After having survived a few AI winters - times when funding dried up and researchers lost interest in the complicated and sophisticated technology - artificial intelligence or AI is now having its moment in the spotlight. Right now, it is the buzzword of all business buzzwords. AI is being touted as a cure-all for just about every business problem out there, which, obviously, can't be true.
Both the hardware and the software that runs AI needed time to catch up to the technology's incredible potential, and now it finally has. AI's time has come, and businesses wishing to utilize AI and its corresponding technologies, namely machine learning and deep learning, need to understand AI is an extremely difficult technology to implement (only 1-in-3 projects succeed), and the AI vendor landscape is filled with companies promising a world where AI does almost anything while running neural nets, generative adversarial networks, long short term memory models, etc., etc., when all they're really selling is a black box containing some logistic regression and/or k-means clustering models.
AI is not the technological godsend for every problem as its champions have promised, but it is a revolutionary tool that can help businesses in a multitude of ways. It is useful in industries as far-ranging as automobile manufacturing, aviation, data centers, e-commerce, finance, gaming, insurance, manufacturing, media, retail, telco, social networks, and pretty much any customer-facing industry that wants to impress its patrons with personalization marketing.
Basically, AI can be broken down into the following five different categories, along with their subcategories:
- Voice recognition
- Voice search
- Flaw detection
- Log analysis/risk detection
- Enterprise resource planning
- Business and economic analysis
- Recommendation engine
- Sentiment analysis
- Augmented search
- Fraud detection
- Facial recognition
- Image search
- Machine vision
- Personalized advertising
- Motion detection
- Real-time threat detection
This is one the most recognizable forms of AI, even if most users of it aren't aware that the technology underpinning Amazon's Alexa, Apple's SIRI, and Google's Voice Assistant is AI and machine learning. Advances in voice recognition, distributed computing, and mobile technology have made it common for people to get their information through voice-activated technology by simply speaking into their phone, computer, or speaker.
Voice search is also audio AI and is becoming a growing area of search engine optimization (SEO), and it can provide businesses with a powerful competitive differentiator. When a mobile user asks for a list of restaurants, Google voice provides them with one -- and only one -- answer, meaning the company that gets atop the search rankings for that request gets the business. The importance of being top of the Google rankings is doubly important when a user is only given one choice, and businesses should recognize this importance because voice search may soon become the way most people search business listings, especially users on the move.
Audio AI can also be used to spot flaws in machines, specifically in aircraft engines, which can emit sounds that point to pending problems.
A time series is an arranged sequence of events or variables in a uniformly spaced time interval that can predict future behavior. There are two types, continuous and discrete. With a continuous time-series, observations are made continuously in time as they happen, while a discrete-time series takes observations at specific, usually equally spaced, times. A time-series analysis can be helpful in forecasting sales, projecting yields, and analyzing budgets.
For IT departments, AIOps (AI in Operations) can be successfully used to monitor a time series of information about properly functioning systems. Should the AIOps program spot anything out of the ordinary, adjustments to the system's processes can be made to correct the mistake. If the problem is beyond the capacity of the AIOps program, alerts can be sent to necessary parties so the issue can be fixed manually.
Chatbots, Natural Language Processing (NLP), sentiment analysis, augmented search, and language translation are all examples of text AI. Chatbots imitate human intelligence by interpreting a user's query and then attempting to answer that query, which can help with customer service questions, completing customer orders, tracking packages, as well as a whole host of other tasks that are currently being handled by humans.
AI and machine learning can also be used to spot fraud. Supervised machine learning algorithms can self-learn from text data sets and recognize patterns that fall within a normal range, then send out alerts for anything that doesn't fit the observed norm. Unsupervised machine learning, however, can cull through text data, define what are acceptable parameters in the data should be, and then, potentially, uncover things that are fraudulent or look suspicious.
One of the major text use cases for AI is sentiment analysis, which uses Natural Language Processing (NLP) to glean insight into how a business might be seen on social media. NLP extracts information from either the spoken or written word, attempting to understand the intent of a phrase or sentence. For example, NLP can be used by social media listening platforms to capture a reviewer's sentiment in a hotel or car review. This is useful information because it tends to be highly trustable information. Few businesses in the western world hire fake reviewers to slam competitor products, therefore businesses can take any criticism it receives on its products or services to heart and adjust their business accordingly.
Image recognition could be characterized as one of the foundational successes of machine learning. A subcategory of computer vision, image recognition can detect and analyze images, such as people, places, and things, to enable the automation of a specific task. Self-driving cars utilize highly sophisticated image recognition to keep a car on the road.
Rapid advancements in image recognition have improved facial recognition technology to the point where a single face can be singled out against millions of others in less than a second. Using image recognition with edge analytics, the technology compares visual data taken directly from a camera with a person's face in real-time. The obvious use case for facial recognition technology is security and facial recognition technology is evolving to point where emotions can be understood as well. This can help not only identify customers in a store but also understand their emotions, which certainly has uses in advertising.
Visual search is the ability to use an image to search for another similar visual asset and it is now being embraced by retailers worldwide because it can simplify the buying process. With visual search, a buyer doesn't need to try and guess a brand, a style, or a shop where a product came from. A potential buyer can simply snap a picture of an item, upload it to the Google image search engine or a retailer's website, and immediately find a match to the searched-for item. All the friction of the normal buying process is gone.
Machine vision utilizes image recognition to automatically inspect and analyze applications and processing, including automatic inspection, sorting, process control, and robotic guidance. The technology can be used on assembly lines, robots can spot flaws in products, or even pick things up in a pick and place way. Machine vision can be used in a wide range of industries, including automotive, broadcasting, food and beverage, manufacturing, smart city, medical imaging, and printing, and pharmaceuticals.
Although not fully integrated yet, the future of advertising could be with videos augmented by AI. With personalization becoming an omnipresent word in advertising, AI could be the newest addition to this customization movement. Today, creating a digital video ad is a long, laborious, and involving process, a process that could be simplified by AI. Multiple iterations of an advertisement could be shot at once. For example, in one, a twenty-something Asian actor plays the lead, in another, the lead is a Caucasian woman in her forties. Once the advertisement is shot, the editor goes to work, but it's not a human editor, it's AI, cutting the ad into multiple versions, each with its own set of unique actors for different demographics. Once the ad airs, AI can be performed to instantly analyze viewer behavior, an analysis which can also be used to personalize future company commercials.
AI has had a long and winding road to success, and it's definitely been worth the wait. For a technology that promises so much, it's understandable that the demands put on the hardware and software that runs the system would be quite extreme. Unquestionably, AI can radically revolutionize business. From website morphing, customer and media recommendations, purchase prediction, demand forecasting, programmatic advertising, as well as social listening, the potential use cases are almost endless. However, it's still important to separate AI fact from fiction. As with any revolutionary technology, those who walk in with their eyes open won't end up in a technological nightmare of their own making.
If you happen to be in the market for a software development partner you can count on to tackle complex solutions, Envative is the leading expert in all development areas of web, mobile and IoT technologies. Contact us today to see how easy we are to work with.
Tagged as: A.I., Tech Use-Cases
About the Author:
Marc Mastrella is Business Relationship Manager at Envative. He regularly engages with potential clients to discuss how software can solve real-life problems within organizations. He connects those pursuing a software solution for their business or looking to bring a mobile app/IoT idea to life with the talented developers at Envative for brainstorming and consultation. Marc sees first-hand what a difference the right technology can do for a business and does all he can to help make the process of getting started easy.