A Complete Introduction to AI and Machine Learning
What is Artificial Intelligence?
Artificial intelligence (AI) is a type of computer science that has become hugely popular in the business world in recent years. But as this introduction to AI will show, the term’s modern usage is very different from its philosophical origins.
Alan Turing is credited with giving rise to the field of AI with his 1950 paper, Computing Machinery and Intelligence. In it, the Enigma code-cracking mathematician asks: “Can machines think?”
The term ‘artificial intelligence’ can be traced back to a 1956 conference at Dartmouth College in New Hampshire, where the famous “Turing Test” was proposed as a way to determine if a machine really is thinking.
Today, AI is more commonly used as a ‘catch all’ term for software designed to perform complex tasks that humans find challenging. AI programs can automatically complete these tasks more efficiently than humans can, helping businesses to operate more efficiently.
Artificial Intelligence Definition
A computer system has artificial intelligence if it thinks like a human. But as things stand, no computer exists that can convincingly simulate an entire human consciousness. Instead, AI software is usually designed to perform one narrowly defined ‘human like’ task extremely well.
AI-driven tools range in complexity from programs containing groups of ‘if-then’ statements to complex statistical models designed to collect, integrate or analyze large volumes of data.
These models may be designed to capture information, execute rules, find patterns, spot atypical datapoints, correct errors, make rational deductions or inferences and more besides.
Read on to discover more about:
- How artificial intelligence works
- Machine learning deployment
- How AI and machine learning are used today
- Machine learning and AI conferences
[3d-flip-book mode="fullscreen" urlparam="fb3d-page" id="8904" title="false" template="black-book-view"]
How Artificial Intelligence Works
There are many different types of AI technology, including machine learning, deep learning, natural language processing, computer vision, robotic process automation and reinforcement learning.
Because the range of things that can be called ‘AI’ is so broad, the best way to understand how artificial intelligence works is to look at some examples of the most common types of AI.
Machine Learning Definition
You can think of deep learning, machine learning and AI as a set of Russian dolls nested within each other. Deep learning is a subset of machine learning, and machine learning is a subset of AI.
Machine learning algorithms are statistical models that can alter themselves. A data scientist will define the algorithm’s initial state and give it access to data. The program will then use this ‘training data’ to make generalizations that can be applied to new datapoints.
The fundamental goal of machine learning is to create programs that can successfully process or interpret new data they’ve never seen before.
How Machine Learning Works
Types of machine learning can be grouped by learning style. For example, ‘supervised learning’ algorithms are taught using training examples that pair ‘input objects’ with desired ‘output values’. The algorithm learns to perform the illustrated transformation on any data that’s input into the system.
Computer vision technology may use this approach to analyze pictures and summarize their contents in natural language.
‘Unsupervised learning’ algorithms use the same ‘input data’, but with no corresponding output values. Instead of being taught, they are left to their own devices to find interesting patterns in data.
‘Principal component analysis’, ‘cluster analysis’ and ‘reinforcement learning’ are two common unsupervised machine learning techniques.
Cluster analysis involves teaching algorithms to group or segment datapoints based on their characteristics. Common uses for this technique include for pattern recognition, image analysis and data mining.
Meanwhile, principal component analysis involves transforming complex, multivariant datasets to reduce the number of attributes they contain, removing redundant variables and making them easier to interpret.
Examples of machine learning’s many applications include data compression, image processing, data visualization, exploratory data analysis, pattern recognition and time series forecasting.
Machine learning algorithms can equally be grouped according to their functions. Common machine learning functions include statistical classification, regression analysis, decision tree learning and deep learning.
So, What is Deep Learning?
Deep Learning is a subfield of machine learning. Deep learning algorithms use artificial neural networks that are loosely inspired by the structure and function of the human brain.
Just like an organic brain, neural networks consist of many interconnected processor units organized into layers. Each of these layers and processing units is then assigned a ‘weighting’ that determines its effect on the network’s overall output.
When a neural network receives an input, it guesses what the pattern might be. Then, it sees how far its answer was from the correct one and makes an appropriate adjustment to its connection weights. The idea is that these adjustments make it more accurate over time.
While they are far simpler than a human brain, they neural networks are still incredibly complex. This has contributed to the idea that AI algorithms are unexplainable ‘black boxes’.
Machine Learning Deployment
The question of how best to deploy AI or machine learning models across an organization is a tricky one.
It’s only once a model has been made available in a production environment that it can start adding value to a business. But besides the challenges that come with deploying any new code or IT systems, there is a wide range of challenges that are unique to machine learning deployment.
For example, development teams must ensure that their machine learning models make reproduceable and unbiased predictions or decisions in order to deploy them successfully.
Automating model training and publishing, planning for future model updates, creating reusable code modules and setting aside ample time for testing can also help developers ensure the deployment of new machine learning models goes smoothly.
How AI and Machine Learning is Used Today
Businesses have been developing AI capabilities for many years. But the volume of applications for AI technologies that are being brought to market has grown exponentially in recent years.
Today, companies in virtually every industry are using AI to streamline business processes, generate valuable insights, improve decision-making processes or create new revenue streams.
As enterprises respond to disruptive forces in their industries, the use of AI in business will rapidly become mainstream in the years ahead.
AI in Retail
A wealth of accurate customer data has empowered retail companies pursue advanced personalization strategies. They use AI to provide tailored experiences based on consumers’ unique wants and needs.
“The market is changing quite rapidly,” says Juwel Rana, Head of Analytics at fashion retailer Varner. “To serve our customers’ best interests we need to strengthen our knowledge, our capacities, our business processes. That’s why data and AI are so important for us.”
Recommendation engines were among the first AI technologies ecommerce data leaders used to enhance customer experiences. These algorithms use techniques such as audience segmentation and market basket analysis to recommend the most relevant products possible to each customer.
Personalization initiatives have historically focused on customer interactions online. But as retailers begin to experiment with AI-driven technologies such as smart mirrors to digitize in-store experiences, the boundaries between on- and offline shopping will start to blur.
At the same time, AI-driven tools are helping the industry’s customer service professionals. For example, chatbots and robotic process automation technologies are automating simple customer interactions, freeing agents up to focus on more complex queries.
Meanwhile, AI is transforming supply chain management. Demand forecasting, warehouse optimization and pricing strategies can all benefit from insights generated through AI-driven predictive analysis.
AI in Financial Services
Financial services institutions were among the first to embrace data-driven technologies in the wake of the global financial crisis. Today, they use AI for everything from risk management to fraud detection, customer experience optimization and beyond.
“We’ve designed algorithms to try to offer the right product to the right customer at the right time,” explains FNB South Africa CAO, Chief Risk Office Mark Nasila. “We enhance this with financial crime models to make sure we’re not offering products to criminals.”
“The bank is now using an in-house developed AI system to optimize the due diligence forensic review process,” he adds. “This AI system automatically creates a single consolidated report with all the information required, which includes a single view for financial crime risk management.”
Of course, the financial services industry encompasses a wide range of organization types, each with its own level of AI maturity.
Sectors that deal primarily with high volume, low value transactions have generally been quicker to adopt AI. But today, enterprises in all corners of the industry are embracing these new technologies.
“People were aware some years ago that artificial intelligence is coming,” says Jose A Murillo, CAO at banking group Banorte. “Now, people have already incorporated these technologies into their businesses and the early adopters are getting the advantage versus those who are lagging.”
AI in Healthcare
The healthcare industry has been relatively slow to adopt data-driven technologies. But AI will play a key role in making it possible to provide patients with ultra-personalized healthcare services.
“The next big thing will be making data available,” says Morgan Templar, VP, Data Management at HM Health Solutions. “We will use artificial intelligence and we will use machine learning to cut through the mass of data and we’ll be able to show better outcomes.”
“Artificial intelligence applications are being baked into Electronic Health Records,” adds Besa Bauta, Chief Data Officer at child welfare agency MercyFirst. “You’re going to see much more development in that arena, both prescriptive types of applications and decision aids for physicians.”
At the same time, AI technologies have the potential to drive efficiencies and improve outcomes in patient care.
For example, GE Healthcare has partnered with the University of Oxford and a group of NHS Trusts to develop AI algorithms to enhance the image quality of PET scans and automatically flag key findings in critical care X-ray scans.
AI in Insurance
As our 2020 Future of Insurance Data report shows, AI is coming to the disrupt the insurance industry. From Ping An in China to Lemonade in the US, companies across the globe are harnessing AI technologies to drag the sector into the 21st century.
The success of these AI-driven start-ups has served as a wakeup call to established insurers, who have been quick to develop their own capabilities.
“By the end of [2020], AI and machine learning will take a major part of the quoting system,” says Alan Luu, AVP of Advanced Digital Analytics at Chubb. “All the major insurance companies will either be there or must catch up – otherwise, they'll be way behind.”
While AI is already used by some insurers to handle fraud detection when customers make claims, Luu anticipates that the greatest impact of AI will be on the underwriting process itself.
AI in Recruitment
It’s undeniable that companies must recruit the right talent to succeed. Yet, recruiters have historically depended on their instincts and experience to navigate the jobs marketplace. AI promises to change that by transforming the recruitment process into an exact science.
Automated candidate sourcing and matching algorithms can help recruiters automate the laborious process of vetting potential job candidates.
Meanwhile, AI can be used to track how prospects behave on recruitment websites and automatically target them with content and offers tailored to their specific interests or needs.
What’s more, passive talent identification AIs could soon help companies find the best talent in their workforce’s network and automatically engage the best employees to make the referrals.
Crucially, AI also has the potential to combat biases in the hiring process through anonymizing candidates or even removing humans from the equation in the early stages of recruitment. This could prove useful for ensuring companies hire the best candidate for any given job.
Machine Learning and AI Conferences
With enterprises across the globe racing to rollout AI and machine learning capabilities, these topics now feature prominently at virtually every major conference for data and analytics leaders. Click here to see the full list of Corinium’s upcoming global events.