PwC estimates AI will add $15.7 trillion to the global economy by 2030. Seeing the writing on the wall, over 90% of the Fortune 1000 are increasing their big data and AI spending. If you are a country, company, or individual hoping to stay technologically relevant through the coming decade, understanding AI is imperative.
The good news is that you don’t need to build neural networks from the ground up to take part. Being able to differentiate core concepts and understanding how AI is used in real business cases (and how you can use it too) are half the battle. That’s where this article will take you.
We use “AI” as a loose term indicating the ecosystem, that includes applications in computer vision, natural language processing, speech recognition, predictive analytics and other practices. Machine learning, including deep learning are the technologies underpinning these applications. First let’s take a quick look of the layman’s term of some of the popular vocabulary, feel free to skip if you are already familiar.
Definitions used within AI and data science
Machine learning is the use of algorithms to perform AI functions, often with labelled data to make numeric or grouping predictions with rules. Deep learning is one specific type of machine learning utilizing many layers of neural networks. For example, feeding a deep learning algorithm 1000 labelled pictures of people allows security cameras to predict that a camera feed includes a person to prompt an intruder alert.
Natural language processing (NLP) is the application of ML to interpret human language. For example, feeding a NLP algorithm tweets to identify sentiment with positive and negative sentence structures and words. Machine learning are used in NLP tasks.
Computer vision is the application of machine learning to interpret images or video streams. For example the 1000 labelled pictures of people mentioned above is an example of using machine learning to achieve computer vision.
1) Google Search post-2015
(deep learning + natural language processing)
Google struggled with the ~15% of inquiries that it had never received before. In order to reduce the number of times users thought “how do I word this so Google understands,” natural language processing and machine learning were enlisted.
AI enhancement: Google search breaks down each sentence mapping relationships between words to draw on similar sentences in which the model is familiar. Tech news publisher CIO provides the example of post-2015 Google re-wording the user input "What's the title of the consumer at the highest level of a food chain?" to the more easily understood "top level of the food chain" before listing search results.
2) Netflix Recommendation Engine post-2012
(deep learning + natural language processing + computer vision)
Netflix used a combination of IMBD scores and self reported user data to curate content recommendations.
AI enhancement: Netflix is able to tear apart a piece of content identifying female protagonist, action level, tone, language, setting etc. These features are turned into key words which are mapped to user preferences.
3) Telefonica Satellite Business Development (machine learning + computer vision)
Telecom companies like Telefonica focus on urban areas with high population densities and adequate infrastructure to install necessary telecom equipment. Rural areas do not have these qualities and suffer.
AI enhancement: Telefonica is able to use satellite imaging and computer vision to identify underserved areas (adequate population density) and map logistic routes to these areas. This has expanded the geography that is economically viable for Telefonica to service.
4) Verizon Quality Assurance post-2017 (machine learning)
Before AI/machine learning it was near impossible to predict poor service for Verizon’s 150+ million customers. Verizon had no choice but to work reactively to fix problems, reliant on customer complaints.
AI enhancement: Verizon is now able to predict poor service quality with a number of computed real time indicators including weather, hyper-active use, and regular tests to determine “normal” service for a particular locale.
5) Unilever AI recruiting (machine learning)
Unilever used a combination of resume, references, and interviews to evaluate potential hires.
AI enhancement: Candidates were asked to play a series of 12 games, and submit their resume online in an initial screening process. Machine learning was used to break up keywords on resumes and scores from the games to identify desirable recruit qualities. For example, one game rewarded applicants the more they blew up a balloon, and docked them for popping the balloon. This game was used to measure risk tolerance, and together mapped with hundreds of other qualities assessed in the candidate enabled a comprehensive screen for Unilever recruiters. Unilever reports their AI hiring tool saves them 70,000 man hours each year.
Thanks to AI Google’s users search in natural language, Netflix users experience their preferences without needing to understand them, new geographic regions are viable for Telefonica, Verizon’s customers receive automatic support, and Unilever job applicants are matched to their perfect role.
The logical next question is, what can AI do for your business? If you are a CEO or CTO looking to take the next step, book a meeting with us today.