STATE OF THE FIELD
According to Chaudhary, it's estimated that two million children in the United Kingdom use AI in the form of smart speakers. They know how to use smart speakers to listen to music, inquire about the weather, ask it to tell them a joke or a story, ask it questions, and receive help with homework assignments (2019). China is striving to be number one in the AI field. They are developing AI for use in the private and military sectors in order to be the most powerful in the world (Demchak, 2019).
AI research has been sponsored by private and government sectors in the past, but most recently has been sponsored by a non-profit named OpenAI. The OpenAI initiative is backed by Elon Musk and has committed $1 billion to develop AI to benefit all humanity. "Tractica has estimated that the market for enterprise applications of AI will exceed $30 billion by 2025, with a particular focus on increasing the accuracy of analyzing big data and Bank of America has predicted that the market could blossom to $70bn for AI systems (along with $83bn for robotics) by 2020." (De Spiegeleire, Maas, & Sweijs, 2017) According to Lord (2018), Gartner said that AI would generate $1.2 trillion in business worldwide in 2018, which is a 70 percent increase over 2017. According to Gillath (2020), Grand View Research estimates the worldwide market for AI to be $39.9 billion in 2019, with a projected expansion on 42.2 percent from 2020 to 2027.
MACHINE LEARNING (ML)
Machine learning uses prediction more than discovery when dealing with data analysis and processing. ML studies the data in order to build and improve on the rules that the system uses to solve problems by applying algorithms (Kim, 2019b; González García, Núñez Valdéz, García Díaz, Pelayo García-Bustelo, & Cueva Lovelle, 2019). Data is what keeps the ML engine running. If there's no data to analyze and process, then the process becomes idle (De Spiegeleire, Maas,& Sweijs, 2017).
Within ML, there are several algorithms used depending on what feedback is given. Those algorithms are supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, transduction learning, and multitaksing learning. Supervised learning uses what has been learned before and applies it to new data to predict what will happen in the future. Unsupervised learning uses the data to infer the outcome. Semi-supervised learning uses both supervised and unsupervised learning to produce results. Reinforcement learning that uses the environment to produce results and relies on feedback. Transduction learning is when reasoning comes from specific cases and is then applied to specific cases. Multi-task learning solves several tasks at the same time while finding the similarities and differences in them (González García, Núñez Valdéz, García Díaz, Pelayo García-Bustelo, & Cueva Lovelle, 2019).
ML can be used to analyze data to find patterns, predict trends, automate tasks, and autonomously control systems (De Spiegeleire, Maas, & Sweijs, 2017). Some examples include SPAM filters, optical character recognition (OCR), search engines, computer vision, software testing, game playing, derivatives training, self-driving cars, fraud detection, credit scoring, automated video captioning, and data mining (González García, Núñez Valdéz, García Díaz, Pelayo García-Bustelo, & Cueva Lovelle, 2019; De Spiegeleire, Maas, & Sweijs, 2017; Sosnovshchenko, 2018).
DEEP LEARNING
Deep learning falls under ML and uses multi-layered networks to filter data through algorithms to determine the best action (De Spiegeleire, Maas,& Sweijs, 2017). Deep learning methods are used in facial recognition, language translation, and speech recognition (Kim, 2019b).
AI can take on many forms. It can be used for cyber security, virtual assistants, chat bots, video and image production, gaming, email filtering, browser searches, language translation, facial recognition, and healthcare. Click on the button to see what major companies and applications use AI in their technology in some form.