloud based Resources and Infrastructure.  Here are a few areas of Enterprise IT operations management that AI will significantly impact:

System and Security Log Analysis: Every layer of the IT stack – Hardware, Operating Systems, Network, Servers, and Applications generates a data stream that is collected and stored, and which could potentially be processed, and analyzed using ML algorithms. Currently, this data is typically used by IT Professionals to perform Audit Trails and/or Root Cause Analysis (RCA) of an event (caused either by some type of security breach or system failure). Existing cloud based IT log management platforms such as Splunk, Elasticsearch, Data Dog, and New Relic are now beginning to augment their platforms with Machine Learning. By leveraging AI in log analysis, IT Professionals can proactively pin-point anomalies in the systems in real-time before a failure is reported.  In the near future, these AI enabled monitoring and management systems will become smart enough to self-diagnose and self-heal to recover from failures. Major vendors such as IBM and Cisco are already applying AI technology to develop enhanced security offerings such as IBM’s  QRadar Advisor with Watson  and the Cisco Security Management Portfolio

Performance Tuning:  Once an application is deployed in a production environment, a large amount of time is spent in performance tuning. In particular, database engines that deal with large numbers of transactions often experience reduced performance over time. Database Administrators have to step in to manually drop and rebuild indices and clear the logs to free up space and improve performance.  Almost every workload from web applications, mobile applications, Big Data solutions, and line-of-business applications require some form of tuning to achieve or maintain optimal performance.  Machine Learning algorithms can be applied to auto-tune workloads. By analyzing the logs and the time taken for common tasks such as query processing or responding to a data request, Machine Learning algorithms can apply an appropriate fix to the problem.  By automatically taking action instead of simply escalating the issue to the IT team; it can directly lower the cost of IT support services.

Capacity Planning: IT professionals currently spend a significant amount of time planning the resource needs of applications. It is increasingly difficult to define the server specifications for a complex, multi-tier application deployment, where each physical layer of the application needs to be matched with the number of CPU cores, the amount of RAM, storage capacity and network bandwidth.  In public cloud environments, this results in identifying the right VM type for each tier.  Mature IaaS offerings such as Amazon EC2, Azure VMs and Google Compute Engine currently offer many different VM types, making it a difficult choice; and Cloud Service Providers are regularly adding new VM families to support emerging workloads such as: Big Data; 3D game rendering, Parallel Processing, and Data Warehousing.  Machine Learning algorithms learn from existing deployments and their performance to recommend the optimal configuration for each workload.  In the not too distant future, public Cloud Service Providers will be adding an intelligent VM recommendation engine for each running workload. This move will reduce the burden on IT professionals by assisting them in identifying the right configuration and specifications, and may even eventually obviate the need for detailed planning as the systems become capable of intelligent auto-configuration.

Infrastructure Scaling:  As a result of the inherent scalability of the Cloud, IT administrators can define auto scaling for applications. Auto scaling can be configured to be proactive or reactive. In proactive mode, admins will schedule the scale-out operation before a particular event. For example, if a direct mailer campaign triggered every weekend results in additional load, they can configure the infrastructure to scale-out on a Friday evening and scale-in on Sunday. In reactive mode, the underlying monitoring infrastructure will track key metrics such as CPU utilization and memory usage to initiate a scale-out operation. When the load returns to the normalcy, the scale-in operation takes place bringing back the infrastructure to its original form.  With Machine Learning, IT admins can configure predictive scaling that learns from the previous load conditions and usage patterns. The system will become intelligent enough to decide when to scale with no explicit rules. This design complements capacity planning by adjusting the runtime infrastructure needs more accurately.  Over the next few months, Public Cloud Service Providers are expected to start adding predictive scaling to their IaaS offerings.

Cost Management:  Understanding and assessing the cost of infrastructure plays a crucial role in IT architecture design. This is especially true for Public Cloud Service Providers where cost analysis and forecast is complex, and the potential savings through optimization are large. Cloud Service Providers typically charge for a variety of components including the usage of VMs, storage capacity, IOPS, internal and external bandwidth, and API calls made by applications.  Machine Learning algorithms can be used to accurately forecast the cost of infrastructure. By analyzing the workloads and their usage patterns, it then becomes possible to provide a breakup of the cost across various components, applications, departments, and subscription accounts. This helps business customers to forecast  their Cloud IT budgets much more accurately.  We expect that intelligent cost management will become a standard feature of Public CSP offerings.

Energy Efficiency:  Large enterprises and infrastructure providers are continuing to invest in massive Data Centers. One of the most complex challenges of managing data centers is power management. Increasing energy costs combined with a greater emphasis on environmental responsibility has put pressure on the Data Center industry to improve operational efficiency.  By applying Machine Learning algorithms to power management, Data Center administrators can significantly reduce the energy usage. For example, Google is pioneering AI-driven power management through DeepMind, a UK-based company that the search giant acquired in 2014 for $600 million. Google claims that it has reduced the amount of energy used for cooling by up to 40%.

What are the AI enabled Cloud Solution Opportunities?

To understand more about the business opportunities that AI is creating for Cloud Service Providers, and the broader Ecosystem of Cloud Vendors, Developers and Solution Providers explore the following topics:

How large is the AI market?

So just how large is the AI market opportunity?  Despite all of the current interest in AI, the market for AI is still relatively small. Market research firm Tractica estimated 2016 revenues at just $644 million.

However, like many industry analysts, they expect a “hockey stick” style growth over the next few years, and is predicting the AI market to be worth $15 Billion by 2022; and accelerating even more thereafter.

Reference: Global AI Market Opportunity Tractica, (2016)

In late 2016, there were about 1,500 AI-related startups in the U.S. alone, and according to CB Insights total funding in 2016 reached a record $5 Billion.

Google, Facebook, Microsoft,, Apple and other technology companies are snapping up AI software companies; and large, established companies are also recruiting Deep Learning talent, buying AI companies specializing in their markets.

AI is Transforming Industries …

Which industries are expected to be the earliest adopters of AI?  Generally, the industries that will adopt AI the fastest will be those industries with clear use cases, high purchasing power, and high potential rewards for making decisions quickly and/or more accurately.

Here are the segments that we believe will lead AI through 2020, ordered by market opportunity:


  • Smart Assistants – Personal assistants that anticipate, optimize, and automate daily life (e.g. Amazon Alexa, Apple Siri, Google Assistant, Microsoft Cortana, Facebook Jarvis home automation, ai virtual assistant Amy)
  • AI Chatbots – 24/7/365 no waiting access to an informative or helpful agent (e.g. WeChat, Bank of America, Uber, Pizza Hut, Alaska Airlines, Amtrak, etc.)
  • Intelligent Search – The ability to more intelligently search more data types including image, video, context, etc. (e.g. Improved Google search, Google Photos, ReSnap)
  • Personalization – The ability to automatically adjust content and/or recommendations to suit individuals (e.g. Entefy, Netflix recommendation engine, Amazon personalized shopping recommendations)
  • Augmented Reality – Overlay information on our field of view in real-time to identify interesting or undesirable things (e.g. Intel Project Alloy, Google Translate using smartphone camera)
  • Personal Robots – Personal robots that are capable of performing household, yard, or other chores (e.g. Jibo robot for day-to-day functions, and Roomba follow-ons).


Finance Sector:

  • Algorithmic Trading – Using AI to augment rule-based algorithmic trading models and data sources (e.g. Kensho analysis of myriad data to predict stock movement).
  • Fraud Detection – The ability to identify fraudulent transactions and/or claims (e.g. USAA identifies fraud using Intel’s Saffron technology).
  • Financial Research – The ability to intelligently assemble, parse, and extract meaning from troves of data that influence asset prices (e.g. Quid, FSI firm reducing time to insight for portfolio managers through smart knowledge management system).
  • Personal Finance – Providing smarter (and more personalized) recommendations, lower risk lending, greater efficiency (e.g. active portfolio recommendations, quickly parsing more data before issuing loan, automatic reading of check scans, etc.)
  • Risk Mitigation – AI used to detect and identify risk factors and/or reduce the burden of regulation and minimize errors through automated compliance (e.g. IBM and Promontory Financial Group using Natural Language Processing to detect anomalies)

Retail Sector:

  • Customer Support – AI Bots providing shopping, ordering and support in life like interaction (e.g. My Starbucks Barista, KLM Dutch Airline customer support via social media, Nieman Marcus visual search, Pizza Hut order pizza via bot, Adobe Digital’s digital mirror that recommends clothes, intelligent phone menu routing based on NLP, ViSenze recommending similar items based on image, Adobe Digital’s digital mirror that recommends clothes)
  • Customer Experience – Using AI to deliver enhanced consumer experiences in-store (e.g. Amazon Go checkout-free grocery store, Macy’s mobile shopping assistant, Lowes Lowebots that roam stores answering simple questions and tracking inventory)
  • Targeted Marketing – AI algorithms enable precision marketing to consumers, promoting products and services how and where they want to hear (e.g. North Face “Expert Personal Shopper” on website)
  • Merchandising – Better planning through accelerated and expanded insight into consumer buying patterns (e.g. Stitch Fix virtual styling, analyzing clicks in real-time to bring similar catalog items forward, Walmart pairing products that sell together, Cosabella evolutionary website tweaks)
  • Loyalty Programs – Transforming the consumer experience through segmentation (e.g. Under Armour health app that constantly collects user data to deliver personalized fitness recommendations)
  • Supply Chain Management – Using AI to optimize the supply chain and inventory management for efficiency and innovate new business models (e.g. OnProcess technology’s use of predictive analytics for inventory management)
  • Physical and Digital Security – improve security of all consumer and business digital assets, such as real-time shoplifting/lifter detection, multi-factor identity verification, data breach detection (e.g. Mastercard pay with your face, Walmart facial recognition to catch shoplifters)

Government Sector:

  • Defense – Using AI enabled drones, connected soldiers, defense strategy (e.g. military/surveillance drones, autonomous rescue vehicles, augmented connected soldier, real-time threat assessment and strategy recommendation)
  • Data Insights – Analyze massive amounts of data to identify opportunities/inefficiencies in bureaucracy, cybersecurity threats and more, to ultimately implement better systems and policies (e.g. MIT AI that detects cyber security threats)
  • Crime Prevention – Using AI to predict and help recover from disasters thanks to ability to quickly process large amounts of unstructured data and optimize limited resources (e.g. OneConcern, BlueLineGrid)
  • Safety & Security – Using AI for crowd analytics, behavioral/sentiment analytics, social media analytics, face/vehicle recognition, online identity recognition, real-time video analytics, using AI to predict and help recover from disasters thanks to ability to quickly process large amounts of unstructured data and optimize limited resources (e.g. police analyzing social media to adjust police presence, license plate readers in police cars)
  • Citizen Engagement – New tools to facilitate citizen engagement like chatbots, at-risk citizen identification, (e.g. Amelia chatbot in North London Enfield council, North Carolina chatbot to help state employees with IT inquiries)
  • Smarter Cities – Traffic/pedestrian management, lighting management, weather management, energy conservation, services analytics (e.g. San Francisco and Pittsburgh using sensors and AI to optimize traffic flow)

Energy Sector:

Transport & Logistics Sector:

  • Automated Vehicles – Autonomous cars and buses driving on the roadways (e.g. BMW, Google Waymo, Uber, many others)
  • Automated Trucking – Autonomous trucks driving on the roadways (e.g. Daimler), or within mining sites.
  • Aerospace – Autonomous planes and other aerial vehicles (e.g. Boeing’s evolution of autopilot and drones)
  • Shipping – Autonomous package delivery via drone or other vehicle (e.g. Amazon package delivery drone)
  • Search & Rescue – Ability to deploy autonomous robot to search and rescue victims in potentially hazardous environments (e.g. war casualty extraction, miner rescue, firefighting, avalanche rescue

Industrial Sector:

  • Factory Automation – Highly-productive, efficient and safe factories with robots that can see, hear and adapt to their environment to produce goods with incredible quality and speed (e.g. assembly line)
  • Predictive Maintenance – Ability to detect patterns that indicate the likelihood of an upcoming fault that would require maintenance (e.g. airline being able to adjust schedule to perform preventive maintenance before a failure)
  • Precision Agriculture – Ability to deliver the precise amount of water, nutrients, sunlight, weed killer, etc. to a particular crop or individual plant (e.g. farmer using visual weed search to zap only weeds with RoundUp, automated sorting of produce for market)
  • Field Automation – Ability to automate heavy equipment beyond the factory walls (e.g. mining, excavation, construction, road repair)

Other Areas:

Examples of AI: By Type of AI

There are a variety of AI approaches/methodology which can be used to solve a wide range of problem types.

The Practical Implementation of Deep Learning

Here are a few specific practical examples of how the implementation of Deep Learning algorithms is transforming industries.

Potential Applications of Deep Learning

Reference: A Strategists Guide to Artificial Intelligence, PwC (May, 2017)

Where is the AI market opportunity?

According to IDC, the Cognitive Computing / AI use cases that will see the greatest levels of investment this year are:

  • Quality Management Investigation and Recommendation Systems;
  • Diagnosis and Treatment Systems;
  • Automated Customer Service Agents;
  • Automated Threat Intelligence and Prevention Systems; and
  • Fraud Analysis and Investigation.

Combined, these use cases will deliver nearly half of all cognitive/AI systems spending in 2017.  The use cases that are forecast to experience the fastest spending growth over the 2015-2020 forecast period are Public Safety and Emergency Response (85.5% CAGR) and Pharmaceutical Research and Discovery (74.2% CAGR)

AI opportunity Heat Maps

AI Start-Up Investment Heat Map

Analysis of investment in AI related start-ups provides a measure of determining which are potentially the best opportunities for AI. Based on a CBInsights State of AI analysis (May 2017), the key areas for active investment in AI startups include:

  • Financial Technology and Insurance
  • HealthCare
  • Horizontal Applications (Industry Agnostic: e.g. Image Recognition; AI ChatBots)
  • Other (Sales Automation, CRM, Cyber-Security etc.)

Who is investing in and who is implementing AI?

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

How does AI create business value?

Artificial Intelligence can create business value across the value chain in four ways (Project, Produce, Promote, and Provide).

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

AI Investment by Type of AI

External investment in AI-focused companies by technology category, 2016 ($US Billion)

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

According to the McKinsey Global Institute, Machine Learning received the most investment of any area of AI, although the boundaries between the different AI technologies are not always clear-cut.

How AI generates Business Value

Here are some industry specific examples of how AI generates Business Value:

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

AI Industry Adoption Heat Map

Sector-by-sector industry adoption of AI is currently highly uneven, reflecting the broader pattern of adoption of digital technologies in general. The McKinsey Global Institute survey found that larger companies and industries that were earlier adopters of digital technologies in the past, are more likely to be early adopters of AI. For them, AI is the next wave of digitization.

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

The MGI Digitization Index is a composite of 27 indicators that fall into three broad categories: Digital Assets, Digital Usage, and Digital Workers.

The MGI Digitization Index is GDP weighted average of Europe and United States.

Metrics included in the industry AI index

The McKinsey Global Institute AI index measures the extent of AI adoption and usage in 13 sectors across the 10 countries in the survey. (The telecommunications and high tech sectors were merged to align with the MGI Industry Digitization Index.) The index is based on 16 input metrics, divided into three categories: AI assets (three metrics), AI usage (11 metrics), and AI-enabled labor (two metrics). Using principal component analysis, the input metrics were combined into an overall AI adoption score. The data for these metrics were primarily obtained from the AI adoption and use survey, proprietary databases, and the MGI Industry Digitization Index.

AI Industry Adoption Forecast (MGI)

The McKinsey Global Institute survey found that larger companies and industries that were earlier adopters of digital technologies in the past, are more likely to be early adopters of AI. For them, AI is the next wave of digitization.

It is expected that large companies with the most digital experience will be the first movers because they are best able to leverage their technical skills, digital expertise, and data resources to develop and smoothly integrate the most appropriate AI solutions.

Sectors leading in AI adoption today are expected to grow their investment the most in the future.

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

Forecasting the AI applications opportunity

There are three basic forms of AI: Assisted (AI that improves what your business is already doing); Augmented (AI that it enables your business to do things that it could not otherwise do); and Autonomous (AI that acts on its own, choosing actions on behalf of your business goals).

Here are the estimated dates of commercial availability for products and services incorporating the three forms of Artificial Intelligence (Assisted, Augmented, Autonomous).

Reference: Artificial Intelligence: The next Digital Frontier?, MGI (June 2017)

Leveraging AI in your Business

Here are some suggestions on how you can best leverage Artificial Intelligence in your own business.

1.     Develop an AI strategy aligned with your business strategy

  • Integrate AI into your existing digital and analytics plans
  • Decide which businesses to disrupt and which to enhance
  • Consider new business models based on improved productivity
  • Plan long-term investments in autonomous intelligence

2.     Create an Enterprise-wide AI capability

  • Redesign products and services to incorporate Machine Learning
  • Use AI to upgrade your most critical distinctive capabilities
  • Use automation to improve your current decision making processes
  • Automate your existing business processes and/or develop new ones
  • Recruit staff who understand AI

3.     Institutionalize your portfolio of AI capabilities

  • Embed AI throughout your business processes
  • Embrace Cloud platforms and specialized AI hardware
  • Foster a culture of decision making that is open to ideas from AI

4.     Ensure appropriate governance

  • Establish clear policies with respect to data privacy, decision rights, and transparency
  • Set up governance structures to monitor possible errors and problems (for example, overreach in automated program trading)
  • Set up communications practices to explain AI-related decisions
  • Consider the impact on employment and invest in developing the workforce that AI will complement

Implications for Software Developers, IT and Business Decision Makers

So what does this all mean for Software Developers, IT Professionals and Business Decision Makers?

  1. The industries and organizations that have already embraced Digital Transformation, are in the best position to leverage Artificial Intelligence; and are likely to be the early adopters of advanced AI tools, AI technologies, and AI enabled solutions.
  2. Some industries, for example in Healthcare, are large and stand to benefit enormously from AI solutions, and have a supply of interested early adopters.  However, they are often slower to adopt the technology solutions more broadly. You will therefore need to focus your attention on working with the identified early adopters, and be prepared for a longer wait before the solutions that you develop become more broadly deployed.
  3. You first need to fully embrace AI opportunities within your own organization by having an AI strategy that is aligned to your business; before you can embark on developing and leveraging AI based  solutions for your customers.

What’s Next in AI?

So what can we expect from AI in the near future?

China’s AI Boom

China has identified AI and machine learning as the next big areas of innovation.  Baidu, has had an AI focused lab for some time, and it is reaping the rewards in terms of improvements in technologies such as voice recognition and Natural Language Processing, as well as a better-optimized advertising business.  Tencent opened an AI lab in 2016. Didi, the ride-sharing giant that bought Uber’s Chinese operations earlier this year, is also building out a lab in Silicon Valley and is working on its own driverless cars.

Chinese investors are now pouring money into AI-focused startups, and the Chinese government has signaled a desire to see the country’s AI industry blossom, pledging to invest about $15 billion by 2018.

Economic Considerations

The economic impact of increased AI Automation will be the subject of significant media discussion.

Expect to hear less about malevolent “AI taking over the world”, and more about quantifying the economic impacts of AI. Concerns about AI stealing jobs are nothing new, but should give way to deeper, more nuanced conversations on what AI will mean economically in terms of improving productivity and driving GDP growth.

Expect to see a lot more specialized AI systems.

Don’t expect to see large, general purpose AI systems, at least not in the short-term. Rather we anticipate an explosion in special purpose targeted AI systems, in domains such as:

  • Robotics: Personal, Industrial, and Retail
  • Autonomous vehicles: (Trucks, Buses, Drones, and Cars)
  • Bots: AI enabled Chat-bots and Virtual Assistants in Customer Relationship Management and Customer Engagement Solutions
  • Industry-specific AI: for Finance Services (e.g. Anti-Money Laundering and Fraud Detection), Health (Diagnosis), Security (Real-time Intrusion Detection, and Retail (Customer Analytics).

Human-machine interaction will become richer

In machine intelligence, there is a spectrum that ranges from pure machine intelligence to human augmentation. Advances in emotional intelligence and detection and in human-assisted solutions will allow for a richer interaction between human and machine intelligence.

The rise of natural user interfaces (voice, chat and vision) provide very interesting options and opportunities for humans to interact with virtual assistants (Apple Siri, Amazon Alexa, and Microsoft Cortana). With AI expanding upon the role of Curator, to Advisor, and eventually Orchestrator.

Emerging AI related Ecosystems

There are a number of emerging AI related ecosystems that have become the new “battle ground” for major vendors.  

In particular: Autonomous Vehicles; AI Chatbots; Connected Devices (such as Google Home and Amazon Dot/Echo), and the Connected Home; Mobile devices and intelligent mobile applications; AR/VR solutions; and the broader Developer community that is just beginning to engage on AI related solutions development leveraging Cloud based API’s from Microsoft, Amazon, Google, IBM and Apple.

It is important to understand the characteristics and importance of these ecosystems so that you are able to navigate your way through the challenges and opportunities that they represent.

Addressing AI’s “Dark Secret”: Algorithmic Transparency

Deep learning is responsible for today’s explosion of AI. It has given computers extraordinary powers, like the ability to recognize spoken words almost as well as a person could, a skill too complex to code into the machine by hand. Deep learning has transformed computer vision and dramatically improved machine translation. It is now being used to guide all sorts of key decisions in medicine, finance, manufacturing—and beyond.  Unfortunately, the workings of any Machine Learning technology are inherently more opaque, even to computer scientists, than a hand-coded (rules based) system. By its very nature, Deep Learning is a particularly dark black box.

Explainability is a key stumbling block to the broader acceptance of Machine Learning methods.

Understanding the reasoning behind an output (Algorithmic Transparency) is also going to be crucial if AI technology is to become a common and useful part of our daily lives. Although there is a lot of research attempting to interpret how Machine Learning algorithms generate specific results, at some stage we may have to simply trust AI’s judgment or do without using it. Just as society is built upon a contract of expected behavior, AI systems will need to be designed to respect and fit in with established social norms.

The explainability of Deep Learning algorithms is currently an active area of research within the AI community.


Background Resources

The following links provide relevant background information on the history of Artificial Intelligence; a summary of the key open source AI software and tools; together with formal definitions of AI terminology and examples of commonly used AI categories and classifications.

A Brief History of Artificial Intelligence (AI)

The origins of AI: from 1950 to 2017

The concept of Artificial Intelligence is not new, in fact it was first raised in the 1950’s in a research paper by mathematician Alan Turing entitled “Computing Machinery & Intelligence“, although it existed in science fiction literature much earlier.

Throughout the 60’s, 70’s, 80’s and 90’s AI was an active area of academic research, and there were a few isolated AI projects that achieved fame.

In the 1960’s, the First International Joint Conference on Artificial Intelligence was held at Stanford University in California, USA.  This decade also saw the first (text based) program ELIZA that was able to carry out an interactive (typed) dialogue in English.

Date   AI Development
1960s Ray Solomonoff lays the foundations of a mathematical theory of AI, introducing universal Bayesian methods for inductive inference and prediction.
1961 Unimation‘s industrial robot Unimate worked on a General Motors automobileassembly line.
1963 Thomas Evans’ program, ANALOGY, written as part of his PhD work at MIT, demonstrated that computers can solve the same analogy problems as are given on IQ tests.
1963 Edward Feigenbaum and Julian Feldman published Computers and Thought, the first collection of articles about artificial intelligence.
1964 Danny Bobrow’s dissertation at MIT (technical report #1 from MIT’s AI group, Project MAC), shows that computers can understand natural language well enough to solve algebra word problems correctly.
1965 Joseph Weizenbaum (MIT) built ELIZA, an interactive program that carries on a dialogue in English language on any topic. It was a popular toy at AI centers on the ARPANET when a version that “simulated” the dialogue of a psychotherapist was programmed.
1965 Edward Feigenbaum initiated Dendral, a ten-year effort to develop software to deduce the molecular structure of organic compounds using scientific instrument data. It was the first expert system.
1966 Ross Quillian (PhD dissertation, Carnegie Inst. of Technology, now CMU) demonstrated semantic nets.
1966 Machine Intelligence workshop at Edinburgh – the first of an influential annual series organized by Donald Michie and others.
1967 Dendral program (Edward Feigenbaum, Joshua Lederberg, Bruce Buchanan, Georgia Sutherland at Stanford University) demonstrated to interpret mass spectra on organic chemical compounds. First successful knowledge-based program for scientific reasoning.
1968 Joel Moses (PhD work at MIT) demonstrated the power of symbolic reasoningfor integration problems in the Macsyma program. First successful knowledge-based program in mathematics.
1968 Richard Greenblatt (programmer) at MIT built a knowledge-based chess-playing programMacHack, that was good enough to achieve a class-C rating in tournament play.
1969 Stanford Research Institute (SRI): Shakey the Robot, demonstrated combining animal locomotionperception and problem solving.
1969 Roger Schank (Stanford) defined conceptual dependency model for natural language understanding.
1969 First International Joint Conference on Artificial Intelligence (IJCAI) held at Stanford University.

The 1970’s saw the introduction of the PROLOG programming language, a Language for Logic Programming and Symbolic Computation; which along with LISP (short for List Processing), became two of the most important programming languages used in Artificial Intelligence in this era.    The 1970’s also saw the introduction of the first “Autonomous Vehicle” … a mobile cart that was able to navigate a chair filled room.

Date    AI Development
Early 1970s Jane Robinson and Don Walker established an influential Natural Language Processing group at SRI.
1970 Jaime Carbonell (Sr.) developed SCHOLAR, an interactive program for computer assisted instruction based on semantic nets as the representation of knowledge.
1971 Terry Winograd‘s PhD thesis (MIT) demonstrated the ability of computers to understand English sentences in a restricted world of children’s blocks, in a coupling of his language with a robot arm that carried out instructions typed in English.
1972 Prolog programming language developed by Alain Colmerauer.
1973 The Assembly Robotics Group at University of Edinburgh builds Freddy Robot, capable of using visual perception to locate and assemble models. (See Edinburgh Freddy Assembly Robot: a versatile computer-controlled assembly system.)
1973 The Lighthill report gives a largely negative verdict on AI research in Great Britain and forms the basis for the decision by the British government to discontinue support for AI research in all but two universities.
1974 Ted Shortliffe‘s PhD dissertation on the MYCIN program (Stanford) demonstrated a very practical rule-based approach to medical diagnoses, even in the presence of uncertainty. While it borrowed from DENDRAL, its own contributions strongly influenced the future of expert system development.
1975 Marvin Minsky published his widely read and influential article on Frames as a representation of knowledge, in which many ideas about schemas and semantic links are brought together.
1975 The Meta-Dendral learning program produced new results in chemistry (some rules of mass spectrometry) the first scientific discoveries by a computer to be published in a refereed journal.
Mid-1970s Barbara Grosz (SRI) established limits to traditional AI approaches to discourse modeling. Subsequent work by Grosz, Bonnie Webber and Candace Sidner developed the notion of “centering”, used in establishing focus of discourse and anaphoric references in Natural language processing.
1978 Herbert A. Simon wins the Nobel Prize in Economics for his theory of bounded rationality, one of the cornerstones of AI known as “satisficing“.
1978 The MOLGEN program, written at Stanford by Mark Stefik and Peter Friedland, demonstrated that an object-oriented programming representation of knowledge can be used to plan gene-cloning experiments.
1979 Bill VanMelle’s PhD dissertation at Stanford demonstrated the generality of MYCIN‘s representation of knowledge and style of reasoning in his EMYCIN program, the model for many commercial expert system “shells”.
1979 Jack Myers and Harry Pople at University of Pittsburgh developed INTERNIST, a knowledge-based medical diagnosis program based on Dr. Myers’ clinical knowledge.
1979 The Stanford Cart, built by Hans Moravec, becomes the first computer-controlled, autonomous vehicle when it successfully traverses a chair-filled room and circumnavigates the Stanford AI Lab.
1979 BKG, a backgammon program written by Hans Berliner at CMU, defeats the reigning world champion.
Late 1970s Stanford’s SUMEX-AIM resource, headed by Ed Feigenbaum and Joshua Lederberg, demonstrates the power of the ARPAnet for scientific collaboration.

In the 1980’s we saw the introduction of dedicated LISP machines for AI computation, as well as the development of practical Expert System shells and associated commercial applications.   Autonomous vehicles had also improved, with the University of Munich building the a robot car that drove at speeds of 55 mph on (understandably) empty streets

Date   AI Development
1980s Lisp machines developed and marketed. First expert system shells and commercial applications.
1980 First National Conference of the American Association for Artificial Intelligence (AAAI) held at Stanford.
1981 Danny Hillis designs the connection machine, which utilizes Parallel computing to bring new power to AI, and to computation in general. (Later founds Thinking Machines Corporation)
1982 The Fifth Generation Computer Systems project (FGCS), an initiative by Japan’s Ministry of International Trade and Industry, begun in 1982, to create a “fifth generation computer” (see history of computing hardware) which was supposed to perform much calculation utilizing massive parallelism.
1983 John Laird and Paul Rosenbloom, working with Allen Newell, complete CMU dissertations on Soar (program).
1983 James F. Allen invents the Interval Calculus, the first widely used formalization of temporal events.
Mid-1980s Neural Networks become widely used with the Backpropagation algorithm (first described by Paul Werbos in 1974).
1985 The autonomous drawing program, AARON, created by Harold Cohen, is demonstrated at the AAAI National Conference (based on more than a decade of work, and with subsequent work showing major developments).
1986 The team of Ernst Dickmanns at Bundeswehr University of Munich builds the first robot cars, driving up to 55 mph on empty streets.
1986 Barbara Grosz and Candace Sidner create the first computation model of discourse, establishing the field of research.
1987 Marvin Minsky published The Society of Mind, a theoretical description of the mind as a collection of cooperating agents. He had been lecturing on the idea for years before the book came out (c.f. Doyle 1983).
1987 Around the same time, Rodney Brooks introduced the subsumption architecture and behavior-based robotics as a more minimalist modular model of natural intelligence; Nouvelle AI.
1987 Commercial launch of generation 2.0 of Alacrity by Alacritous Inc./Allstar Advice Inc. Toronto, the first commercial strategic and managerial advisory system. The system was based upon a forward-chaining, self-developed expert system with 3,000 rules about the evolution of markets and competitive strategies and co-authored by Alistair Davidson and Mary Chung, founders of the firm with the underlying engine developed by Paul Tarvydas. The Alacrity system also included a small financial expert system that interpreted financial statements and models.
1989 Dean Pomerleau at CMU creates ALVINN (An Autonomous Land Vehicle in a Neural Network).

In the 1990’s we see AI being successfully deployed as part of an advanced logistics scheduling application during the Gulf War … and generating enough value to pay back the USA Defense Advance Research Project Agency’s prior 30 year investment in AI research.

This decade also saw the release of the first consumer AI product, the Tiger Electronic’s Furby.

Date   AI Development
Early 1990s TD-Gammon, a backgammon program written by Gerry Tesauro, demonstrates that reinforcement (learning) is powerful enough to create a championship-level game-playing program by competing favorably with world-class players.
1990s Major advances in all areas of AI, with significant demonstrations in machine learning, intelligent tutoring, case-based reasoning, multi-agent planning, scheduling, uncertain reasoning, data mining, natural language understanding and translation, vision, virtual reality, games, and other topics.
1991 DART scheduling application deployed in the first Gulf War paid back DARPA’sinvestment of 30 years in AI research.
1993 Ian Horswill extended behavior-based robotics by creating Polly, the first robot to navigate using vision and operate at animal-like speeds (1 meter/second).
1993 Rodney BrooksLynn Andrea Stein and Cynthia Breazeal started the widely publicized MIT Cog project with numerous collaborators, in an attempt to build a humanoid robot child in just five years.
1993 ISX corporation wins “DARPA contractor of the year” for the Dynamic Analysis and Replanning Tool (DART) which reportedly repaid the US government’s entire investment in AI research since the 1950s.
1994 With passengers on board, the twin robot cars VaMP and VITA-2 of Ernst Dickmanns and Daimler-Benz drive more than one thousand kilometers on a Paris three-lane highway in standard heavy traffic at speeds up to 130 km/h. They demonstrate autonomous driving in free lanes, convoy driving, and lane changes left and right with autonomous passing of other cars.
1994 English draughts (checkers) world champion Tinsley resigned a match against computer program Chinook. Chinook defeated 2nd highest rated player, Lafferty. Chinook won the USA National Tournament by the widest margin ever.
1995 “No Hands Across America”: A semi-autonomous car drove coast-to-coast across the United States with computer-controlled steering for 2,797 miles (4,501 km) of the 2,849 miles (4,585 km). Throttle and brakes were controlled by a human driver.
1995 One of Ernst Dickmanns‘ robot cars (with robot-controlled throttle and brakes) drove more than 1000 miles from Munich to Copenhagen and back, in traffic, at up to 120 mph, occasionally executing maneuvers to pass other cars (only in a few critical situations a safety driver took over). Active vision was used to deal with rapidly changing street scenes.
1997 The Deep Blue chess machine (IBM) defeats the (then) world chesschampion, Garry Kasparov.
1997 First official RoboCup football (soccer) match featuring table-top matches with 40 teams of interacting robots and over 5000 spectators.
1997 Computer Othello program Logistello defeated the world champion Takeshi Murakami with a score of 6–0.
1998 Tiger Electronics‘ Furby is released, and becomes the first successful attempt at producing a type of A.I to reach a domestic environment.
1998 Tim Berners-Lee published his Semantic Web Road map paper.
1998 Leslie P. KaelblingMichael Littman, and Anthony Cassandra introduce the first method for solving POMDPs offline, jumpstarting widespread use in robotics and automated planning and scheduling
1999 Sony introduces an improved domestic robot similar to a Furby, the AIBO becomes one of the first artificially intelligent “pets” that is also autonomous.
Late 1990s Web crawlers and other AI-based information extraction programs become essential in widespread use of the World Wide Web.

In the 2000’s we saw the introduction of “Smart Toys”, interactive robo-pets that could react and learn new behaviors.

This decade also saw the launch of DARPA’s Urban Challenge for autonomous vehicles; and with Google building its first self-driving car.

Date   AI Development
2000 Interactive robopets (“smart toys“) become commercially available, realizing the vision of the 18th century novelty toy makers.
2000 Cynthia Breazeal at MIT publishes her dissertation on Sociable machines, describing Kismet (robot), with a face that expresses emotions.
2000 The Nomad robot explores remote regions of Antarctica looking for meteorite samples.
2002 iRobot‘s Roomba autonomously vacuums the floor while navigating and avoiding obstacles.
2004 OWL Web Ontology Language W3C Recommendation (10 February 2004).
2004 DARPA introduces the DARPA Grand Challenge requiring competitors to produce autonomous vehicles for prize money.
2004 NASA‘s robotic exploration rovers Spirit and Opportunity autonomously navigate the surface of Mars.
2005 Honda‘s ASIMO robot, an artificially intelligent humanoid robot, is able to walk as fast as a human, delivering trays to customers in restaurant settings.
2005 Recommendation technology based on tracking web activity or media usage brings AI to marketing. See TiVo Suggestions.
2005 Blue Brain is born, a project to simulate the brain at molecular detail.
2006 The Dartmouth Artificial Intelligence Conference: The Next 50 Years (AI@50) AI@50 (14–16 July 2006)
2007 Philosophical Transactions of the Royal Society, B – Biology, one of the world’s oldest scientific journals, puts out a special issue on using AI to understand biological intelligence, titled Models of Natural Action Selection
2007 Checkers is solved by a team of researchers at the University of Alberta.
2007 DARPA launches the Urban Challenge for autonomous cars to obey traffic rules and operate in an urban environment.
2009 Google builds self driving car.

In the current decade we have seen the introduction of Machine Learning and Deep Learning algorithms, initially running on super-computers, with IBM’s Watson computer defeating the Jeopardy Game show champions in 2011; followed by more recently Google’s DeepMind AlphaGo defeating the world Go champion Ke Jie.

Date  AI Development
2010 Microsoft launched Kinect for Xbox 360, the first gaming device to track human body movement, using just a 3D camera and infra-red detection, enabling users to play their Xbox 360 wirelessly. The award-winning machine learning for human motion capture technology for this device was developed by the Computer Vision group at Microsoft Research, Cambridge.
2011 IBM‘s Watson computer defeated television game show Jeopardy! champions Rutter and Jennings.
2011 Apple‘s SiriGoogle‘s Google Now and Microsoft‘s Cortana are smartphoneapps that use natural language to answer questions, make recommendations and perform actions.
2013 Robot HRP-2 built by SCHAFT Inc of Japan, a subsidiary of Google, defeats 15 teams to win DARPA’s Robotics Challenge Trials. HRP-2 scored 27 out of 32 points in 8 tasks needed in disaster response. Tasks are drive a vehicle, walk over debris, climb a ladder, remove debris, walk through doors, cut through a wall, close valves and connect a hose.
2013 NEIL, the Never Ending Image Learner, is released at Carnegie Mellon University to constantly compare and analyze relationships between different images.
2015 An open letter to ban development and use of autonomous weapons signed by HawkingMuskWozniak and 3,000 researchers in AI and Robotics.
2015 Google DeepMind‘s AlphaGo defeated 3 time European Go champion 2 dan professional Fan Hui by 5 games to 0.
2016 Google DeepMind‘s AlphaGo defeated Lee Sedol 4-1. Lee Sedol is a 9 dan professional Korean Go champion who won 27 major tournaments from 2002 to 2016.  Before the match with AlphaGo, Lee Sedol was confident in predicting an easy 5-0 or 4-1 victory.
2017 Google DeepMind‘s AlphaGo won 60-0 rounds on two public Go websites including 3 wins against world Go champion Ke Jie.
2017 Libratus, designed by Carnegie Mellon professor Tuomas Sandholm and his graduate student Noam Brown won against four top players at no-limit Texas hold ’em, a very challenging version of poker. Unlike Go and Chess, Poker is a game in which some information is hidden (the cards of the other player) which makes it much harder to model.

The State of AI Today … the Good Headlines

So where are we with AI today?  Let’s take a look at some of the (good) headlines.  The promise of AI is being realized in fields as disparate as Health & Medicine, Security Systems, Language Analysis & Creative Writing; and even Scientific discovery.

The State of AI Today … the Bad Headlines

However, not all of the headline news about AI is good.  There is growing concern about the potential harm of AI and Robotics; from replacing jobs … to replacing humans!

Although there are some justifiable concerns about the ethical issues of relying on AI in “life and death” decision scenarios; the overall impact of AI is that it is more likely to usher in a new wave of innovative services and improved productivity & efficiency, that will ultimately benefit mankind.

Key Open Source AI Software and Tools

The importance of Open Source Software and Tools to the development of AI.

Largely as a result of AI’s rich heritage as an area of academic research over the past 50 years, much of the AI algorithms, software and associated tools have been developed by the Open Source Community and/or available in the public domain.  This has had a democratizing effect, allowing smaller organizations and individuals to build intelligent applications leveraging the most advanced tools.

Let us now look at some of the more commonly used Open Source AI software and tools.

Distributed Machine Learning Toolkit

Distributed Machine Learning Toolkit (DMTK) is one of Microsoft’s open source artificial intelligence tools. Designed for use in big data applications, it aims to make it faster to train AI systems. It consists of three key components: the DMTK framework, the LightLDA topic model algorithm, and the Distributed (Multisense) Word Embedding algorithm. As proof of DMTK’s speed, Microsoft says that on an eight-cluster machine, it can “train a topic model with 1 million topics and a 10-million-word vocabulary (for a total of 10 trillion parameters), on a document collection with over 100-billion tokens,” a feat that is unparalleled by other tools.

Microsoft Cognitive Toolkit (CNTK)

Computational Network Toolkit, CNTK is one of Microsoft’s open source artificial intelligence tools. It boasts outstanding performance whether it is running on a system with only CPUs, a single GPU, multiple GPUs or multiple machines with multiple GPUs. Microsoft has primarily utilized it for research into speech recognition, but it is also useful for applications like machine translation, image recognition, image captioning, text processing, language understanding and language modeling.


Deeplearning4j is an open source deep learning library for the Java Virtual Machine (JVM). It runs in distributed environments and integrates with both Hadoop and Apache Spark. It makes it possible to configure deep neural networks, and it’s compatible with Java, Scala and other JVM languages. The project is managed by a commercial company called Skymind, which offers paid support, training and an enterprise distribution of Deeplearning4j.


Caffe is a deep learning framework based on expressive architecture and extensible code. It’s claim to fame is its speed, which makes it popular with both researchers and enterprise users. Caffe can process more than 60 million images in a single day using just one NVIDIA K40 GPU. It is managed by the Berkeley Vision and Learning Center (BVLC), and companies like NVIDIA and Amazon have made grants to support its development.


Focused more on enterprise uses for AI than on research, H2O has large companies like Capital One, Cisco, Nielsen Catalina, PayPal and Transamerica among its users. It claims to make is possible for anyone to use the power of machine learning and predictive analytics to solve business problems. It can be used for predictive modeling, risk and fraud analysis, insurance analytics, advertising technology, healthcare and customer intelligence. It comes in two open source versions: standard H2O and Sparkling Water, which is integrated with Apache Spark. Paid enterprise support is also available.


An Apache Foundation project, Mahout is an open source machine learning framework. According to its website, it offers three major features: a programming environment for building scalable algorithms, premade algorithms for tools like Spark and H2O, and a vector-math experimentation environment called Samsara. Companies using Mahout include Adobe, Accenture, Foursquare, Intel, LinkedIn, Twitter, Yahoo and many others. Professional support is available through third parties listed on the website.


MLlib is Apache Spark’s scalable machine learning library. It integrates with Hadoop and interoperates with both NumPy and R. It includes a host of machine learning algorithms for classification, regression, decision trees, recommendation, clustering, topic modeling, feature transformations, model evaluation, ML pipeline construction, ML persistence, survival analysis, frequent itemset and sequential pattern mining, distributed linear algebra and statistics.


Managed by Numenta, NuPIC is an open source artificial intelligence project based on a theory called Hierarchical Temporal Memory, or HTM. Essentially, HTM is an attempt to create a computer system modeled after the human neocortex. The goal is to create machines that “approach or exceed human level performance for many cognitive tasks.” In addition to the open source license, Numenta also offers NuPic under a commercial license, and it also offers licenses on the patents that underlie the technology.


Developed by Cycorp, OpenCyc provides access to the Cyc knowledge base and commonsense reasoning engine. It includes more than 239,000 terms, about 2,093,000 triples, and about 69,000 links to external semantic data namespaces. It is useful for rich domain modeling, semantic data integration, text understanding, domain-specific expert systems and game AIs. The company also offers two other versions of Cyc: one for researchers that is free but not open source, and one for enterprise use that requires a fee.


Designed for researchers and developers with advanced understanding of artificial intelligence, OpenNN is a C++ programming library for implementing neural networks. Its key features include deep architectures and fast performance. Extensive documentation is available on the website, including an introductory tutorial that explains the basics of neural networks. Paid support for OpenNN is available through Artelnics, a Spain-based firm that specializes in predictive analytics.

Oryx 2

Built on top of Apache Spark and Kafka, Oryx 2 is a specialized application development framework for large-scale machine learning. It utilizes a unique lambda architecture with three tiers. Developers can use Oryx 2 to create new applications, and it also includes some pre-built applications for common big data tasks like collaborative filtering, classification, regression and clustering. The big data tool vendor Cloudera created the original Oryx 1 project and has been heavily involved in continuing development.


Salesforce bought PredictionIO, and then in July 2016, it contributed the platform and its trademark to the Apache Foundation, which accepted it as an incubator project. So while Salesforce is using PredictionIO technology to advance its own machine learning capabilities, work will also continue on the open source version. It helps users create predictive engines with machine learning capabilities that can be used to deploy Web services that respond to dynamic queries in real time.


Originally developed by IBM, SystemML is now an Apache big data project. It offers a highly-scalable platform that can implement high-level math and algorithms written in R or a Python-like syntax. Enterprises are already using it to track customer service on auto repairs, to direct airport traffic and to link social media data with banking customers. It can run on top of Spark or Hadoop.


An open-source software library for Machine Intelligence. One of Google’s open source artificial intelligence tools. It offers a library for numerical computation using data flow graphs. It can run on a wide variety of different systems with single- or multi-CPUs and GPUs and even runs on mobile devices. It boasts deep flexibility, true portability, automatic differential capabilities and support for Python and C++. The website includes a very extensive list of tutorials and how-tos for developers or researchers interested in using or extending its capabilities.


Theano is a Python library that lets you to define, optimize, and evaluate mathematical expressions, especially ones with multi-dimensional arrays (numpy.ndarray). Using Theano it is possible to attain speeds rivaling hand-crafted C implementations for problems involving large amounts of data.

Theano can also surpass C on a CPU by many orders of magnitude by taking advantage of GPUs.

Theano requirements

  • git: A distributed revision control system (RCS).
  • nosetests: A system for unit tests.
  • numpy: A library for efficient numerical computing.
  • python: The programming language Theano is for.
  • scipy: A library for scientific computing.

Anaconda (Continuum Analytics)

Anaconda (“Anaconda Distribution”) is a free, easy-to-install package manager, environment manager, Python distribution, and collection of over 1000 open source packages with free community support.

With over 4.5 million users, Anaconda is the world’s most popular Data Science ecosystem, and is supported by a wide variety of vendors including Intel, IBM, and Microsoft.

Continuum Analytics offers products and services that help support, govern, scale, assure, customize and secure Anaconda for enterprise deployment.

The Anaconda Ecosystem:

The Microsoft Data Science Virtual Machine

A Virtual Machine with all the tools you need for Data Science modeling and development.

The Microsoft Data Science Virtual Machine runs either on Windows Server or Linux and contains popular tools for data exploration, modeling and development activities.

The main tools included are Microsoft R Server Developer Edition (An enterprise ready scalable R framework), Anaconda Python distribution, Julia Pro developer edition, Jupyter notebooks for R, Python and Julia, Visual Studio Community Edition with Python, R and node.js tools, Power BI desktop, SQL Server 2016 Developer edition including support In-Database analytics using Microsoft R Server (Windows version only).

It also includes open source deep learning tools like Microsoft Cognitive Toolkit (CNTK 2.0) and mxnet; ML algorithms like xgboost, Vowpal Wabbit. The Azure SDK and libraries on the VM allows you to build your applications using various services in the cloud that are part of the Cortana Analytics Suite which includes Azure Machine Learning, Azure data factory, Stream Analytics and SQL Datawarehouse, Hadoop, Data Lake, Spark and more.

You can deploy models as web services in the cloud on Azure Machine Learning OR deploy them either on the cloud or on-premises using the Microsoft R Server operationalization.

Intel DAAL (Data Analytics Acceleration Library)

The Intel Data Analytics Acceleration Library (DAAL) is an open source High Performance Machine Learning and Data Analytics library that provides the building blocks for all stages of the data analytics process: Pre-processing and Transformation of Data, Data Analysis, Modeling and Validation, as well as Decision making tools.   Intel DAAL incorporates data preparation, Data Mining, and Machine Learning tools.

Key features include:

  • Open Source Apache 2.0 License
  • Common Python, Java and C++ APIs across all Intel hardware
  • Optimized for large data sets including streaming and distributed processing
  • Flexible interfaces to leading big data platforms including Spark and range of data formats (CSV, SQL, etc.)

Intel Development Tools for Artificial Intelligence

Here is a summary of Intel Development Tools for Artificial Intelligence applications:

Intel also provides access to open-source frameworks for Machine Learning and Deep Learning, as well as code and reference architectures for distributed training and scoring.

For more information check out the hyperlinks to the Open Source projects in the supporting documentation, or explore Intel’s support for Open Source AI tools on their web-site.


AI Defined: Terminology and Common Used Classifications

Explanation of Key AI Terms

Artificial Intelligence

Artificial Intelligence (AI) is the simulation of human intelligence processes by machines (computers). AI processes include learning (the acquisition of information and rules for using the information), reasoning (using the rules to reach approximate or definite conclusions), and self-correction. Specific applications of AI include Expert Systems, speech and language recognition (Natural Language Processing), and Machine Vision.

Artificial General Intelligence

Artificial General Intelligence (AGI) is the intelligence of a machine that could successfully perform any intellectual task that a human being can. It is a primary (but unattained) goal of some areas of Artificial Intelligence research, and is a common topic in science fiction and future studies. Artificial General Intelligence is also referred to as “strong AI“, “full AI” or as the ability of a machine to perform “general intelligent action“.

Machine Learning

Machine Learning is a type of Artificial Intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. Machine Learning focuses on the development of computer programs that can change when exposed to new data.   The process of Machine Learning is similar to that of data mining. Both systems search through data to look for patterns. However, instead of extracting data for human comprehension — as is the case in data mining applications — Machine Learning uses that data to detect patterns in data and adjust program actions accordingly.  Machine Learning algorithms are often categorized as being supervised or unsupervised. Supervised algorithms can apply what has been learned in the past to new data. Unsupervised algorithms can draw inferences from data sets.

Deep Learning / Deep Neural Networks

Deep Learning is a sub-discipline of Machine Learning concerned with algorithms inspired by the structure and function of the brain (often referred to as Artificial Neural Networks).  A Deep Neural Network (DNN) is an Artificial Neural Network (ANN), that has multiple hidden layers of units between the input and output layers, which allows DNNs to model complex non-linear relationships.  The term “Deep” is in reference to the number of layers.  Conventional Machine Learning algorithms often plateau on analytics performance after processing a certain amount of data. The reason is that when an algorithm is directed to look for correlations among specific variables, those correlations become apparent fairly quickly. The performance of Deep Learning algorithms, often improves exponentially when given more training data to analyze. They take a neural-network approach to look for patterns and correlations that can be more, and that become clearer only with the use of more data.

Cognitive Computing

Cognitive computing is the simulation of human thought processes in a computerized model. Cognitive computing involves self-learning systems that use data mining, pattern recognition and natural language processing to mimic the way the human brain works. The goal of cognitive computing is to create automated IT systems that are capable of solving problems without human assistance.  Cognitive computing is used in numerous artificial intelligence (AI) applications, including expert systems, natural language programming, neural networks, robotics and virtual reality. The term cognitive computing is closely associated with IBM’s cognitive computer system, Watson.


Cognification is the process of making objects smarter and smarter by connecting, integrating sensors and building Artificial Intelligence software into them (both locally and Cloud-based).

AI Common Use Patterns and Categories

There are a number of general areas for the application of Artificial intelligence:

Perception: involves collecting and interpreting information to sense the world and describe it. These capabilities include Natural Language Processing, Computer Vision, and Audio Processing.

Prediction: involves using reasoning to anticipate behaviors and results. For example, to develop precisely targeted advertising for specific customers by predicting their response to offers based on an analysis of historical data.

Prescription: is principally concerned with what to do to achieve goals. It has a variety of use cases, including route planning, drug discovery, and dynamic pricing.

Integrated Solutions: AI can be combined with complementary technologies such as robotics to provide integrated solutions. These include autonomous driving, robotic surgery, and household robots that respond to stimuli.

These can be classified into the following functional categories:

AI is being leveraged in the following ways:

Natural Language Generation: Producing text from computer data. Currently used in customer service, report generation, and summarizing business intelligence insights. Vendors include: Attivio, Automated Insights, Cambridge Semantics, Digital Reasoning, Lucidworks, Narrative Science, SAS, and Yseop.

Speech Recognition: Transcribe and transform human speech into a format that is useful for computer applications. Currently used in Interactive Voice Response (IVR) systems and mobile applications. Key vendors include: NICE, Nuance Communications, OpenText, Verint Systems.

Virtual Agents: Evolving from simple chatbots to advanced systems that can network with humans. Currently used in customer service and support applications, and as a smart home manager. Vendors include: Amazon, Apple, Artificial Solutions, Assist AI, Creative Virtual, Google, IBM, IPsoft, Microsoft, and Satisfi.

Text Analytics and NLP: Natural Language Processing (NLP) uses and supports text analytics by facilitating the understanding of sentence structure and meaning, sentiment, and intent through statistical and machine learning methods. Currently used in fraud detection and security, a wide range of automated assistants, and applications for mining unstructured data. Vendors include: Basis Technology, Coveo, Expert System, Indico, Knime, Lexalytics, Linguamatics, Mindbreeze, Sinequa, Stratifyd, and Synapsify.

Decision Management: Engines that insert rules and logic into AI systems and used for initial setup/training and ongoing maintenance and tuning. A mature technology, it is used in a wide variety of enterprise applications, assisting in or performing automated decision-making. Vendors include: Advanced Systems Concepts, Informatica, Maana, Pegasystems, and UiPath.

Machine Learning platforms: Providing algorithms, APIs, development and training toolkits, data, as well as computing power to design, train, and deploy models into applications, processes, and other machines. Currently used in a wide range of enterprise applications, mostly involving prediction or classification. Vendors include: Amazon, Fractal Analytics, Google,, Microsoft, SAS, and Skytree.

Deep Learning platforms: A special type of machine learning consisting of artificial neural networks with multiple abstraction layers. Currently primarily used in pattern recognition and classification applications supported by very large data sets. Vendors include: Deep Instinct, Ersatz Labs, Fluid AI, MathWorks, Peltarion, Saffron Technology, and Sentient Technologies.

Biometrics: Enable more natural interactions between humans and machines, including but not limited to image and touch recognition, speech, and body language. Currently used primarily in market research. Vendors include: 3VR, Affectiva, Agnitio, FaceFirst, Sensory, Synqera, Tahzoo.

AI-optimized Hardware: Graphics processing units (GPU) and appliances specifically designed and architected to efficiently run AI-oriented computational jobs. Currently primarily making a difference in Deep Learning applications. Vendors include: Alluviate, Cray, Google, IBM, Intel, and Nvidia.

Robotic Process Automation: Using scripts and other methods to automate human action to support efficient business processes. Currently used where it’s too expensive or inefficient for humans to execute a task or a process. Vendors include: Advanced Systems Concepts, Automation Anywhere, Blue Prism, UiPath, and WorkFusion.

As a field of academic research, Artificial Intelligence encompasses the following branches:


Intel Cloud Insider Program

Why Join the Intel® Cloud Insider Program?

We help our members stay competitive and successful with access and insights to proven technologies, tools to maximize revenue and expand your market share, and exclusive networking opportunities.

Learn More >>

International Planning and Research

Key Partner for 40 years

For over 40 years, IPR has been the key partner successful companies rely on to make more intelligent and strategic business decisions, delivering data-driven, actionable insight.   

To find out more visit

Videos in series