Artificial Intelligence Versus Machine Learning and Top Practical Applications

AI versus ML

There is not a universally accepted definition for artificial intelligence. So there is a lot of confusion and misconception over the meaning of AI. MIT offers this succinct definition for AI: “Machines acting in ways that seem intelligent”. The meaning of artificial intelligence can be broken down into two key areas of study: Artificial General Intelligence (AGI) and Artificial Narrow Intelligence (ANI). More on ANI shortly.

AGI, also know as Artificial Super Intelligence is that thing Elon Musk warns us about. Artificial General Intelligence is an artificial intelligence that can successfully perform any intellectual task that a human being can, including learning, planning and decision-making under uncertainty, communicating in natural language, making jokes, manipulating people, trading stocks, or… reprogramming itself. We may be 50 years away from AGI. Back to the reality of today.

Artificial Narrow Intelligence (ANI) can be described as machines acting in ways that seem intelligent, which can effectively perform a narrowly defined task. Which brings us to Machine Learning (ML), the actual AI that can be applied today. ML is a special algorithm that learns on its own. Machine Learning enables computers to identify patterns in observed data, build models that explain the world, and predict things without having explicit pre-programmed rules and models. Today ML is being applied across industries on narrowly defined use cases, driving big time value.

For more terminology check out this helpful resource.

Practical Applications of Machine Learning

We at KUNGFU.AI believe in practical AI. Meaning successful adoption starts with the business problem and applies the simplest, most creative application of artificial intelligence. It is a point solution applied to a narrow use cases. We’ve observed patterns emerging from companies who are publicly discussing their AI initiatives. 

benefits for artificial intelligence

These companies are graduating their machine learning models from R&D and creating solutions that augment their intelligence and capabilities. Once AI goes from theoretical to actually providing value to people, it becomes practical. We thought it would be helpful to share some of the top practical applications for artificial intelligence. These use cases are general and apply towards many (if not all) industries.

Generally, companies are reporting many benefits resulting from AI programs. The use cases resulting in these benefits vary. Below are several common use cases for machine learning being applied across industries:

Identifying Objects in Images and Videos

Computer vision, an AI technology that allows computers to understand and label images (and video objects), is now used in convenience stores, driverless car testing, daily medical diagnostics, and in monitoring the health of crops and livestock.

Using computer vision technology, Amazon Go Stores have cameras with the capability to determine when an object is taken from a shelf and who has taken it. If an item is returned to the shelf, the system is also able to remove that item from a customer’s virtual basket. The network of cameras allows the app to track people in the store at all times, ensuring it bills the right items to the right shopper when they walk out, without having to use facial recognition.

Extracting Written or Spoken Text from Images, Audio, or Video

NLP is used by computers to manipulate human language, whether to extract meaning, generate text, or for any other purpose. The interaction computer-language is categorized according to the task that needs to be accomplished: summarizing a long document, translating between two human languages, or detecting spam email are all examples of tasks that today can be decently accomplished by a machine.

A telco company may want to know under what circumstances different telecom competitors are mentioned in phone conversations. For example, are competitors typically mentioned during refund requests, billings issues, or service requests? When are these competitors being mentioned as a comparison on service and price, as a casual mention, or as a threat of defection?

Much of the worlds information is held on paper and PDFs, or are simply scans of physical documents. You can also apply machine learning to extract data from flat files. For documents that are not machine readable — like those that are scanned as PDFs — optical character recognition (OCR) is the key means for text recognition and is the conversion of characters in a digital image to digital text. Once unlocked and machine readable, there are a lot of things that can be done with documents using what’s called text mining or text analytics. For example, we build an OCR model for Real Estate Company, Keller Williams, to help them extract data from offer contracts to make predictions on what offers will ensure a sale.

This is the same application of machine learning enables you to take a picture of your check and make a deposit.

Identifying Categories or Segments in User Behavior

Identifying the right customer at the right time keeps every business up at night. There are many ways to solve this problem. Machine learning is proving to be a more effective tool for identifying the ideal customer. For example:

Psychographic segmentation involves dividing your market into segments based upon different personality traits, values, attitudes, interests, and lifestyles of consumers. The capability goes even further to include predicting personality traits, aptitudes, and abilities in what is called psychometric profiling. This data traditionally could only be found through specifically designed tests and questionnaires but, through machine learning, has now been correlated to data that can predict things like how neurotic you are, how open you are to new experiences or whether you are contentious. Today, Twitter is an open data source advertisers leverage to understand individuals motivations and desires. This data can then be used to infer additional unknown, information about you like your gender, political affiliation, your current emotional state and how reliable you are. They can even be used to leverage AI, making predictions on what you’re likely going to buy next.

Psychographic models

Propensity models use machine learning techniques and historical data, such as bookings, orders, size of the company, number of employees, and their likelihood to buy from you. This allows the salesperson (or marketer) to focus on only the subset of accounts that are likely to make a purchase. Without this model, the company will use its own business rules.

For example, we are working with a beverage company to create psychographic profiles to identify behaviors of a new market segment. We will use that data to develop a propensity model that predicts which sponsorship opportunities will make the greatest impact.

Detecting Anomalies or Erratic Behavior in a System

Large-scale machine learning technology can collect huge amounts of data relating to fraudulent activities worldwide and analyze it instantly. Thousands of traces left behind by fraudsters, which might otherwise have remained unconnected in the vast ocean of data, are now linked to produce a clear predictor of fraud threats. As the AI crunches through all this data, it is able to detect anomalies and indicate probable incidences of abuse. 

For example, many sites now have algorithms designed to electronically screen for false reviews, and Amazon recently sued more than 1,100 creators of fakes.

Chatbots for Automated Conversations

One of the most common use cases for AI today are chat bots. A chat bot is a computer program designed to simulate conversation with human users, especially over the Internet. Chat bots, a form of natural language processing (NLP) and natural language generation, are used across the web to automate sales and customer service engagements. Some even use chatbots to augment and inform human service agents.

Chat bots are becoming popular for customer support. IBM estimates that 265 billion customer support tickets and calls are made globally every year, resulting in $1.3 trillion in customer service costs. For example, Digital Genius offers a software which they claim can help businesses such as travel agencies both automate their repetitive customer support inquiries and assist human agents with answering customer questions using NLP. Essentially, Digital Genius offers a chatbot that allows for greater user input, a service they call “Co-Pilot.” Digital Genius primarily integrates with Salesforce and Zendesk.

Accessing User Emotions, Opinions, and Attitudes

Machine learning is being leveraged to understand peoples emotional state using a technique called sentiment analysis. the process of computationally identifying and categorizing opinions expressed in a piece of text, especially in order to determine whether the writer’s attitude towards a particular topic, product, etc., is positive, negative, or neutral.

Sentiment analysis is a process for answering the question: “How do they feel?” There is sentiment that expresses anger, disappointment, joy, anxiety, and more – and these feelings are important for understanding customers.

The airline may use natural language processing to detect a higher frequency of positive sentiment about flight delays, despite delay frequencies that are about equal to its other flights. The company could investigate whether this apparent reduction in complaints is due to sparse data, or due to better customer service or improved satisfaction at the gates – resulting in noticeably happier customers.

Personalized Information, Products, and Service Recommendations

There’s been a lot of advancement in ML around product recommendations. We now have more data on customer and products with more compute power to process it all. We can now collect and process petabytes of data and append mass-consumer purchase behavior to an individual’s purchase history to offer relevant and helpful product recommendations.

For example, Amazon’s uses machine learning to drive product recommendations. They use a combination of Collaborative Filtering and Next-in-Sequence models to make predictions on goods an individual consumer may need next. Amazon possesses a massive database of consumer purchase behavior to power its predictions.

Product Recommendation Systems

Video Analytics and Event Detection

Companies are using a Deep Learning technique, computer vision, to better understand data inside video files and feeds. For example, the HomeCourt app take photos and video of basketball players, processing the data in real-time to measure performance accuracy and movement. Evolv Technology claims to offer a physical security system for facial recognition, allowing 600 to 900 people to walk through per hour, identifying at least one person per second. Their technology could be used at airports, major events, schools, commercial and business buildings, government offices, and other public places.

computer vision and video analytics in sports

Uber’s ETA for Rides and UberEats

Rideshare apps used machine learning to dethrone the incumbent cab industry seemingly overnight. ML enables Ridesharers to optimize pricing, minimize wait time, optimize passenger pick up, and find the optimal route.

For example, Engineering Lead for Uber ATC Jeff Schneider  discussed in an NPR interview how the company uses ML to predict rider demand to ensure that “surge pricing” (short periods of sharp price increases to decrease rider demand and increase driver supply) will soon no longer be necessary. Uber’s Head of Machine Learning Danny Lange confirmed Uber’s use of machine learning for ETAs for rides, estimated meal delivery times on UberEATS, computing optimal pickup locations, as well as for fraud detection.

predictive delivery

Compliance Automation

In large companies which deal with several thousands of contracts everyday, human contract experts might not be a feasibly scalable option. It follows that AI would find its way into lease and contract abstraction applications for finance and law.

For example, KPMG claims users can upload contract images or PDF files into their software. The software uses OCR to convert the input documents into a machine-readable format. The contracts are then automatically segmented into different categories based on their content by IBM Watson’s algorithm. Financial experts at KPMG then review the extracted and segmented contracts to correct any errors in the process, training the software to learn how to extract information more accurately. The system then provides a file with the extracted information from contracts in a structured format, which can then be uploaded into a lease management software.

Content Personalization

Predictive content personalization, also referred to as machine-learning personalization, is the more advanced and AI-driven way to dynamically display the most relevant content to each user.

Unlike the rules-based method, it does not target whole segments; instead, users are identified at a more granular level, and a more personalized website experience is created for them. It puts more focus on displaying content and messages to users based on their intent, rather than just on the readily available information about their interests and previous behavior.

With AI, marketers can vastly expand the data used to get individual personalization right, including critical customer variables like real-time location, context, behavior and values. These variables make the difference in what you offer and when.

Location, for example, can help you target customers in a specific geographical area. Context can affect your messaging based on factors like the season or weather, or whether a customer is traveling with family when they normally travel alone for business. Understanding values, like whether a customer is vegetarian, can ensure you avoid making offensive offers. 

Voice Assistants

Voice assistants like Siri, Alexa, Google Now, and Cortana are the new frontier of intelligent augmentation. These AI tools may help you do everything from book a meeting to order a pizza. Voice assistants use natural language understanding and processing to understand the speakers intent and automate performance of tasks. 

how chat bots work

Pricing and Promotions

Traditionally, retailers run promotional efforts around local events or holidays. And typically, these promotions are repeated annually (with little variance to the promotion). Retailers often lack the complete picture when it comes to understanding how effective promotional campaigns are, or if the promotion of certain products might only be ultimately reducing overall profits by causing sales of related category products to drop.

One approach is centered around the reinforcement learning (RL) branch of AI. After aggregating past promotions, market factors, sales, and manufacturing data, machine learning can run billions of simulations and, through trial and error, determines the products and prices that are most likely to deliver the maximum return. Sequences of prices and products are forecast in a huge number of scenarios – just like AlphaGo might look 10 or 100 moves ahead in a Chess game to predict possible outcomes.

Trend Sensing

Deep Learning algorithms to give retailers the benefit of collecting and analyzing customer data points — specific keywords, online navigation patterns, price points on their shopping cart, actual purchases, and their “likes,” among others — to determine what users are searching for, and anticipate the next trend.

For example, Intelligence Node is reported to be able to track exact and close matches to their product, potentially giving retailers the opportunity to recommend and promote items that match emerging trends.

Create Unique Artworks in the Style of Famous Artists

This past October, the first AI generated art went to auction. Christie’s sold the painting created by artificial intelligence for $432,500.

Artificial intelligence can produce a legitimately classical-looking portrait, using a model called a GAN (generative adversarial network). By feeding an algorithm a training data set tens of thousands of portraits created between the 14th and 20th centuries. Using these images, the algorithm was able to “generate” new images similar to the ones it had been fed. These new portraits were then presented to another algorithm (the “adversarial” part of the GAN acronym) that was trained to distinguish between images produced by humans versus those by machines—a Turing-like test for artworks—until the generated portraits could fool this discriminator into thinking they were “real,” too.

AI generated art

This is a non-exhaustive list of practical AI use cases. However the ones selected above were chosen because they transfer to other industries and data sets. For many, AI is a question of where to start. To this end, we have collaborated with Voltage Control to develop a canvas for AI adoption. If you don’t know where to start, it’s helps develop strategies and identity use cases for AI. When you are ready to explore finding applications of machine learning in your world, start with the problem first. Hone in on your biggest problem with the most data available. End with selecting the application for machine learning to maximize likelihood for success.